# customer support

Published articles for customer support.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Reliability is a user-specific property of large language models

DevFeed: [Reliability is a user-specific property of large language models](<https://devfeed.tech/articles/reliability-is-a-user-specific-property-of-large-language-models-142194.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23732>)

Author: Xingyu Zhao

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [LLM security](<https://devfeed.tech/topics/llm-security.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Uncertainty quantification LLMs](<https://devfeed.tech/topics/uncertainty-quantification-llms.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [capability](<https://devfeed.tech/tags/capability.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [failure](<https://devfeed.tech/tags/failure.md>), [inference](<https://devfeed.tech/tags/inference.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [llm](<https://devfeed.tech/tags/llm.md>), [modelling](<https://devfeed.tech/tags/modelling.md>), [operational](<https://devfeed.tech/tags/operational.md>), [probabilistic](<https://devfeed.tech/tags/probabilistic.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [requirements](<https://devfeed.tech/tags/requirements.md>), [research](<https://devfeed.tech/tags/research.md>), [roadmap](<https://devfeed.tech/tags/roadmap.md>)

### AI overview

The abstract argues that LLM reliability should be claimed for specified users and operating conditions, because capability rankings alone do not establish how often a service will fail in future use. It distinguishes fault-finding tests from reliability assessment and proposes six research workstreams: specifying claims, modelling changing operational profiles, handling graded outcomes and noisy judgements, measuring local probabilistic robustness, accounting for interaction dependence, and applying statistical inference with reassessment. Published findings and a customer-support example motivate the roadmap, but integrated validation remains open. Reliability claims are presented as complements to separate safety and security arguments.

### Source excerpt

arXiv:2611.23732v1 Announce Type: new Abstract: Public benchmarks support comparisons of large language model (LLM) capabilities, yet users can experience the same highly ranked model differently. Drawing on traditional software reliability, we argue that the reliability of an LLM service should be assessed for specified users and conditions of use. This requires more than a capability score: evaluation evidence must support a claim about failure during future operation. We distinguish testing to discover faults from testing to assess reliability and identify ten modelling considerations behind such claims. We propose a six-workstream research roadmap linking claim specification, dynamic operational profiles, graded outcomes and noisy judgements, local probabilistic robustness, interaction dependence, and statistical inference with reassessment. Local probabilistic robustness addresses the behavioural variation hidden when different concrete inputs are treated as the same task. Published findings and an illustrative customer-support example motivate the roadmap; its integrated validation remains an open research task. Reliability claims complement, but do not replace, separate safety and security arguments. Capability scores should therefore be complemented by versioned operational claims that state whose requirements are covered, under which conditions, and with what uncertainty.

## Fivetran Adds Staff and Cloud Region in Singapore | Blog | Fivetran

DevFeed: [Fivetran Adds Staff and Cloud Region in Singapore | Blog | Fivetran](<https://devfeed.tech/articles/fivetran-adds-staff-and-cloud-region-in-singapore-blog-fivetran-140255.md>)

Original publisher: [Read original article](<https://wf.fivetran.com/blog/singapore-expansion>)

Author: T.J. Chandler

Published: 2026-11-18T02:45:59Z

Content type: news

Language: en

Sources: [Fivetran Blog](<https://devfeed.tech/sources/fivetran-blog.md>)

Topics: [Cloud](<https://devfeed.tech/topics/cloud.md>), [cloud-infrastructure](<https://devfeed.tech/topics/cloud-infrastructure.md>), [Cloud Dataflow](<https://devfeed.tech/topics/cloud-dataflow.md>)

Tags: [2020](<https://devfeed.tech/tags/2020.md>), [act](<https://devfeed.tech/tags/act.md>), [allows](<https://devfeed.tech/tags/allows.md>), [also](<https://devfeed.tech/tags/also.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [announce](<https://devfeed.tech/tags/announce.md>), [apac](<https://devfeed.tech/tags/apac.md>), [api](<https://devfeed.tech/tags/api.md>), [asia-pacific](<https://devfeed.tech/tags/asia-pacific.md>), [blog](<https://devfeed.tech/tags/blog.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [fivetran](<https://devfeed.tech/tags/fivetran.md>), [product](<https://devfeed.tech/tags/product.md>), [region](<https://devfeed.tech/tags/region.md>), [singapore](<https://devfeed.tech/tags/singapore.md>), [staff](<https://devfeed.tech/tags/staff.md>), [support](<https://devfeed.tech/tags/support.md>)

### AI overview

Fivetran opened a cloud region and formed a customer-facing team in Singapore to support local customers and partners, extend global support coverage, and serve customers needing locally delivered data pipelines. The region joins existing infrastructure in the US, Canada, the EU, UK and Australia, and the company says it can facilitate compliance with Singapore's PDPA. The team covers sales, marketing, partnerships and customer service. The excerpt also describes prebuilt connectors that adapt to API and schema changes, automated in-warehouse transformations through dbt packages, and consumption-based pricing with uptime and data-delivery SLAs.

### Source excerpt

We're continuing to deepen our product offering and improve customer support in the Asia-Pacific region.

## Securing Your Agents: A Hands-On Guide to Agent Anomaly Detection

DevFeed: [Securing Your Agents: A Hands-On Guide to Agent Anomaly Detection](<https://devfeed.tech/articles/securing-your-agents-a-hands-on-guide-to-agent-anomaly-detection-136557.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/securing-your-agents-a-hands-on-guide-to-agent-anomaly-detection/>)

Author: Achuth Narayan Rajagopal; Derek Luan

Published: 2026-11-17T05:43:32.994674Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [agent observability](<https://devfeed.tech/topics/agent-observability.md>), [threat detection](<https://devfeed.tech/topics/threat-detection.md>), [Cloud Monitoring](<https://devfeed.tech/topics/cloud-monitoring.md>), [ai observability](<https://devfeed.tech/topics/ai-observability.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Detection engineering](<https://devfeed.tech/topics/detection-engineering.md>), [alert triage](<https://devfeed.tech/topics/alert-triage.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>)

Tags: [adk](<https://devfeed.tech/tags/adk.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [alerts](<https://devfeed.tech/tags/alerts.md>), [anomaly-detection](<https://devfeed.tech/tags/anomaly-detection.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-monitoring](<https://devfeed.tech/tags/cloud-monitoring.md>), [configure](<https://devfeed.tech/tags/configure.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [detection](<https://devfeed.tech/tags/detection.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [guide](<https://devfeed.tech/tags/guide.md>), [monitor](<https://devfeed.tech/tags/monitor.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [platform](<https://devfeed.tech/tags/platform.md>), [sample](<https://devfeed.tech/tags/sample.md>), [security](<https://devfeed.tech/tags/security.md>), [telemetry](<https://devfeed.tech/tags/telemetry.md>)

### AI overview

Agent Anomaly Detection on the Gemini Platform analyzes agent traces and logs to flag sessions that stray beyond an agent's intended job. The guide demonstrates it with a deliberately over-permissive customer support agent using canned in-memory data: after telemetry and message logging are enabled, a normal session and a simulated payroll-data request are processed, and the latter is flagged for tool misuse. Setup requires a Cloud Logging bucket and Observability dataset in the same US or EU multi-region; monitored sessions are analyzed after setup, which takes about 30 minutes, and new sessions are scanned within about five minutes. The example finding was High severity with 95% probability, but detector results and severities can vary because the analysis models are non-deterministic. Cloud Monitoring metrics support anomaly and coverage alerts; Security Command Center can receive anomaly findings when enabled for the organization. The excerpt cautions against copying the intentionally unsafe demo tool into production.

### Source excerpt

This guide walks developers through setting up Agent Anomaly Detection on the Gemini Enterprise Agent Platform to monitor AI agent telemetry and catch behavioral risks. It demonstrates how to deploy a sample customer support agent, trigger out-of-scope security findings, and configure proactive alerts using Cloud Monitoring and Security Command Center.

## Microsoft 365 Outage SLA Credit Claims: MSPs Describe a Gulf Between the Work and the Refund

DevFeed: [Microsoft 365 Outage SLA Credit Claims: MSPs Describe a Gulf Between the Work and the Refund](<https://devfeed.tech/articles/microsoft-365-outage-sla-credit-claims-msps-describe-a-gulf-between-the-work-and-the-refund-134164.md>)

Original publisher: [Read original article](<https://joho-todai.com/en/microsoft-365-outage-sla-credit-claims-msp-experience/>)

Author: Toru Misaki

Published: 2026-11-14T22:20:12Z

Content type: news

Language: en

Sources: [Joho no Todai](<https://devfeed.tech/sources/joho-no-todai.md>)

Topics: [microsoft 365](<https://devfeed.tech/topics/microsoft-365.md>), [Microsoft](<https://devfeed.tech/topics/microsoft.md>)

Tags: [billing](<https://devfeed.tech/tags/billing.md>), [copilot](<https://devfeed.tech/tags/copilot.md>), [costs](<https://devfeed.tech/tags/costs.md>), [csp](<https://devfeed.tech/tags/csp.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [defender](<https://devfeed.tech/tags/defender.md>), [email](<https://devfeed.tech/tags/email.md>), [email-delivery](<https://devfeed.tech/tags/email-delivery.md>), [exchange](<https://devfeed.tech/tags/exchange.md>), [exchange-online](<https://devfeed.tech/tags/exchange-online.md>), [global-outage](<https://devfeed.tech/tags/global-outage.md>), [labor](<https://devfeed.tech/tags/labor.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-365](<https://devfeed.tech/tags/microsoft-365.md>), [microsoft-365-outage](<https://devfeed.tech/tags/microsoft-365-outage.md>), [msps](<https://devfeed.tech/tags/msps.md>), [outage](<https://devfeed.tech/tags/outage.md>), [partners](<https://devfeed.tech/tags/partners.md>), [process](<https://devfeed.tech/tags/process.md>), [refunds](<https://devfeed.tech/tags/refunds.md>)

### AI overview

Microsoft 365 SLA credits are limited to the prorated monthly fee for the affected service, so a suite outage does not automatically earn a credit on the whole license. Claims require the tenant GUID, incident ID, evidence of impact, and a request email from the affected customer's domain; only CSP direct and indirect providers may submit them. The non-Azure claim deadline for this incident is December 31, 2026. MSPs describe claims taking six to 18 months, with outcomes ranging from small credits to rejection despite extensive evidence and effort. Separate claims are needed for each customer, and credits do not cover lost business time or support labor. The excerpt presents filing as a choice between pursuing a potentially small credit and accepting the workload and uncertainty.

### Source excerpt

Following the major Microsoft 365 outage on November 12, CSP partners and MSPs are encouraging each other to file SLA credit claims. MSPs who have gone through the process before, however, describe a stark gap between the effort required and the refunds received.

## AMD openSIL Achieves Its Production Goal With 6th Gen EPYC "Venice"

DevFeed: [AMD openSIL Achieves Its Production Goal With 6th Gen EPYC "Venice"](<https://devfeed.tech/articles/amd-opensil-achieves-its-production-goal-with-6th-gen-epyc-venice-132601.md>)

Original publisher: [Read original article](<https://www.phoronix.com/news/AMD-openSIL-Venice-Goal>)

Author: Michael Larabel

Published: 2026-11-13T18:20:29Z

Content type: news

Language: en

Sources: [Phoronix](<https://devfeed.tech/sources/phoronix.md>)

Topics: [coreboot](<https://devfeed.tech/topics/coreboot.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [amd](<https://devfeed.tech/tags/amd.md>), [amd-epyc](<https://devfeed.tech/tags/amd-epyc.md>), [bmc](<https://devfeed.tech/tags/bmc.md>), [code](<https://devfeed.tech/tags/code.md>), [coreboot](<https://devfeed.tech/tags/coreboot.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [desktop-linux](<https://devfeed.tech/tags/desktop-linux.md>), [epyc](<https://devfeed.tech/tags/epyc.md>), [epyc-9006](<https://devfeed.tech/tags/epyc-9006.md>), [genoa](<https://devfeed.tech/tags/genoa.md>), [linux-benchmarking](<https://devfeed.tech/tags/linux-benchmarking.md>), [linux-hardware-benchmarks](<https://devfeed.tech/tags/linux-hardware-benchmarks.md>), [linux-hardware-reviews](<https://devfeed.tech/tags/linux-hardware-reviews.md>), [linux-how-to](<https://devfeed.tech/tags/linux-how-to.md>), [linux-performance](<https://devfeed.tech/tags/linux-performance.md>), [linux-server-benchmarks](<https://devfeed.tech/tags/linux-server-benchmarks.md>), [motherboard](<https://devfeed.tech/tags/motherboard.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-code](<https://devfeed.tech/tags/open-source-code.md>), [open-source-graphics](<https://devfeed.tech/tags/open-source-graphics.md>), [performance](<https://devfeed.tech/tags/performance.md>), [phoronix](<https://devfeed.tech/tags/phoronix.md>), [phoronix-test-suite](<https://devfeed.tech/tags/phoronix-test-suite.md>), [production](<https://devfeed.tech/tags/production.md>), [server](<https://devfeed.tech/tags/server.md>), [silicon](<https://devfeed.tech/tags/silicon.md>), [sp7](<https://devfeed.tech/tags/sp7.md>), [ubuntu-benchmarks](<https://devfeed.tech/tags/ubuntu-benchmarks.md>), [ubuntu-hardware](<https://devfeed.tech/tags/ubuntu-hardware.md>), [venice](<https://devfeed.tech/tags/venice.md>), [zen-6](<https://devfeed.tech/tags/zen-6.md>)

### AI overview

AMD says openSIL met its production goal with 6th Gen EPYC "Venice," and that no Venice platform has booted without it. The company plans to release the Venice open-source code within one quarter of the product launch. AMD is monitoring customer interest in Coreboot-based EPYC 9006 servers; the code release could enable independent efforts such as 3mdeb's earlier Dasharo work. EPYC 9006 reference motherboards continue to use OpenBMC.

### Source excerpt

Since 2023 I and many open-source advocates have been very eager for AMD's openSIL project for delivering open-source CPU silicon initialization as a replacement to the traditional AGESA code. With Zen 5 there was proof-of-concept support that we have enjoyed with Coreboot on an MSI Ryzen motherboard and a Gigabyte EPYC server motherboard thanks to the Coreboot-based Dasharo ports by 3mdeb. The goal for production readiness was for Zen 6, but with this week's EPYC 9006 SP7 announcement the openSIL topic wasn't touched on. So I asked around...

## AI Automation Safety: When to Keep a Human in the Loop

DevFeed: [AI Automation Safety: When to Keep a Human in the Loop](<https://devfeed.tech/articles/ai-automation-safety-when-to-keep-a-human-in-the-loop-130182.md>)

Original publisher: [Read original article](<https://hackernoon.com/ai-automation-safety-when-to-keep-a-human-in-the-loop?source=rss>)

Author: Muneeb Ashraf

Published: 2026-11-12T07:33:50Z

Content type: tutorial

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [human-in-the-loop UX](<https://devfeed.tech/topics/human-in-the-loop-ux.md>), [human review](<https://devfeed.tech/topics/human-review.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [workflow automation](<https://devfeed.tech/topics/workflow-automation.md>), [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [Automation](<https://devfeed.tech/topics/automation.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-automation](<https://devfeed.tech/tags/ai-automation.md>), [ai-automation-agency](<https://devfeed.tech/tags/ai-automation-agency.md>), [ai-customer-support](<https://devfeed.tech/tags/ai-customer-support.md>), [ai-safety](<https://devfeed.tech/tags/ai-safety.md>), [ai-workflow](<https://devfeed.tech/tags/ai-workflow.md>), [airtable](<https://devfeed.tech/tags/airtable.md>), [audit-trail](<https://devfeed.tech/tags/audit-trail.md>), [automation](<https://devfeed.tech/tags/automation.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [dashboard](<https://devfeed.tech/tags/dashboard.md>), [dutch](<https://devfeed.tech/tags/dutch.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [invoice-automation](<https://devfeed.tech/tags/invoice-automation.md>), [workflow-automation](<https://devfeed.tech/tags/workflow-automation.md>), [zapier](<https://devfeed.tech/tags/zapier.md>)

### AI overview

The article recommends keeping human approval on actions that are hard to reverse, leave the organization's control, or carry costs that outweigh a short delay. Its examples include automatically answering tested routine support questions while routing unclear or money-related cases to reviewers, and holding invoices when bank details differ from supplier records until someone verifies the change through a contact already on file. It also describes silent workflow failure despite an "active" status, recommending checks for expected output, and treats retired-model replacements as new releases requiring about three weeks of testing across languages, outputs, and human-reviewed live runs. In Zapier, Human in the Loop pauses a Zap for a reviewer's decision; later steps can use the decision and reviewer note, with approvals and declines logged. The article says the feature requires a paid plan, reviewers need Zapier accounts and shared access, Pro allows requests only to yourself, and the step cannot be inside or after a Looping by Zapier step.

### Source excerpt

AI can send routine replies, but it never pays or deletes. How we decide which automation steps need a human approval, and how to build one in Zapier.

## Automated Root-Cause Subclassification and No-Code Fix Generation for Invalid Bug Reports

DevFeed: [Automated Root-Cause Subclassification and No-Code Fix Generation for Invalid Bug Reports](<https://devfeed.tech/articles/automated-root-cause-subclassification-and-no-code-fix-generation-for-invalid-bug-reports-128109.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.17561>)

Author: Mahmut Furkan Gon, Emre Dinc, Tevfik Emre Sungur, Eray Tuzun

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [automated](<https://devfeed.tech/tags/automated.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bug](<https://devfeed.tech/tags/bug.md>), [bug-reports](<https://devfeed.tech/tags/bug-reports.md>), [cause](<https://devfeed.tech/tags/cause.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [detection](<https://devfeed.tech/tags/detection.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [feature](<https://devfeed.tech/tags/feature.md>), [generate](<https://devfeed.tech/tags/generate.md>), [generation](<https://devfeed.tech/tags/generation.md>), [llms](<https://devfeed.tech/tags/llms.md>), [no-code](<https://devfeed.tech/tags/no-code.md>), [performance](<https://devfeed.tech/tags/performance.md>), [question](<https://devfeed.tech/tags/question.md>), [rag](<https://devfeed.tech/tags/rag.md>), [rag-applications](<https://devfeed.tech/tags/rag-applications.md>), [reproducibility](<https://devfeed.tech/tags/reproducibility.md>), [root](<https://devfeed.tech/tags/root.md>)

### AI overview

On the manually curated benchmark, retrieval augmented generation led invalid bug report subclassification with 0.66 weighted F1, narrowly ahead of vanilla LLMs (0.65) and agentic web search (0.64). Detection varied by subclass: Non-reproducibility reached 0.85 F1, while Wrong Version was hardest, scoring 0.00-0.29. For no-code fix suggestions, agentic web search had the highest Judge LLM success rate (68.9%), versus 64.9% for vanilla LLMs and 64.4% for RAG. Its strongest subclass results were Working as Designed (87.4%) and Question (72.2%). The study also assessed suggestions with BERTScore; the supplied abstract gives no corresponding scores.

### Source excerpt

arXiv:2605.17561v3 Announce Type: replace-cross Abstract: Issues faced when using software are reported in the form of bug reports. However, many bug reports are invalid, meaning they do not require code changes, and are resolved with a no-code fix. Manually determining the root cause of the invalid bug reports and providing actionable resolutions by the customer support causes a serious waste of resources. Our goal is to introduce a standardized taxonomy for root-cause oriented invalid bug report subclassification, and perform experiments to test the accuracy of various approaches on invalid subclassification and no-code fix generation. We study how different configurations perform on a gold-standard benchmark we have created. Using a manually curated benchmark for higher quality analysis, we experimented with vanilla LLMs, Retrieval Augmented Generation, and agentic web search to identify invalid subclasses and generate no-code fixes. We evaluated the results against manually labeled ground truth data that includes the invalid subclass and no-code fixes from the original bug reports. We measured subclass detection performance with weighted F1-Score, and assessed no-code fix suggestions using BERTScore and Judge LLM success rates. For subclassification, retrieval augmented generation achieves the highest overall performance with 0.66 weighted F1, slightly outperforming vanilla LLMs at 0.65 and agentic web search at 0.64. At the subclass level, performance peaks at 0.85 F1 for Non-reproducibility and 0.79 for Feature Request and Question, while Wrong Version remains the most challenging with scores between 0.00 and 0.29. For no-code fix generation, agentic web search achieves the highest overall Judge LLM success rate at 68.9%, compared to 64.4% for RAG applications and 64.9% for vanilla LLMs, with subclass-level peaks of 87.4% for Working as Designed and 72.2% for Question.

## A letter from our founders

DevFeed: [A letter from our founders](<https://devfeed.tech/articles/a-letter-from-our-founders-129184.md>)

Original publisher: [Read original article](<https://www.getdbt.com/blog/founders-letter>)

Author: George Fraser, Tristan Handy, Taylor Brown

Published: 2026-11-11T00:00:00Z

Content type: opinion

Language: en

Sources: [dbt Labs](<https://devfeed.tech/sources/dbt-labs.md>)

Topics: [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [ai-governance](<https://devfeed.tech/topics/ai-governance.md>), [Digital Public Good](<https://devfeed.tech/topics/digital-public-goods.md>), [digital sovereignty](<https://devfeed.tech/topics/digital-sovereignty.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [analytics-engineering](<https://devfeed.tech/tags/analytics-engineering.md>), [apis](<https://devfeed.tech/tags/apis.md>), [automate](<https://devfeed.tech/tags/automate.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-agents](<https://devfeed.tech/tags/autonomous-agents.md>), [company](<https://devfeed.tech/tags/company.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [data](<https://devfeed.tech/tags/data.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [founders](<https://devfeed.tech/tags/founders.md>), [future](<https://devfeed.tech/tags/future.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [open](<https://devfeed.tech/tags/open.md>), [open-data](<https://devfeed.tech/tags/open-data.md>)

### AI overview

The founders argue that useful agents combine intelligence with tools and data, and that their effectiveness depends on access to business context. They distinguish data-heavy analysis, which may need structured data plus OKRs, notes, and initiatives, from tasks such as contract review that require a more limited set of materials. Agents could expand access to analysis by being available on demand and able to investigate for longer, but the authors say this requires infrastructure such as evaluations and shared memory. They favor connecting agents to operational systems and the definitions and checks that explain the data, rather than relying only on documents and messages. The letter also describes a shift from AI augmenting workers to agents owning processes end to end, and names four responsibilities for agent engineers: process and systems engineering, evaluation and operations, and change management. It advocates open standards, portability, data ownership, and governance to avoid dependence on one vendor, while emphasizing continued investment in human practitioners.

### Source excerpt

Our founders on the age of intelligence: agents, open data infrastructure, and the humans building the agentic future.

## Customer support is the gnarliest AI problem

DevFeed: [Customer support is the gnarliest AI problem](<https://devfeed.tech/articles/customer-support-is-the-gnarliest-ai-problem-127396.md>)

Original publisher: [Read original article](<https://stackoverflow.blog/2026/11/10/customer-support-is-the-gnarliest-ai-problem/>)

Author: Phoebe Sajor

Published: 2026-11-10T08:40:00Z

Content type: article

Language: en

Sources: [Stack Overflow](<https://devfeed.tech/sources/stack-overflow-blog.md>)

Topics: [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [AI tools for developers](<https://devfeed.tech/topics/ai-tools-for-developers.md>), [Software Engineering](<https://devfeed.tech/topics/software-engineering.md>), [engineering-leadership](<https://devfeed.tech/topics/engineering-leadership.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ahmad-al-dahle](<https://devfeed.tech/tags/ahmad-al-dahle.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-coding-tools](<https://devfeed.tech/tags/ai-coding-tools.md>), [airbnb](<https://devfeed.tech/tags/airbnb.md>), [balance](<https://devfeed.tech/tags/balance.md>), [blog](<https://devfeed.tech/tags/blog.md>), [cto](<https://devfeed.tech/tags/cto.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [engineers](<https://devfeed.tech/tags/engineers.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [implement](<https://devfeed.tech/tags/implement.md>), [latency](<https://devfeed.tech/tags/latency.md>), [podcast](<https://devfeed.tech/tags/podcast.md>), [se-stackoverflow](<https://devfeed.tech/tags/se-stackoverflow.md>), [se-tech](<https://devfeed.tech/tags/se-tech.md>), [software](<https://devfeed.tech/tags/software.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [teams](<https://devfeed.tech/tags/teams.md>), [tools](<https://devfeed.tech/tags/tools.md>)

### AI overview

Airbnb CTO Ahmad Al-Dahle describes customer support as a particularly difficult AI problem and discusses deploying multi-party AI agents to address it. He explains how evolving, rigorous evaluations help engineers balance response latency with accuracy, and argues that AI coding tools require software teams to exercise greater judgment. The supplied text gives no further details about the agents, evaluations, or engineering practices.

### Source excerpt

Ryan chats with Ahmad Al-Dahle, CTO at Airbnb, about deploying multi-party AI agents to tackle the uniquely gnarly problem of customer support, how a moving target of rigorous evaluations helps engineers balance latency against accuracy, and the ways in which AI coding tools demand an even higher level of judgment from software engineering teams.

## Singapore Airlines elevates travel & work with OpenAI technology

DevFeed: [Singapore Airlines elevates travel & work with OpenAI technology](<https://devfeed.tech/articles/singapore-airlines-elevates-travel-work-with-openai-technology-127938.md>)

Original publisher: [Read original article](<https://openai.com/index/singapore-airlines-travel-work>)

Published: 2026-11-10T08:00:00Z

Content type: article

Language: en

Sources: [OpenAI News](<https://devfeed.tech/sources/openai-news.md>)

Topics: [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [OpenAI](<https://devfeed.tech/topics/openai.md>), [Chat Bot](<https://devfeed.tech/topics/chatbot.md>), [Developer experience](<https://devfeed.tech/topics/developer-experience.md>), [developer-productivity](<https://devfeed.tech/topics/developer-productivity.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-applications](<https://devfeed.tech/tags/ai-applications.md>), [airlines](<https://devfeed.tech/tags/airlines.md>), [api](<https://devfeed.tech/tags/api.md>), [chatbot](<https://devfeed.tech/tags/chatbot.md>), [csat](<https://devfeed.tech/tags/csat.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [discovery](<https://devfeed.tech/tags/discovery.md>), [employee](<https://devfeed.tech/tags/employee.md>), [employee-productivity](<https://devfeed.tech/tags/employee-productivity.md>), [openai](<https://devfeed.tech/tags/openai.md>), [openai-models](<https://devfeed.tech/tags/openai-models.md>), [productivity](<https://devfeed.tech/tags/productivity.md>), [singapore](<https://devfeed.tech/tags/singapore.md>), [support](<https://devfeed.tech/tags/support.md>), [technology](<https://devfeed.tech/tags/technology.md>), [travel](<https://devfeed.tech/tags/travel.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

Singapore Airlines reports a 2x increase in customer satisfaction after replacing Kris's predefined dialog flows with an OpenAI API-powered chatbot that interprets natural-language intent, grounds answers in a knowledge base, handles multi-topic conversations, and passes context to agents when needed. AI-assisted customer service tools, including call summaries, are associated with about 10% lower average handling time. Its Flight Recommender connects an orchestration layer to fare and availability data to return natural-language recommendations; monthly users grew nearly 3x over the past year. The airline also uses OpenAI models across software delivery tasks, reporting gains of up to 20% in selected activities, and in Jarvis, a governed internal platform used monthly by more than 60% of employees. The article describes controls including validation, data access limits, PII handling, and human review. Booking changes and baggage transactions for Kris, and a more conversational Recommender, are planned expansions.

### Source excerpt

Discover how Singapore Airlines uses OpenAI models and API to enhance customer support, travel discovery, and employee productivity, doubling CSAT for its Kris chatbot.

## Call-Level Revision Contracts for LLM-Authored Workflows

DevFeed: [Call-Level Revision Contracts for LLM-Authored Workflows](<https://devfeed.tech/articles/call-level-revision-contracts-for-llm-authored-workflows-120803.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06967>)

Author: Chenyang Jiang, Benyang Pan, Puqu Wu

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [call hierarchy](<https://devfeed.tech/topics/call-hierarchy.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [call](<https://devfeed.tech/tags/call.md>), [change](<https://devfeed.tech/tags/change.md>), [container](<https://devfeed.tech/tags/container.md>), [contracts](<https://devfeed.tech/tags/contracts.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [digest](<https://devfeed.tech/tags/digest.md>), [durable-execution](<https://devfeed.tech/tags/durable-execution.md>), [engines](<https://devfeed.tech/tags/engines.md>), [execution](<https://devfeed.tech/tags/execution.md>), [expression](<https://devfeed.tech/tags/expression.md>), [failed](<https://devfeed.tech/tags/failed.md>), [gate](<https://devfeed.tech/tags/gate.md>), [history](<https://devfeed.tech/tags/history.md>), [llm](<https://devfeed.tech/tags/llm.md>), [revision](<https://devfeed.tech/tags/revision.md>), [workflows](<https://devfeed.tech/tags/workflows.md>)

### AI overview

The revision contract aims to let models repair failed workflows without repaying successful calls or silently reusing results whose payloads changed. It identifies each capability call by callsite, fanout position, and occurrence count; a static gate rejects changes to committed calls before execution, while a runtime digest catches changed payloads and an occurrence ledger tracks repeated paid payloads. The call expression is frozen, but surrounding orchestration code can change. Under three assumptions, accepted repairs replay invoked committed calls, run only new work, and report skipped identities. The abstract reports 22 media-workflow repairs replaying 258 committed identities without duplicates, and a live six-way fanout costing seven calls instead of twelve; a 12-task customer-support study used the same checks on two models. Relocating a call can trigger a pre-spend stop because no local identity is both relocation-invariant and free of wrong reuse.

### Source excerpt

arXiv:2611.06967v1 Announce Type: new Abstract: Durable execution engines replay a recorded history and do not compare activity payloads. When a model repairs a workflow that failed partway through, it must either pay again for calls that succeeded or accept a recorded result whose payload changed. We define a revision contract for workflows whose orchestration code is written by a model. Each capability call carries a structural identity built from its callsite, its fanout position, and its occurrence count, and the identity stays stable under rewrites that leave the call in place, so committed identities form a frozen prefix. The frozen unit is the call expression, and the surrounding code stays writable. A static gate rejects a rewrite that would change a committed call before a container starts. A runtime digest stops a changed payload at a committed identity, and an occurrence ledger names a repeated paid payload. Under three assumptions, a candidate the gate accepts replays every committed call it invokes, executes only new work, and reports the identities it skips. No local identity is both relocation-invariant and free of wrong reuse, so callsite identity turns a relocation into a pre-spend stop. On media workflows, 22 gate-passed repairs replay 258 committed identities with no duplicate, and a live six-way fanout makes seven paid calls instead of twelve. A 12-task customer-support study applies the same identity, replay, and boundary checks on two models.

## The next chapter of Redis Iris: what agents need to work in the real world

DevFeed: [The next chapter of Redis Iris: what agents need to work in the real world](<https://devfeed.tech/articles/the-next-chapter-of-redis-iris-what-agents-need-to-work-in-the-real-world-119008.md>)

Original publisher: [Read original article](<https://redis.io/blog/the-next-chapter-of-redis-iris-what-agents-need-to-work-in-the-real-world/>)

Author: Simba Khadder

Published: 2026-11-04T00:00:00Z

Content type: release

Language: en

Sources: [Redis Blog](<https://devfeed.tech/sources/redis-blog-2.md>), [Redis Blog](<https://devfeed.tech/sources/redis-blog.md>)

Topics: [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [agent to agent communication](<https://devfeed.tech/topics/agent-to-agent-communication.md>), [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [agent orchestration](<https://devfeed.tech/topics/agent-orchestration.md>), [Redis](<https://devfeed.tech/topics/redis.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [a2a](<https://devfeed.tech/tags/a2a.md>), [adaptive](<https://devfeed.tech/tags/adaptive.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-memory](<https://devfeed.tech/tags/agent-memory.md>), [agent-mesh](<https://devfeed.tech/tags/agent-mesh.md>), [agents](<https://devfeed.tech/tags/agents.md>), [agents-in-production](<https://devfeed.tech/tags/agents-in-production.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-stack](<https://devfeed.tech/tags/ai-stack.md>), [audit](<https://devfeed.tech/tags/audit.md>), [audit-trail](<https://devfeed.tech/tags/audit-trail.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [beta](<https://devfeed.tech/tags/beta.md>), [built-in](<https://devfeed.tech/tags/built-in.md>), [cache](<https://devfeed.tech/tags/cache.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [engineering-ai](<https://devfeed.tech/tags/engineering-ai.md>), [redis](<https://devfeed.tech/tags/redis.md>), [tech](<https://devfeed.tech/tags/tech.md>)

### AI overview

Redis Iris is presented as a context and runtime layer intended to make AI agents more reliable in production by giving them relevant context quickly, retaining useful history, and reusing established procedures. Its six capabilities include intent classification for routing; Agent Playbooks that reuse stable answers or procedures and improve through a learning loop; Agent Memory that builds, organizes, audits, and can forget sensitive or irrelevant information; and Context Gateway, which connects context sources while handling identity and authorization. Lens (beta) traces outcomes to routes, playbooks, memories, and source sessions, with task, token, and latency visibility. Agent Mesh (beta) provides durable SSE streams and A2A communication for handoffs. The article illustrates these components in a hypothetical account-charge support interaction. Capabilities can be used together or independently; claims of customer results are vendor-reported.

### Source excerpt

The other day, I was chatting with a customer support AI agent dealing with a cancelled flight. The experience was terrible. You could tell the agent realistically only had a few functions, and this was clearly not one of them. I'm sure when the devel...

## Salesforce to buy customer service automation startup Cresta for reported $2.7B

DevFeed: [Salesforce to buy customer service automation startup Cresta for reported $2.7B](<https://devfeed.tech/articles/salesforce-to-buy-customer-service-automation-startup-cresta-for-reported-2-7b-115618.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/03/salesforce-to-buy-customer-service-automation-startup-cresta-for-reported-2-7b/>)

Author: Maria Deutscher

Published: 2026-11-03T21:20:22Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [acquisition](<https://devfeed.tech/tags/acquisition.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-platform](<https://devfeed.tech/tags/ai-platform.md>), [ai-video](<https://devfeed.tech/tags/ai-video.md>), [ai-voice-agents](<https://devfeed.tech/tags/ai-voice-agents.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [automation](<https://devfeed.tech/tags/automation.md>), [conversation-intelligence](<https://devfeed.tech/tags/conversation-intelligence.md>), [cresta](<https://devfeed.tech/tags/cresta.md>), [customer-service](<https://devfeed.tech/tags/customer-service.md>), [customer-service-automation](<https://devfeed.tech/tags/customer-service-automation.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [deals](<https://devfeed.tech/tags/deals.md>), [news](<https://devfeed.tech/tags/news.md>), [nyse-crm](<https://devfeed.tech/tags/nyse-crm.md>), [salesforce](<https://devfeed.tech/tags/salesforce.md>), [startup](<https://devfeed.tech/tags/startup.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-6](<https://devfeed.tech/tags/top-story-6.md>)

### AI overview

Salesforce agreed to acquire customer service software provider Cresta; Salesforce did not disclose the price, while The Information reported $2.7 billion. Cresta uses AI to answer inquiries and handle tasks such as appointment scheduling, tailoring responses to customer preferences and prior interactions. Companies can configure voice agents for specific support categories, including their AI models and compliance guardrails. Cresta also coordinates multiple agents on complex requests, offers Deputy to gather information and manage employee follow-up, and gives human representatives real-time guidance. Its Opera tool fine-tunes models for customer service, while Synthetic Customers generates conversations for agent testing. Salesforce expects to close the deal in its 2028 fiscal year, starting next February.

### Source excerpt

Salesforce Inc. has agreed to buy Cresta Intelligence Inc., a venture-backed provider of customer service software. The cloud giant didn't disclose the financial terms in its late Monday announcement of the deal. The Information reported that Salesforce is paying $2.7 billion. Cresta's platform uses artificial intelligence to answer customer inquiries. It can also perform more [...] The post Salesforce to buy customer service automation startup Cresta for reported $2.7B appeared first on SiliconANGLE.

## How a Bank Serving 100 Million Users Tests Agentic AI in Production

DevFeed: [How a Bank Serving 100 Million Users Tests Agentic AI in Production](<https://devfeed.tech/articles/how-a-bank-serving-100-million-users-tests-agentic-ai-in-production-109110.md>)

Original publisher: [Read original article](<https://hackernoon.com/how-a-bank-serving-100-million-users-tests-agentic-ai-in-production?source=rss>)

Author: sarahevans

Published: 2026-10-30T13:05:58Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>), [blue green and canary deployments](<https://devfeed.tech/topics/blue-green-and-canary-deployments.md>), [experiments](<https://devfeed.tech/topics/experiments.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-research](<https://devfeed.tech/tags/ai-research.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [bank](<https://devfeed.tech/tags/bank.md>), [business](<https://devfeed.tech/tags/business.md>), [customer-experience](<https://devfeed.tech/tags/customer-experience.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [fintech](<https://devfeed.tech/tags/fintech.md>), [llm-evaluation](<https://devfeed.tech/tags/llm-evaluation.md>), [production](<https://devfeed.tech/tags/production.md>), [tests](<https://devfeed.tech/tags/tests.md>)

### AI overview

The bank's five customer support agents improved transactional NPS in online A/B tests, but all remained below separately measured human benchmarks. To assess new versions, the team first validates automated graders against expert labels: one grader's initial 68.88% accuracy trailed a 77.78% majority-label baseline, then rose to 88.89% after prompt optimization. That score measures agreement with expert labels, not agent accuracy. New versions begin at 1% to 5% customer exposure, with the team monitoring self-service, satisfaction, escalations, and errors. Agents hand off cases when requests exceed their routines, required data is unavailable, or confidence is low; adding a card-reissue capability removed a transfer from the most common card-delivery resolution path. The excerpt stresses that satisfaction scores alone can obscure workload shifts: harder cases left for human agents may take longer, so teams need to examine why cases escalate. The graders' cross-model tests covered models from one vendor, with broader coverage left for future work.

### Source excerpt

A KDD 2026 paper shows how a bank tests agentic AI in production: graders checked against experts, 1% to 5% rollouts, and the human handoffs behind the scores.

## Your Agent's Tools Can Make Decisions, Too

DevFeed: [Your Agent's Tools Can Make Decisions, Too](<https://devfeed.tech/articles/your-agent-s-tools-can-make-decisions-too-107783.md>)

Original publisher: [Read original article](<https://towardsdatascience.com/your-agents-tools-can-make-decisions-too/>)

Author: Shuai Guo

Published: 2026-10-30T12:30:00Z

Content type: tutorial

Language: en

Sources: [Towards Data Science](<https://devfeed.tech/sources/towards-data-science.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [Multi-step AI processes](<https://devfeed.tech/topics/multi-step-ai-processes.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [software-architecture](<https://devfeed.tech/topics/software-architecture.md>), [OpenAI](<https://devfeed.tech/topics/openai.md>), [Agent Harness](<https://devfeed.tech/topics/agent-harness.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-tools](<https://devfeed.tech/tags/agent-tools.md>), [agents-sdk](<https://devfeed.tech/tags/agents-sdk.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [azure](<https://devfeed.tech/tags/azure.md>), [case-study](<https://devfeed.tech/tags/case-study.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [decision-models](<https://devfeed.tech/tags/decision-models.md>), [delay](<https://devfeed.tech/tags/delay.md>), [editor-s-picks](<https://devfeed.tech/tags/editor-s-picks.md>), [jev](<https://devfeed.tech/tags/jev.md>), [llm-agents](<https://devfeed.tech/tags/llm-agents.md>), [llm-applications](<https://devfeed.tech/tags/llm-applications.md>), [llm-workflows](<https://devfeed.tech/tags/llm-workflows.md>), [tools](<https://devfeed.tech/tags/tools.md>)

### AI overview

The decision-model-as-a-tool pattern moves routine lookup choices into a tool: an internal decision model selects and runs backend functions in a loop, then returns collected evidence to the LLM agent. In the order-support example, the specialist exposes six information lookups behind one interface; the agent provides an order ID and question, while the tool tracks evidence and decisions. The model can stop with `done` when it has enough evidence or `blocked` when available operations cannot resolve the issue. Using Decision-1 on Azure, the example completed in 8.9 seconds with one Luna call; an otherwise similarly configured agent calling the six functions directly made six Luna calls and took 25.4 seconds. Both reached the same core recommendation; the author reports 2.85x faster execution and 24.7% lower model cost. The functions use a small synthetic dataset, so these results describe the case study.

### Source excerpt

Decision models as tools: a hands-on guide The post Your Agent's Tools Can Make Decisions, Too appeared first on Towards Data Science.

## The support case of a customer who said that all they could see was a black screen

DevFeed: [The support case of a customer who said that all they could see was a black screen](<https://devfeed.tech/articles/the-support-case-of-a-customer-who-said-that-all-they-could-see-was-a-black-screen-107321.md>)

Original publisher: [Read original article](<https://devblogs.microsoft.com/oldnewthing/20261029-00/?p=112808/>)

Author: Raymond Chen

Published: 2026-10-29T14:00:00Z

Content type: article

Language: en

Sources: [Raymond Chen](<https://devfeed.tech/sources/raymond-chen.md>)

Topics: [Support](<https://devfeed.tech/topics/support.md>), [screen](<https://devfeed.tech/topics/screen.md>)

Tags: [analysis](<https://devfeed.tech/tags/analysis.md>), [background](<https://devfeed.tech/tags/background.md>), [black](<https://devfeed.tech/tags/black.md>), [black-screen](<https://devfeed.tech/tags/black-screen.md>), [bug](<https://devfeed.tech/tags/bug.md>), [builds](<https://devfeed.tech/tags/builds.md>), [button](<https://devfeed.tech/tags/button.md>), [color](<https://devfeed.tech/tags/color.md>), [colors](<https://devfeed.tech/tags/colors.md>), [command](<https://devfeed.tech/tags/command.md>), [compositor](<https://devfeed.tech/tags/compositor.md>), [computer](<https://devfeed.tech/tags/computer.md>), [control-panel](<https://devfeed.tech/tags/control-panel.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [desktop](<https://devfeed.tech/tags/desktop.md>), [drawing](<https://devfeed.tech/tags/drawing.md>), [driver](<https://devfeed.tech/tags/driver.md>), [history](<https://devfeed.tech/tags/history.md>), [old-new-thing](<https://devfeed.tech/tags/old-new-thing.md>), [screen](<https://devfeed.tech/tags/screen.md>), [support](<https://devfeed.tech/tags/support.md>)

### AI overview

Holding a PC's power button for eight seconds prompts Windows to capture an emergency kernel dump; a 12-second hold forces hardware-level shutdown. Dumps that show an all-black screen are classified as black screen LPBH failures and routed to a specialist team. Causes can range from a broken video driver to the compositor supplying black pixels or no content being drawn, including because the usual drawing process is hung. The excerpt also recounts a Windows 3.1 case where a customer set every color, including text and backgrounds, to black, making the interface invisible. The author guesses support may have restored defaults by deleting the `[colors]` section from `win.ini`, but says the actual fix is unknown.

### Source excerpt

Then again, that is what they asked for. The post The support case of a customer who said that all they could see was a black screen appeared first on The Old New Thing.

## How to build multilingual AI agents for customer support

DevFeed: [How to build multilingual AI agents for customer support](<https://devfeed.tech/articles/how-to-build-multilingual-ai-agents-for-customer-support-109682.md>)

Original publisher: [Read original article](<https://www.twilio.com/en-us/blog/insights/multilingual-ai-agents>)

Author: Jesse Sumrak

Published: 2026-10-29T00:00:00Z

Content type: tutorial

Language: en

Sources: [Twilio](<https://devfeed.tech/sources/twilio-blog.md>), [Twilio Blog](<https://devfeed.tech/sources/twilio-blog-2.md>)

Topics: [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>), [twilio api](<https://devfeed.tech/topics/twilio-api.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-voice-agent](<https://devfeed.tech/tags/ai-voice-agent.md>), [ai-voice-agents](<https://devfeed.tech/tags/ai-voice-agents.md>), [api](<https://devfeed.tech/tags/api.md>), [apis](<https://devfeed.tech/tags/apis.md>), [audio](<https://devfeed.tech/tags/audio.md>), [automatically](<https://devfeed.tech/tags/automatically.md>), [build](<https://devfeed.tech/tags/build.md>), [build-ai-agents](<https://devfeed.tech/tags/build-ai-agents.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [industry-insights](<https://devfeed.tech/tags/industry-insights.md>), [language](<https://devfeed.tech/tags/language.md>), [messaging](<https://devfeed.tech/tags/messaging.md>), [multilingual-ai](<https://devfeed.tech/tags/multilingual-ai.md>), [switch](<https://devfeed.tech/tags/switch.md>), [twilio](<https://devfeed.tech/tags/twilio.md>), [voice](<https://devfeed.tech/tags/voice.md>)

### AI overview

Twilio ConversationRelay can detect language per utterance and switch a voice agent mid-call without restarting: configure a <Language> element for each supported language, use Deepgram with speech-to-text language set to multi, then read each transcript's lang tag and send a language message to update recognition and text-to-speech. Detected tags use primary codes such as es, without regional suffixes. Language and voice coverage and quality depend on the selected providers; ElevenLabs text-to-speech can also detect language when set to multi, and Deepgram Flux can use configured languages as recognition hints. Pass the current language into the LLM context and define how it handles mixed-language input; keep factual values sourced from tools or a knowledge base. The excerpt also describes storing language preference in Twilio Conversation Memory for future calls and using Agent Connect across voice and messaging channels. For human escalation, pass the detected language as a routing attribute to Flex so it can route to a matching representative with conversation history.

### Source excerpt

Learn how to build AI agents that detect a caller's language and switch mid-conversation (across voice and messaging) with Twilio ConversationRelay.

## Brilliant Builds: Building for Seamlessness

DevFeed: [Brilliant Builds: Building for Seamlessness](<https://devfeed.tech/articles/brilliant-builds-building-for-seamlessness-105803.md>)

Original publisher: [Read original article](<https://www.twilio.com/en-us/blog/events/Brilliant-builds-episode-3>)

Author: Twilio

Published: 2026-10-29T00:00:00Z

Content type: article

Language: en

Sources: [Twilio](<https://devfeed.tech/sources/twilio-blog.md>), [Twilio Blog](<https://devfeed.tech/sources/twilio-blog-2.md>)

Topics: [ai regulation](<https://devfeed.tech/topics/ai-regulation.md>)

Tags: [2025](<https://devfeed.tech/tags/2025.md>), [3](<https://devfeed.tech/tags/3.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-code](<https://devfeed.tech/tags/ai-code.md>), [ai-data](<https://devfeed.tech/tags/ai-data.md>), [ai-leadership](<https://devfeed.tech/tags/ai-leadership.md>), [ai-regulation](<https://devfeed.tech/tags/ai-regulation.md>), [ai-strategy](<https://devfeed.tech/tags/ai-strategy.md>), [brilliant-builds](<https://devfeed.tech/tags/brilliant-builds.md>), [ceo](<https://devfeed.tech/tags/ceo.md>), [context](<https://devfeed.tech/tags/context.md>), [conversations](<https://devfeed.tech/tags/conversations.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [eu](<https://devfeed.tech/tags/eu.md>), [events-community](<https://devfeed.tech/tags/events-community.md>), [iso](<https://devfeed.tech/tags/iso.md>), [omnichannel-orchestration](<https://devfeed.tech/tags/omnichannel-orchestration.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [rcs](<https://devfeed.tech/tags/rcs.md>), [sms](<https://devfeed.tech/tags/sms.md>)

### AI overview

Episode 3 features Matthew Blakemore on omnichannel orchestration: connecting customer support channels while preserving conversation context, so customers do not have to repeat their issue as interactions move across SMS, Voice, RCS, and WhatsApp. The excerpt identifies Blakemore as an AI strategy lecturer and contributor to the EU AI Office's General-Purpose AI Code of Practice. It gives no detailed implementation guidance or comparisons between channels.

### Source excerpt

In Episode 3, we sit down with Matthew Blakemore--AI Leadership & AI Strategy Lecturer and key contributor to the EU AI Office's General-Purpose AI Code of Practice.

## KORA: Deterministic-First Execution Control for AI Workloads

DevFeed: [KORA: Deterministic-First Execution Control for AI Workloads](<https://devfeed.tech/articles/kora-deterministic-first-execution-control-for-ai-workloads-103618.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.35869>)

Author: Albert Heekwan Kim

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [LLM observability](<https://devfeed.tech/topics/llm-observability.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-applications](<https://devfeed.tech/tags/ai-applications.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [execution](<https://devfeed.tech/tags/execution.md>), [faq](<https://devfeed.tech/tags/faq.md>), [inference](<https://devfeed.tech/tags/inference.md>), [lookup](<https://devfeed.tech/tags/lookup.md>), [policy](<https://devfeed.tech/tags/policy.md>), [schema](<https://devfeed.tech/tags/schema.md>), [telemetry](<https://devfeed.tech/tags/telemetry.md>), [transformation](<https://devfeed.tech/tags/transformation.md>), [validation](<https://devfeed.tech/tags/validation.md>)

### AI overview

KORA adds a pre-inference control layer that tries deterministic or structured task routes, validates their results, records telemetry, and escalates to a model-candidate path when those routes are insufficient. In a reproducible, deliberately deterministic-heavy 100-task benchmark, it avoided 80 simulated model invocations versus a direct-model baseline, with zero mismatches in the deterministic subset. A separate synthetic, local/no-network customer-support workload demonstrates route measurement and avoided-call accounting. The evidence is limited to these documented workloads; it does not establish production behavior, real-provider outcomes, energy or API-cost reductions, or broad superiority.

### Source excerpt

arXiv:2610.35869v1 Announce Type: cross Abstract: AI applications often invoke a model before they determine whether inference is necessary. That default is useful for ambiguous generative tasks, but it is poorly matched to work that is structured, deterministic, policy-bound, lookup-oriented, or directly verifiable. KORA introduces deterministic-first execution control: a pre-inference control layer that represents requests as task paths, attempts deterministic or structured routes first, validates route outcomes, records telemetry, and escalates to a model-candidate path only when deterministic handling is insufficient. This paper presents KORA as a deterministic-first execution-control architecture and reports bounded evidence from reproducible benchmark and validation workloads. In a reproducible 100-task deterministic-heavy benchmark, KORA-controlled execution avoided 80 of 100 simulated model invocations versus a naive direct baseline, with zero mismatches in the deterministic subset. The workload is intentionally deterministic-heavy: it includes categories such as FAQ lookup, policy/rule checking, schema or format validation, deterministic transformation, and a smaller set of model-needed ambiguous or generative tasks. A separate synthetic customer-support local/no-network validation workload demonstrates route measurement and avoided model-call accounting in an application-shaped setting. These results support bounded model-invocation reduction and validation visibility in documented workloads. They do not establish production behavior, real-provider outcomes, energy outcomes, API-cost reduction, or broad superiority.

## Why AI Agent Pricing Is Still Unsolved and What It Means for How You Track Costs

DevFeed: [Why AI Agent Pricing Is Still Unsolved and What It Means for How You Track Costs](<https://devfeed.tech/articles/why-ai-agent-pricing-is-still-unsolved-and-what-it-means-for-how-you-track-costs-93935.md>)

Original publisher: [Read original article](<https://hackernoon.com/why-ai-agent-pricing-is-still-unsolved-and-what-it-means-for-how-you-track-costs?source=rss>)

Author: Deniz Akbasaran

Published: 2026-10-21T21:56:11Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Finance](<https://devfeed.tech/topics/finance.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-agent-pricing](<https://devfeed.tech/tags/ai-agent-pricing.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-pricing](<https://devfeed.tech/tags/ai-pricing.md>), [ai-pricing-models](<https://devfeed.tech/tags/ai-pricing-models.md>), [billing](<https://devfeed.tech/tags/billing.md>), [coding](<https://devfeed.tech/tags/coding.md>), [cost-per-resolution](<https://devfeed.tech/tags/cost-per-resolution.md>), [costs](<https://devfeed.tech/tags/costs.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [llm-cost-control](<https://devfeed.tech/tags/llm-cost-control.md>), [pricing](<https://devfeed.tech/tags/pricing.md>), [saas](<https://devfeed.tech/tags/saas.md>), [value](<https://devfeed.tech/tags/value.md>)

### AI overview

AI agent pricing remains unsettled because customer value varies while per-interaction LLM costs shift. The article compares four models: resolution billing offers a clear ROI proxy but can reward closing tickets over solving them and penalizes appropriate escalation; conversation billing avoids that incentive and is easier to forecast, but assumes every interaction has value; cost-plus pricing follows unstable inference costs and may miss value created; seat pricing is predictable but tracks access rather than agent activity. The pricing metric should determine cost monitoring: resolution vendors need to track cost per resolution, since falling resolution rates can erode margins while cost per conversation looks stable. Feature and cohort level analysis can reveal losses hidden in aggregate results. In one reported case, moving a money-losing proactive feature to a smaller model saved about $800,000 annually. The author recommends learning what value customers experience, choosing a fitting model, and building cost monitoring around it from the start; no universal pricing model is offered.

### Source excerpt

Unlike traditional SaaS, AI agents deliver uncertain value at a cost that keeps shifting, and that combination breaks the pricing logic SaaS was built on.

## Printerpix and the not so private photo album

DevFeed: [Printerpix and the not so private photo album](<https://devfeed.tech/articles/printerpix-and-the-not-so-private-photo-album-92783.md>)

Original publisher: [Read original article](<https://www.pentestpartners.com/security-blog/printerpix-and-the-not-so-private-photo-album/>)

Author: Alex Wallace

Published: 2026-10-21T13:42:24Z

Content type: news

Language: en

Sources: [Pen Test Partners](<https://devfeed.tech/sources/pen-test-partners.md>)

Topics: [hosting](<https://devfeed.tech/topics/hosting.md>), [Support](<https://devfeed.tech/topics/support.md>), [Image](<https://devfeed.tech/topics/image.md>), [Image processing](<https://devfeed.tech/topics/image-processing.md>)

Tags: [access-tokens](<https://devfeed.tech/tags/access-tokens.md>), [accessible](<https://devfeed.tech/tags/accessible.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-s3](<https://devfeed.tech/tags/aws-s3.md>), [books](<https://devfeed.tech/tags/books.md>), [bounty](<https://devfeed.tech/tags/bounty.md>), [bug](<https://devfeed.tech/tags/bug.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-security](<https://devfeed.tech/tags/cloud-security.md>), [cloud-storage](<https://devfeed.tech/tags/cloud-storage.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [data](<https://devfeed.tech/tags/data.md>), [dns](<https://devfeed.tech/tags/dns.md>), [download](<https://devfeed.tech/tags/download.md>), [exploit](<https://devfeed.tech/tags/exploit.md>), [photo](<https://devfeed.tech/tags/photo.md>), [photos](<https://devfeed.tech/tags/photos.md>), [printing](<https://devfeed.tech/tags/printing.md>), [private](<https://devfeed.tech/tags/private.md>), [services](<https://devfeed.tech/tags/services.md>), [vulnerabilities-and-disclosures](<https://devfeed.tech/tags/vulnerabilities-and-disclosures.md>)

### AI overview

Printerpix photos uploaded for editing were stored in an S3 bucket whose public index listed file keys, allowing internet users to enumerate and view images. The upload flow used signed, token-protected URLs, but later image access was public; complex filenames did not prevent discovery because indexing exposed the keys. The researcher reported millions of files dating back to 2021, though the excerpt presents this as an estimate. Disclosure was hampered by the lack of a security contact and misrouted customer-support emails. The author says Printerpix moved image storage to a content delivery network that does not allow indexing; the issue was found resolved on 20 October 2026, without notification to the researcher.

### Source excerpt

TL;DR A bucketful of photos There are many services that allow digital photos to be converted to prints. These include a plethora of other printing services so that customer photos may be converted to picture books, mugs, magnetic photo tiles, clothing or even a throw. With the combination of destination, these services allow photos to [...] The post Printerpix and the not so private photo album appeared first on Pen Test Partners.

## How Fraudulent Airline Support Numbers Exploit Search Results and AI-Generated Answers

DevFeed: [How Fraudulent Airline Support Numbers Exploit Search Results and AI-Generated Answers](<https://devfeed.tech/articles/how-fraudulent-airline-support-numbers-exploit-search-results-and-ai-generated-answers-92891.md>)

Original publisher: [Read original article](<https://blog.checkpoint.com/research/how-fraudulent-airline-support-numbers-exploit-search-results-and-ai-generated-answers/>)

Author: laurama@checkpoint.com

Published: 2026-10-21T13:00:48Z

Content type: news

Language: en

Sources: [Check Point Blog](<https://devfeed.tech/sources/check-point-blog.md>)

Topics: [search-engines](<https://devfeed.tech/topics/search-engines.md>), [Search engine optimization (SEO)](<https://devfeed.tech/topics/seo.md>), [online-fraud-prevention](<https://devfeed.tech/topics/online-fraud-prevention.md>), [Vishing](<https://devfeed.tech/topics/vishing.md>), [Threat monitoring](<https://devfeed.tech/topics/threat-monitoring.md>), [Instagram](<https://devfeed.tech/topics/instagram.md>), [X (Twitter)](<https://devfeed.tech/topics/twitter.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [airline](<https://devfeed.tech/tags/airline.md>), [campaign](<https://devfeed.tech/tags/campaign.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [exploit](<https://devfeed.tech/tags/exploit.md>), [research](<https://devfeed.tech/tags/research.md>), [results](<https://devfeed.tech/tags/results.md>), [search](<https://devfeed.tech/tags/search.md>)

### AI overview

Check Point researchers describe an active campaign that uses mass-produced PDFs and social posts to place fraudulent travel support numbers where travelers search for help. The materials impersonate more than 20 brands and appear on trusted university, government, and commercial sites, often through public upload features. Repeated numbers and search-focused titles help the content appear in search results, social posts, and AI-generated summaries. The five-stage operation manufactures and seeds the content, relies on indexing and trusted domains to attract callers, then may seek fees or booking and payment data. The PDFs contain no malicious code, so scans by 61-64 antivirus engines did not detect them; the threat is the fraudulent information and borrowed credibility. The excerpt advises organizations to monitor public channels and restrict or review exposed uploads, and travelers to use contact details from official sites, apps, or booking confirmations.

### Source excerpt

A canceled flight. A missed connection. A refund that has yet to arrive. When travelers need urgent help, they often turn to search engines, social media, or AI tools to find a customer support number. Threat actors are increasingly exploiting that moment of urgency. In our previous investigation, Scammers Are Watching Airline Complaints and Posing as Customer Support, Check Point Exposure Management researchers uncovered scammers monitoring airline complaints on social media and impersonating customer support representatives. Check Point Exposure Management researchers identified a related campaign that targets an earlier point in the support journey: the search for help itself. Threat [...] The post How Fraudulent Airline Support Numbers Exploit Search Results and AI-Generated Answers appeared first on Check Point Blog.

## Is the ad a scam? How to tell before you click

DevFeed: [Is the ad a scam? How to tell before you click](<https://devfeed.tech/articles/is-the-ad-a-scam-how-to-tell-before-you-click-91866.md>)

Original publisher: [Read original article](<https://www.welivesecurity.com/en/social-media/is-ad-scam-how-tell-before-click/>)

Author: Phil Muncaster

Published: 2026-10-20T09:00:00Z

Content type: article

Language: en

Sources: [WeLiveSecurity](<https://devfeed.tech/sources/welivesecurity.md>)

Topics: [fake account fraud](<https://devfeed.tech/topics/fake-account-fraud.md>), [Credential theft](<https://devfeed.tech/topics/credential-theft.md>), [trust](<https://devfeed.tech/topics/trust.md>)

Tags: [2025](<https://devfeed.tech/tags/2025.md>), [2fa](<https://devfeed.tech/tags/2fa.md>), [ads](<https://devfeed.tech/tags/ads.md>), [advertising](<https://devfeed.tech/tags/advertising.md>), [ai](<https://devfeed.tech/tags/ai.md>), [app](<https://devfeed.tech/tags/app.md>), [apps](<https://devfeed.tech/tags/apps.md>), [bank](<https://devfeed.tech/tags/bank.md>), [banks](<https://devfeed.tech/tags/banks.md>), [browser](<https://devfeed.tech/tags/browser.md>), [card](<https://devfeed.tech/tags/card.md>), [chargeback](<https://devfeed.tech/tags/chargeback.md>), [computer](<https://devfeed.tech/tags/computer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [media](<https://devfeed.tech/tags/media.md>), [platform](<https://devfeed.tech/tags/platform.md>), [scam](<https://devfeed.tech/tags/scam.md>), [social-media](<https://devfeed.tech/tags/social-media.md>), [trust](<https://devfeed.tech/tags/trust.md>)

### AI overview

Scam ads can appear on trusted platforms, and automated ad tools may help generate and target them without confirming legitimacy. Watch for urgency, implausible offers, requests for sensitive information, suspicious support numbers, pressure to download software, sensational clickbait, and celebrity endorsements that may be deepfaked. Ads can lead to malware, phishing, fake purchases or investments, and theft of payment details. If you engage, stop contact with the scammer, monitor accounts, contact your bank or card provider promptly, change exposed passwords and enable 2FA, scan downloaded files with reputable security software, preserve evidence, and report the ad and scam. Depending on location and payment type, an APP claim or card chargeback may be possible; social platforms are not generally legally required to reimburse victims, and the FTC is considering possible platform duties.

### Source excerpt

You can trust a social media platform and still get scammed by an ad running on it

## How to handle token limits in long AI support conversations

DevFeed: [How to handle token limits in long AI support conversations](<https://devfeed.tech/articles/how-to-handle-token-limits-in-long-ai-support-conversations-92610.md>)

Original publisher: [Read original article](<https://www.twilio.com/en-us/blog/insights/token-limits-ai-support-conversations>)

Author: Jesse Sumrak

Published: 2026-10-20T00:00:00Z

Content type: article

Language: en

Sources: [Twilio](<https://devfeed.tech/sources/twilio-blog.md>), [Twilio Blog](<https://devfeed.tech/sources/twilio-blog-2.md>)

Topics: [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [context window](<https://devfeed.tech/topics/context-window.md>), [Conversational AI](<https://devfeed.tech/topics/conversational-ai.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Support](<https://devfeed.tech/topics/support.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [context-window](<https://devfeed.tech/tags/context-window.md>), [conversation-memory](<https://devfeed.tech/tags/conversation-memory.md>), [conversations](<https://devfeed.tech/tags/conversations.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [industry-insights](<https://devfeed.tech/tags/industry-insights.md>), [limits](<https://devfeed.tech/tags/limits.md>), [llm](<https://devfeed.tech/tags/llm.md>), [support](<https://devfeed.tech/tags/support.md>), [token](<https://devfeed.tech/tags/token.md>), [twilio](<https://devfeed.tech/tags/twilio.md>)

### AI overview

Long support chats can exceed an LLM's context window, causing requests to fail or old turns--including the customer's original issue--to be dropped. Resending the full history also raises token costs and can slow responses. The article outlines four tactics: sliding windows keep recent turns but lose early facts; rolling summaries preserve the gist but may omit identifiers unless instructed; retrieval fetches relevant stored facts but requires extraction and storage; structured state efficiently tracks known fields but misses details not anticipated. It recommends combining structured state, a sliding window, and retrieval. Twilio Agent Connect leaves live-history trimming to the agent's code, while Conversation Memory stores observations and summaries and recalls relevant past context. Recall supports semantic and lexical search, with limits and query options to bound results. On messaging it runs each turn; on voice, developers choose the frequency to balance fresh context against speed. The article notes that larger context windows do not remove the cost of resending history.

### Source excerpt

Long support conversations outgrow an LLM's context window. Learn 4 ways to manage token limits and how to handle context across conversations with Twilio.

[Next page](<https://devfeed.tech/tags/customer-support.md?cursor=WyIyMDI2LTEwLTIwVDAwOjAwOjAwKzAwOjAwIiwgIjM3MWY4YjVhLWEyOGYtNDBkOS05OTFkLWRjZjBiODg4ZDE0NSJd>)