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.
Our main story today is about why customer support is starting to look like precrime.
Why Customer Support Is Starting to Look Like Precrime
In Minority Report, the precrime unit stopped murders before they happened. Three precogs floated in a pool, catching visions of violence days in advance, and a police unit acted on those visions before the crime occurred. The premise sounds like science fiction until you notice how much of customer support is moving in the same direction. A failed identity check, an abandoned cart, a series of clicks without actions happening: these are the precogs of customer experience, and more support platforms are starting to build systems that reach out before the customer files a ticket at all.
Fin’s recent update to its Procedures feature is one example of the pattern. A Procedure used to wait for a customer to start a conversation. Now a signal from a company’s own systems can trigger it directly, with the Agent already knowing why it is reaching out.
At Frame.io, I built a customer signal tagging system that let us flag friction as customers sent messages. This evolved into a dashboard showing issues as they happened – A failed upload, a stalled render, a repeated attempt at the same broken workflow.
At the time, the tech wasn’t advanced enough to predict issues while we could see it happening in real time. The value was about giving a Support Specialist enough context to open a conversation already understanding the problem, the same shift Fin is now trying to build directly into its product.
The operational case for this kind of proactive support is straightforward. When actions happen that aren’t supposed to, an AI chat pops up providing guidance towards the solution. Customers receive guidance before having to look for the solution themselves. Done well, this solves recurring issues in the moment rather than frustrating the customer to the point of reaching out for help.
But precrime had a flaw, and it is the same flaw that shows up whenever a system starts acting on a prediction instead of a fact. The precogs were not always right. Sometimes they disagreed with each other, and someone still got arrested for a future that never happened. A support system that reaches out based on a signal has the same failure mode, just with lower stakes. A customer who was not actually struggling gets an unsolicited message about a problem they were never having, and what was meant to feel helpful instead feels like the company is watching too closely and guessing wrong.
That gap between the signal and the customer’s actual state is where proactive support earns its reputation, one way or another. A well tuned signal, like the failed identity check Fin uses as its example, is close to unambiguous: something broke, and reaching out is almost certainly welcome. A vaguer signal, like a long pause on a pricing page, is a guess dressed up as a prediction, and guessing wrong at scale trains customers to distrust every proactive message that follows, including the ones that were right.
For CX leaders looking at this shift, the practical question is not whether to build proactive support but which signals earn the trigger. A system failure, a broken workflow, an identity check that failed: these are close enough to fact that acting on them costs little if the read is wrong. Behavioral signals that infer intent rather than observe a failure deserve more skepticism, and probably a lighter touch, until a team has enough data to know how often the read is actually correct. The teams getting this right are not the ones automating the most outreach. They are the ones being honest with themselves about which signals are precogs and which are just guesses.
Customer support has spent decades getting better at responding fast. The next year is going to be about deciding when it is right to respond before being asked at all, and building enough humility into that decision to know the signal worth acting on.
Gorgias Redesigns Chat to Drive Purchases
Gorgias, a helpdesk platform built for ecommerce brands, announced a redesign of its Chat product aimed at driving shoppers toward purchase rather than only answering questions. The new Chat includes clickable replies, contextual prompts, and shoppable product cards embedded directly in the conversation. Gorgias reported that customers click on products shown in Chat 36 percent more often under the new design, with 2.25 times more shoppers adding a product to cart directly from a conversation.
The redesign reflects a broader shift in what brands expect from a chat channel, since Gorgias said Chat is now growing 2.5 times faster than email for its customer base. Support Specialists working ecommerce queues gain a tool that can move a conversation toward checkout without leaving the chat window, which reduces the back and forth of pointing customers to a product page. Teams that treat Chat mainly as a deflection channel will need to reconsider how they staff and script it now that it carries commerce features.
Brands should audit their product catalog data before enabling shoppable cards, since incomplete or outdated product information will surface directly inside customer conversations. The reported click and cart lift figures are Gorgias’s own numbers and are likely to vary by catalog size and traffic source, so teams should track their own before and after metrics rather than assume the same lift. Support Specialists will need a short training pass on the new interface, since clickable replies and inline product cards change how a conversation is composed compared to the previous chat experience.
Notion Adds Developer Section to Workspace Sidebar
Notion, a workspace and knowledge platform used across product, support, and documentation teams, added a Developer section to the workspace sidebar. The new area centralizes management of Workers, connections, and personal access tokens, and includes a developer bar that surfaces IDs for pages, databases, blocks, workspaces, and users for one click copying.
Teams that have connected Notion to internal tools or built Custom Agents on top of Notion’s developer platform now have a single place to see what is deployed and review logs behind each run, rather than hunting through settings menus. For support and knowledge management teams building automations on top of Notion, such as syncing help center content or triggering workflows from ticket data, the developer bar removes a common friction point of manually locating object IDs for API calls.
Access to the Developer section requires enabling developer features under Settings, so it will not appear by default for most users. Teams with several people building integrations should agree on naming and cleanup conventions for connections and tokens now that they are visible in one place, since sprawl in this area has historically been invisible until something breaks.
Linear Agent Now Sets Up and Tests Code
Linear, an issue tracking platform increasingly used to manage AI coding work, updated its coding sessions so Linear Agent can set up, run, and test code before returning it for review. Coding sessions now automatically detect a codebase’s toolchain and install needed dependencies across languages including Python, Ruby, and Go, and Linear introduced lower, more transparent pricing for the AI credits these sessions consume.
Fewer handoffs is the practical benefit here, since a coding session that already runs and tests before returning work gives a human reviewer something closer to a finished change rather than a first draft. Teams that pin runtime versions or add setup scripts get more predictable results, since coding sessions will now respect that configuration instead of guessing at environment setup on every run.
Teams evaluating coding sessions for the first time should expect an adjustment period around AI credit usage now that pricing has changed, and should review the new pricing details before assuming past cost patterns will hold. Support and CX teams that use Linear primarily for issue tracking rather than code review will see limited direct impact from this release, though it is worth understanding whether engineering counterparts are adopting agentic coding sessions that touch the same issues customer facing teams file.
Sierra Adds Release Governance for AI Agents
Sierra, a platform for building AI Agents for customer experience, introduced release governance for changes made inside its Agent Studio. The feature adds checks, approvals, and staged rollouts to move a change from a builder’s workspace into a live customer conversation, applying standard software release discipline to agent configuration changes.
Teams running Agents in production have had limited formal controls over how a change to agent behavior gets tested and approved before it reaches customers, which creates risk when multiple people are editing the same agent. Staged rollouts let a team test a change against a portion of live traffic before committing to it fully, a meaningful step for organizations that have been reluctant to let non-engineering staff edit agent behavior directly.
Teams should define who holds approval authority for agent changes before turning this on, since the feature is only useful if the approval step reflects real accountability rather than becoming a formality. Organizations with a single person managing their Sierra agent will get less immediate value from staged rollouts than teams with multiple builders working on the same agent, though the audit trail benefit still applies.
Attio Ships Workflow, Sidebar, and Developer Updates
Attio, a CRM built around AI Agents and workflow automation, shipped a release covering new Workflows apps and blocks, the ability to rename system objects, a refreshed sidebar, developer platform updates, and dictation support inside Ask Attio. The apps and blocks additions expand what workflows can connect to and control, while renaming system objects lets teams match Attio’s default terminology to how their own organization talks about records.
For CX and revenue teams that use Attio alongside a support tool, the ability to rename system objects means the CRM can reflect internal terminology instead of forcing teams to translate between Attio’s defaults and their own vocabulary in every meeting and report. Expanded workflow blocks give teams more ways to trigger actions from customer data without custom code, which matters for smaller teams that do not have engineering time to spend on integration work.
Renaming system objects is workspace wide, so teams should agree on naming conventions before making the change, since a rename affects every user and every existing view built on that object. Workflow blocks that are new this release should be tested against a small segment before being applied broadly, given that workflow automation mistakes can propagate quickly across a CRM full of live customer records.
Capacity Expands Voice AI Across Channels
Capacity, a customer experience automation platform, expanded its voice offering with a neural voice AI Agent that draws on the same knowledge layer already powering its chat, SMS, email, and Support Specialist assist tools. The company positioned the update around cost, stating that live voice interactions run seven to thirteen dollars compared to fifty cents to two dollars for an interaction an AI Agent handles.
Voice is the channel customers escalate to when self-service and chat have already failed, so a voice Agent that shares context with other channels can pick up where those earlier attempts left off instead of starting the conversation over. Teams that have already trained a knowledge base for Capacity’s chat Agent do not need to build a separate one for voice, which lowers the setup cost of adding an AI voice channel.
The cost figures Capacity cites are the company’s own estimates and will not match every organization’s actual cost structure, so teams should model their own numbers using current contact center pricing before treating the comparison as a business case. Voice still requires complex issues to route to a human, so teams should plan for that handoff experience rather than assume the neural voice Agent replaces phone Support Specialists entirely.
Pendo Adds Session Replay for Jetpack Compose
Pendo, a product experience platform used by product and CX teams to track user behavior, added support for Session Replay in apps built with Jetpack Compose, extending replay coverage to Android’s newer low-code UI framework. The same mobile release also added support for images and animated GIFs inside tooltip guides.
Teams that have moved Android development to Jetpack Compose previously had a gap in session replay coverage, so this closes that gap for CX and product teams trying to understand where mobile users get stuck. Adding visuals to tooltip guides gives onboarding and support teams a way to show rather than describe a step, useful for features that are hard to explain in text alone.
Teams should confirm their mobile app has updated to a supported SDK version before expecting Compose based screens to appear in Session Replay, since older SDK versions will not pick up the new capability automatically. Image and GIF file sizes in tooltip guides should be kept small, as large media assets in an in-app guide can slow load time on the exact screen a team is trying to make clearer.
Document360 Adds Collaborative Editing and Publish Checks
Document360, a knowledge base platform, shipped two releases within the same week. One redesigned the Publish dialog to surface pre-publish checks such as broken links and unresolved comments alongside one click AI suggestions, and the other added collaborative editing so multiple contributors can work on the same article at the same time.
Catching broken links and unresolved comments before an article goes live reduces the number of published articles that need a same day correction, a common source of embarrassment for support teams whose help center is customer facing. Collaborative editing removes a bottleneck where only one writer could work on an article at a time, which matters for teams that review documentation as part of a broader content or support workflow.
Teams should walk through the new Publish dialog once as a group before relying on it for a real release, since the pre-publish checks are only useful if the team trusts and acts on what they surface. Collaborative editing changes how simultaneous edits are resolved, so teams with strict review processes should confirm the new editing behavior does not conflict with existing approval workflows before turning it on broadly.
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.
