Customer Experience News is a weekly newsletter about the most important news and discussions for Customer Experience and Customer Support Leaders.
This is all the weekly news you need in around 5 minutes.
A year ago, an AI Agent in customer support meant a chatbot that answered a question from a knowledge base, and handed off anything it did not recognize.
Across the CX products this year, Agents started fixing their own mistakes, closing sales inside a support conversation, working through business logic that used to require a static form, coordinating with sales and success on the same account, investigating a ticket before a Support Specialist ever opens it, dialing outside companies and sitting through their hold music, and even patching a software bug before an engineer sees the report.
Read the full article: How AI Agents Evolved for Customer Support in 2026
Zendesk Launches Specialized Agents for Service
Zendesk, a customer service platform, announced Specialized Agents, a set of purpose-built Agents that combine industry expertise, business context, and connected systems to resolve specific customer needs rather than only answering questions. The Agents come in two forms: Industry Agents built for common workflows in a given sector, and Custom Agents built around a company’s own processes and systems.
Industry Agents give teams a faster starting point for automating requests like order tracking, returns, refunds, and account updates, while Custom Agents let teams keep automation aligned to workflows that do not fit an out-of-the-box template. Zendesk positions escalation as preserving full conversation context when a request moves to a Support Specialist, which matters for teams concerned about customers repeating information.
Before rolling Specialized Agents into production, support leaders should test which existing workflows an Industry Agent actually covers against which still need a Custom Agent build, since the line between the two categories will vary by industry and by how far a company’s processes depart from common patterns. Teams should also confirm what customization options are available to Custom Agents, including instructions, policies, and escalation rules, before committing to a build.
Decagon Adds Governance to Self-Improving Agents
Decagon, a conversational Agent platform, announced expanded governance for Duet Autopilot, the self-improving Agent it introduced earlier this year. The update adds Vault, a centralized context layer for grounding Autopilot’s changes in a team’s own documentation and guardrails, along with two parallel improvement tracks and outcome-based measurement.
The update separates two kinds of Agent improvement work that previously competed for the same review queue: trigger-based runs that fix errors the moment an alert fires, and a separate audit process aimed at deflection and CSAT gains that error triage tends to crowd out. Teams see a projected impact on deflection before a change ships and measured results after, giving reviewers more to work with than judgment alone.
This update is built for teams already running Decagon’s Duet Agent, not a general-purpose audit tool, so the governance benefit depends on how much documentation and context a team has already loaded into Vault. Support leaders evaluating self-improving Agents more broadly should treat the oversight shift Decagon describes as a preview of a wider pattern: less review of individual Agent changes, more upfront work specifying objectives and guardrails correctly.
Customer.io Agent Delivers Behavioral Analytics Reports
Customer.io, a customer engagement platform, announced its Agent can now analyze a workspace’s messaging and behavioral data directly, producing custom reports that rank automations, compare inbox providers, track conversions, and report on site traffic and audience growth.
Teams that rely on Customer.io for lifecycle and transactional messaging gain a way to ask plain-language questions about their own data instead of building a custom report from scratch each time a stakeholder asks, which could reduce the analyst time spent on recurring requests about campaign or automation performance.
The feature works through Customer.io’s existing CLI and MCP tools, so teams need those already connected to use it. The quality of the Agent’s answers will track the quality of a workspace’s existing event and attribute data, so teams with inconsistent tracking should expect the reports to reflect that inconsistency rather than correct it.
Salesforce Completes Acquisition of Fin (Intercom)
Fin, a customer agent platform, announced that Salesforce completed its acquisition of the company, bringing Fin’s Agent technology and technical AI team into Salesforce’s customer service portfolio alongside Agentforce. Fin, formerly Intercom’s Fin, serves more than 30,000 companies and resolves customer queries across chat, email, WhatsApp, SMS, voice, and Slack.
For support teams already running Fin, the immediate operational picture stays the same: Fin continues serving its existing customer base and advancing its own Agent capabilities rather than being folded into Agentforce outright. The acquisition does widen the paths into Fin’s technology inside Salesforce, since teams now have two distinct options under one vendor: a fast-to-deploy Agent that layers onto a help desk already in place, or Agentforce’s deeper, custom-built approach.
Leaders running Fin today should watch how the product roadmaps for Fin and Agentforce converge or stay separate over the coming quarters, since maintaining two Agent platforms inside one company is an unusual long-term structure. Teams evaluating either platform now face a harder decision between two options from the same vendor, and pricing, account management, and support structures during the integration period are worth confirming directly with a representative before committing budget.
ReadMe Syncs Changelogs With Git Repositories
ReadMe, a documentation platform, announced that changelog posts now sync bi-directionally with a connected git repository, with the feature rolling out gradually across projects.
Documentation and support teams who maintain a public changelog alongside engineering release notes gain one less manual step: a changelog written in git now flows into ReadMe automatically instead of requiring a separate copy into the docs platform, which reduces the chance that a public changelog drifts out of step with what actually shipped.
Because the rollout is gradual, teams should confirm the feature is active on their project before relying on it, and should plan for a short overlap period where edits made directly in ReadMe and edits made in git could conflict until the sync settles into a steady rhythm.
Linear Expands Loops for Product Management
Linear, a project and issue tracking platform, announced that Loops, its recurring Agent workflows, now respond to more workspace activity, including changes to initiatives, projects, and cycles, and can edit Linear documents and post updates to Slack.
For support teams that route escalations and bug reports into Linear, Loops close a gap that has mostly been handled by hand: when a project’s target date or scope changes, a loop can update the relevant plan document and post a Slack message explaining what changed and who needs to act, instead of a support operations lead tracking that update down after the fact.
Linear extended introductory Loops credits through the end of the year, giving teams a window to test the feature against real escalation and handoff workflows before deciding whether it earns a longer-term commitment. The automatic Slack updates are worth reviewing for tone and audience before turning them on for a broader channel.
Notion Adds Model Controls for Agents
Notion, a workspace and knowledge management platform, announced model controls that let workspace owners choose which AI models are available to Notion Agent and Custom Agents separately, along with a default model for Custom Agents.
Support and knowledge management teams running Notion as a source of truth gain a way to standardize which models touch their content, which matters for teams that have built Custom Agents against internal support documentation and want predictable costs and consistent output rather than whatever model an individual teammate happens to select.
The feature is available on Business and Enterprise plans only, so teams on lower tiers will not see the setting. Workspace admins should audit existing Custom Agents before restricting the model roster, since an Agent built around one model’s behavior can respond differently once the available models change.
Now this is:
Strategic Support for CX Leaders.
You’ve got ambitious Support targets and new metrics but you’re not sure what to prioritize first.
The list is long, the queue is getting longer, and you don’t have time to step back and think about CX strategically.
What if you could pressure-test your thinking with someone who’s spent 20 years building Customer Support operations?
No pitch, just a conversation.
