Customer Experience News is a weekly newsletter and video 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.
Today’s headlines are from:
Sierra Co-Creates Personal Agent Protocol
Sierra, a conversational AI platform, announced the Personal Agent Protocol, an open standard that defines how consumers’ personal agents interact with businesses. Sierra developed the protocol with Meta, and Genesys, Instinct, Rocket, Shopify, Stripe, and Walmart are named as industry partners. A personal agent can reach a business through its website, through its APIs using MCP and OpenAPI, or through the business’s own Agent, with OAuth handling authorization.
Support teams may soon receive requests that arrive from a customer’s personal agent as well as from the customer directly. The protocol lets consumers grant read-only or write access to their accounts. Sessions begin as guest access and move to authenticated access when account data is needed, which mirrors the way many teams already separate general questions from account-specific ones.
The version 0.1 specification was announced as forthcoming, with a reference implementation and design workshops planned, so the standard is at an early stage. The current version covers authentication, visibility, and basic access control. Granular permissions, push notifications, and payment extensions are listed as future work. CX leaders can start by listing which account actions their APIs and Agents already expose, since those determine what a personal agent could request.
Pylon Introduces Nexus 1.0 Agent Harness
Pylon, a B2B customer support platform, announced Nexus 1.0, which it describes as a versioned agent harness for B2B support. Nexus orchestrates AI models using three precomputed context layers: support history, account context, and product knowledge. Pylon states that it uses both frontier and open-weight models and routes between them against quality, latency, cost, and control targets.
Nexus ingests images, videos, and files with parsing and annotation, and it keeps persistent sessions across the life of an issue. An Agent working a technical ticket can carry context from the first report through resolution, and results from background tasks feed back into the session. Pylon also lists output review against policies such as tone and PII requirements, and citation validation before a customer-facing response is sent.
Runtime controls include bounded limits on time, cost, and output, detection of stalled investigations and repeated failed calls, and human approval for sensitive actions. Teams will need to decide which actions count as consequential and who approves them. The Nexus page does not state pricing or an availability timeline, so both are worth confirming with Pylon. The context layers draw on existing support history and product documentation, so an audit of that material is a sensible step before relying on cited answers.
Fin/Intercom Macros Add Fill-In Fields
Intercom, a customer service platform, announced fill-in fields for macros. When a teammate selects a macro that includes fields, a short form asks for the details and only the finished reply is placed in the composer, ready to review and send. Fields can be text, long text, or choice, each optional or required, and macros can include optional sections that a teammate switches on or off while watching a live preview.
A macro that once needed hand-edited placeholders can now collect the variable details up front. For a refund confirmation, a Support Specialist could enter an order number and amount once and review the finished message before sending. Teams that adopted a separate text expansion tool to personalize replies may be able to retire it.
Macros with fields are created in Settings under Macros with the Insert field control, or from the inbox with Cmd+K. Existing macros need to be revisited to add fields, so the highest volume templates are a reasonable starting point. Required fields support consistency, and optional sections allow one macro to cover variants that would otherwise need several near duplicates. The announcement does not name plan availability, so check your plan before scheduling a rollout.
Zendesk October Release Expands Agent Tooling
Zendesk, a customer service platform, announced its October releases. Voice Agents reached general availability for Zendesk Voice customers, with native phone automation, real-time conversations, handoff to Support Specialists with transcripts and summaries, and support for 53 languages. Email Agents gained embedded templates, segments in procedures, a welcome reply, and a knowledge reply. Action flows gained wait steps, action logs, cloning, and a higher step limit.
Wait steps pause an action flow from 1 minute to 31 days, which supports follow-ups and multi-stage escalations without manual tracking. Action logs show runs in real time with 30 days of history and step-by-step traces, which shortens diagnosis when a flow fails, and the step limit rose from 50 to 60. The Universal Connector is generally available for bringing REST and JSON API content into Knowledge for use by Agents, Copilot, and search, and dedicated Helpjuice, WordPress, and ReadMe connectors reached general availability as well. In workforce management, forecast scenario imports and simplified schedule imports reduce manual reconciliation of CSV data.
Several changes call for administrator attention. Sender authentication now defaults to Minimal for all accounts, with a staged rollout for eligible accounts that had it disabled. Accounts with inactivity timeouts of 10 hours or more migrate to an 8-hour timeout, and existing sessions longer than 12 hours are forcibly expired, so Support Specialists will need to sign in again. Voicemail requires a stack upgrade. Draft mode for Auto Assist procedures lets teams save and test procedure updates without affecting live suggestions, which makes it a useful place to rehearse changes. Voice Agents depend on Zendesk Voice, and handoff quality will depend on how well transcripts and summaries fit current escalation procedures.
Gorgias Adds Scheduled Holiday Hours
Gorgias, an ecommerce customer support platform, announced holiday hours scheduling. Teams can set a full-day closure or special opening hours for a date or date range from Settings, in the Business and holiday hours area under the Holidays & exceptions tab. Settings can apply to all integrations or to selected ones such as individual stores or regions, and a yearly recurrence option is available.
Holiday hours carry through to Chat availability, Voice routing, and Rules. When an SLA policy pauses timers outside business hours, the pause also accounts for holiday hours, which keeps SLA reporting aligned with the days a team is closed. Regular weekly hours stay in place outside the exceptions.
Teams with several stores or regions can scope each closure to the relevant integrations, so setup starts with a list of which storefronts follow which calendar. Recurring entries reduce yearly re-entry, and holidays whose dates move from year to year should be checked annually. Many holiday schedules fall near peak ecommerce periods, so entering them early leaves time to test Rules and Voice routing. The announcement does not state plan availability.
Help Scout Redesigns Conversation Sidebar
Help Scout, a customer support platform, announced a redesigned conversation sidebar in Inbox. Conversation details such as status, assignee, SLA, Topics, and Sentiment now sit alongside customer and company information, conversation history, Custom Fields, and apps. Custom Fields are editable from the sidebar, and customer and company profiles open in a unified view.
Support Specialists can see context and make edits without moving between screens. The sidebar can be resized, and sections can be collapsed and reordered, so a team can put the details it uses most at the top. A team working to contractual response times, for example, could place SLA and Custom Fields first.
Teams that rely on Custom Fields for routing or reporting should confirm which fields are worth editing from the sidebar and who may change them. Because Support Specialists can reorder sections, a shared recommendation on section order may help keep coaching sessions and screen shares consistent. The announcement lists no plan restrictions or limitations.
Front Brings Agents Into Team Workflows
Front, a shared inbox and customer communication platform, announced capabilities that let teams run Agents as part of the team inside Front. Supported Agents include custom Autopilot Agents built for specific brands, teams, or regions, and external Agents from Claude, Codex, Sierra, and other providers. The update also adds Compass, Copilot Routines, and an Accounts experience.
Each Agent gets clear ownership and full context, with performance visible next to the work Support Specialists do. Compass scores Agent replies for quality and sentiment and offers improvement recommendations, which gives managers one place to review human and Agent output. Copilot Routines deliver alerts for churn and expansion, along with meeting briefs, to the inbox. The Accounts experience brings omnichannel activity and recommended next steps for a business into one view.
Compass scoring is automated, so leads may want to compare it with their own QA scorecards before using it in performance reviews. Bringing in an external Agent still calls for a named owner and an escalation path. The announcement does not describe usage limits or integration boundaries, so those are worth confirming before rollout.
Decagon Co-Develops PACT for Personal Agents
Decagon, an AI customer service platform, announced the Personal Agent Consent & Trust Protocol, known as PACT, an open-source standard for personal agents that interact with business Agents under explicit customer authorization. Decagon developed it with Instinct, and it builds on the A2A protocol and OAuth. The specification is available at openpactprotocol.org, with code and documentation on GitHub.
PACT separates a personal agent’s identity from its authority to act on a customer’s account. The personal agent finds the business’s A2A Agent Card and authenticates with signed JWTs. The customer then logs in directly with the business through the OAuth device authorization flow and approves specific permissions, so credentials are never shared with the personal agent. Later requests carry both identity and delegation tokens, and the provider verifies authorization before acting within the approved scopes.
Sierra and Decagon have each announced a protocol for this interaction, so CX leaders may see more than one approach emerge, and the announcements do not describe how the two relate. The Decagon page does not detail current limitations. Teams evaluating either one can ask their platform vendor which they plan to support, and can involve security and identity owners early, since consent and token handling sit outside most Support Operations responsibilities.
Ada Adds Voice Proactive Outreach
Ada, an AI customer service platform, announced Voice Proactive Outreach, which lets voice Agents start outbound phone calls. When a person answers, the Agent runs a linked Playbook, and the system includes answering machine detection. The Proactive Outreach feature, formerly called Proactives, also moved to its own configuration page at Config > AI AGENT > Outreach, and existing items continue to work.
A team could trigger a call from an event in its own systems and have the Agent follow a defined Playbook once someone picks up. Two integration paths are documented. The first uses the Conversations API endpoint POST /v2/conversations/proactive/, where Ada places the call and handles detection over the customer’s SIP infrastructure. The second lets a third-party contact center dialer transfer answered calls to Ada with a SIP INVITE that carries the X-Ada-Proactive-Id header.
Access to Voice Proactive Outreach is by request through the Ada team. Both integration paths depend on SIP infrastructure, which brings telephony owners into the project. Outbound calling carries regulatory and consent obligations that vary by region, so legal review of calling practices belongs in the plan before launch.
GitBook Adds Docs Quality Monitoring
GitBook, a documentation platform, announced a quality monitoring layer for published docs. An Improve section consolidates quality checks in one place, and an Analyze section provides rebuilt analytics reports with audience-specific data and period comparisons. Capabilities include content gap detection, broken link detection, style guide checks, and site analytics.
Content gap detection compares visitor questions and support data with existing content and ranks issues by severity, which gives knowledge owners a prioritized list tied to what people ask. Broken link detection finds URLs that return 404 errors and suggests redirects ranked by traffic, grouping related URLs under wildcard redirects. Style guide checks let GitBook Agent generate fixes automatically. Site analytics segment visits by human visitors, Agents, and AI Assistant interactions.
Style guide checks cover up to three spaces per audit. Content gap ranking draws on support data, so teams with inconsistent ticket tagging may need cleanup before the ranking is useful. Fixes generated by GitBook Agent still warrant review by a content owner before publishing.
Featurebase Redesigns Customer Portal
Featurebase, a customer feedback and support platform, announced a redesigned customer portal that brings Support, Help Center, Feedback, and Updates into one location, with unified search and an Ask AI function that returns answers with source links. The release also adds native API documentation, a redesigned Help Center, and new customization options.
API docs import from OpenAPI files or URLs, include code examples and an interactive playground, support multiple APIs and versions, and show a change review before publishing. Fibi, the AI assistant, can answer questions from published API docs. The Help Center gains a tabbed layout, cards, content tabs, steps, accordions, previous and next buttons, related articles, and options to copy a page as Markdown or open it in ChatGPT or Claude. User profiles consolidate requests, tickets, and chats with status indicators.
API docs are available on the Free plan and above. Teams moving existing docs should plan time to review imported OpenAPI content and the new layout. Customer sign-in methods are configurable, so the choice of methods is worth settling before Support and Feedback sit behind one portal. The release also adds llms.txt support for AI navigation, and teams can confirm which content that file exposes. The announcement does not state limits for the other features.
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.
