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.
Gorgias AI Agent Resolves Customer Tickets Autonomously
Gorgias, an e-commerce helpdesk platform, launched Gaia, an AI Agent built to autonomously resolve customer tickets: handling returns, cancellations, and subscription changes within merchant-set guardrails. The Agent can now read and interpret images sent through Chat, enabling it to process visual context such as screenshots, product photos, or damage images. Gorgias also added a Metrics Glossary inside Analytics, providing a single reference that defines how each metric in the platform is calculated.
The image-reading capability addresses a gap that has limited AI ticket resolution in e-commerce: many customer issues arrive with visual evidence. Without the ability to read those images, the Agent had to escalate or ignore them. That changes here. For teams already running Gaia, the Metrics Glossary reduces the back-and-forth that comes when Support Specialists interpret the same metric differently, particularly during QBRs or when building dashboards.
Image reading is only as useful as the customer data behind it. If Gaia lacks context about a product catalog or order details, the image alone may not be enough to resolve the ticket. Teams should test this in lower-stakes ticket types first and set clear escalation rules for cases where visual evidence is ambiguous. Teams that have built custom metric definitions outside the platform will need to reconcile those against Gorgias’s official definitions before relying on dashboards for operational decisions.
Plain Launches Custom Skills for AI Agents
Plain, a customer support platform built for technical teams, announced Custom Skills — a way to extend the AI Agent’s capabilities by connecting it to internal APIs, tools, and workflows. Custom Skills let support teams define what actions the Agent can take, such as triggering a refund, pulling account data, or creating an internal task, without writing code from scratch.
The ability to define custom Agent behaviors without engineering dependency is a meaningful shift for CX ops teams. Most AI Agent platforms require engineering work to configure integrations. Custom Skills move more of that configuration into the hands of support operators, which matters for teams that have been blocked on roadmap prioritization to get their Agent connected to internal systems.
Custom Skills require access to internal APIs, which means security review and API key management are prerequisites. Teams without documented internal APIs will hit that dependency before they can build skills. There is also a reliability question: if a custom skill calls an external system that has downtime, the Agent’s behavior in that scenario needs to be designed explicitly. A skill that fails silently can create worse customer outcomes than no skill at all.
Guru Expands Knowledge Agents Across Enterprise Teams
Guru, a knowledge management platform, announced an expansion of its Knowledge Agents product, positioning it as an AI layer that surfaces verified, role-appropriate knowledge across support, sales, and operations teams. Knowledge Agents pull from connected sources (tickets, docs, Slack, CRMs) and return answers grounded in content that has been reviewed and verified by a human subject-matter expert.
Guru’s verification model is central to how it differentiates from general AI search. Support teams have used AI-assisted search for years, but the recurring problem is answers drawn from outdated or unreviewed content. Guru’s quality automation flags content based on usage signals and verification status, so Agents are not surfacing KB articles that were last reviewed two years ago. For teams running large knowledge bases, automating verification rather than relying on manual review cycles reduces the operational overhead of maintaining accurate content.
Knowledge Agents are only as accurate as the content they reference. Teams with inconsistent, partially migrated, or outdated knowledge bases will see that reflected in answer quality. Guru’s own documentation is clear: the platform cannot manufacture accurate answers from poor source material. Access control is a second consideration: Knowledge Agents inherit source permissions, which means mismatched permissions between connected sources need to be audited before rollout to avoid surfacing restricted content to the wrong audience.
Linear Agent Automates Recurring Work with Loops
Linear, a project management platform, introduced Loops, a feature that enables the Linear Agent to run recurring automated workflows on a schedule, removing manual triggers from repetitive team processes such as standups, retros, and status updates. The Agent can also now edit documents and project descriptions directly, with author names and version history tracking whether a human or a Loop made each change.
For support operations teams that use Linear to manage projects, backlogs, or cross-functional requests, Loops reduces the coordination overhead of recurring processes. A Loop can draft a weekly update, pull recent ticket metrics, and post it to a project before the team’s Monday sync without anyone initiating it. The document editing capability adds AI-generated writing to project artifacts, which is useful for teams that struggle to keep project briefs and specs current as work evolves.
Automated recurring workflows require clear ownership. A Loop that drafts an update no one reviews can create false signals about project health. Teams should establish which Loops are fully automated versus which require a human review step before the output is visible to stakeholders. The Agent’s document editing feature also raises a governance question: teams need version history habits in place before giving the Agent write access to important documents.
Attio Adds Live Formula Attributes to CRM Records
Attio, a CRM platform, launched Formula Attributes: a feature that allows teams to build custom calculated fields directly on CRM records using live data from across the workspace. Formula Attributes can compute revenue figures, date differences, scoring models, or any value derived from existing record data, and update automatically as the underlying data changes.
For CX and CS teams that use Attio to track account health or customer data, Formula Attributes remove a common workaround: exporting data to a spreadsheet to run calculations, then re-importing results. Fields like days since last contact, open ticket count, or average resolution time per account can now live directly on the customer record and stay current without manual updates. This makes account views more reliable for teams that use CRM data to prioritize outreach or escalations.
Formula Attributes are only as reliable as the underlying data they reference. If input fields are inconsistently populated across the account base, formula outputs will be unreliable and potentially misleading. Teams should audit data quality on input fields before building formulas that feed into operational decisions. Formulas that chain multiple calculated fields together can also become difficult to debug when results look wrong, so documenting each formula’s logic is worth the upfront time.
Zipchat Connects to Shopify Sidekick, Adds Customer Memory
Zipchat, an AI Agent platform for e-commerce, integrated with Shopify Sidekick so merchants can manage, debug, and prompt their AI Agent directly from the Shopify admin without switching tools. Zipchat also launched Customer Profiles: persistent memory that tracks each customer’s history across conversations and channels, automatically extracting key facts from each interaction and recognizing the same contact whether they reach out via chat, email, or WhatsApp.
The Shopify Sidekick integration lowers the barrier to managing the AI Agent for merchants who spend most of their time in Shopify admin. Instead of context-switching between platforms to adjust prompts or debug a conversation, the work happens in the interface they are already in. Customer Profiles address a structural gap in most e-commerce AI deployments: the Agent starts each conversation from zero, with no memory of prior interactions. Persistent memory means a customer who reported a defective product last week does not have to re-explain the context when they follow up.
Customer Profiles are built from conversation history, which means profile quality depends on what the Agent has observed and extracted correctly. Early profiles on new customers will be sparse. Teams should also consider what happens when extracted memories are wrong. If the Agent incorrectly records a preference or account detail, it is not yet clear how easily that can be corrected. Privacy implications of storing customer behavior summaries should be reviewed against applicable data regulations, particularly for teams serving EU customers under GDPR.
Miro Adds Native Video Recording to the Canvas
Miro, a collaborative workspace platform, announced native video recording directly within boards, allowing teams to record walkthroughs, async updates, and presentations without leaving the canvas or opening a separate recording tool. Video clips embed directly on the board and can be attached to specific objects, frames, or sections of the workspace.
For CX and support operations teams that use Miro for process documentation or cross-functional alignment, native video removes a friction point: the need to record elsewhere, upload the file, and link it back. A team lead can record a walkthrough of a new escalation process directly on the Miro board where the process is documented, and the video stays with the artifact. As async video increasingly replaces status meetings, having that capability embedded in the workspace where the work lives reduces tool sprawl.
Video recording features add storage implications that enterprise teams should review, particularly on plans with storage limits. Miro is primarily a product and engineering collaboration tool; CX teams that are heavy Miro users tend to be in larger organizations. Teams considering this for process documentation should evaluate whether video walkthroughs will actually be consumed, a recording attached to a board is only useful if teammates know it is there and have a habit of opening it.
Fin Adds Telegram Support and Webhook Pause Capability
Fin/Intercom, an AI-first customer service platform, shipped a series of updates including Telegram as a new supported channel for Fin, and a Wait for Webhook capability that lets Fin pause mid-procedure while an external system completes its work (such as an identity check, payment, or bank linking flow) then resume automatically. Other updates include a redesigned Help Center with persistent sidebar navigation, bulk actions across all conversations matching a search filter, AI-generated pre-fill when converting conversations to tickets, voice transcription in the Messenger, snoozed-time metrics in Custom Reports, and custom business caller ID for outbound calls.
The Wait for Webhook capability is the most operationally significant of this group. Until now, Fin had to hand off to a Support Specialist whenever a procedure depended on an external system completing a step first. That limitation forced escalations in workflows the Agent should otherwise be able to handle end to end: identity verification, payment confirmation, or any multi-system transaction. Removing that constraint expands the range of issues Fin can fully automate. The bulk action update addresses a practical inbox management problem: large backlogs have required selecting and acting on conversations one page at a time. The Telegram channel expansion matters for teams serving consumer audiences where Telegram is a primary communication channel, particularly in Europe and the Middle East.
Wait for Webhook depends on external systems sending a reliable callback signal. If a third-party service times out or sends a malformed response, Fin escalates to a Support Specialist — which may be the right behavior, but teams should map out those failure scenarios before building workflows that rely on it heavily. The redesigned Help Center is opt-in and reversible, which reduces rollout risk; teams should still preview the new layout with their own content before switching, particularly if they have heavily customized navigation. The bulk action feature runs large jobs in the background with a completion notification — teams should set clear policies about who has permission to run bulk operations, since a misapplied bulk close or assignment is time-consuming to reverse.
GitBook Agent Now Enforces Team Style Guides
GitBook, a documentation platform, launched Style Guides, a feature that allows teams to define their writing rules in a dedicated document that the GitBook Agent automatically enforces whenever it writes, edits, or reviews documentation. GitBook also redesigned its app in a sites-first update, organizing each site’s content, settings, and tools together in one sidebar with a new All Content view for finding everything across the organization.
For support and knowledge teams managing large documentation sets with multiple contributors, style consistency is an ongoing maintenance problem. A style guide that humans must remember to follow is different from one that an Agent enforces on every edit. GitBook’s implementation goes beyond a linter. The Agent flags violations with specific rule IDs and suggests fixes, turning the style guide into an active editorial layer rather than a reference document. Teams that produce high volumes of AI-assisted documentation will find this useful: it constrains what the Agent writes to match established voice, terminology, and structure rather than defaulting to generic patterns.
Style Guides are in early access and rolling out gradually, so not all organizations have access yet. The feature’s enforcement quality depends entirely on how the style guide is written — vague rules produce vague enforcement. GitBook’s own documentation is explicit: the Agent enforces only what is written in the style guide, not general good writing practice. Teams investing in this should plan time to write the style guide carefully, particularly the word list and numbered enforceable rules. The Agent reviews up to five violations per pass, which may be limiting for documentation with significant style debt.
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.
