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 the speed of engineering and what it means for how Customer Experience teams operate.
Today’s headlines are from:
The Speed of Engineering Has Changed. Customer Experience Has Not Caught Up.
For most of the last decade, the relationship between engineering and customer experience operated on a predictable rhythm. Engineers built in sprints. Product managers planned quarters. At the end of every cycle, CX received advance notice, updated the knowledge base, revised macros, and prepared Support Specialists for what was coming.
That rhythm is gone.
AI-assisted development has broken the cadence that CX teams depended on. Engineers who previously shipped features on a two-week sprint cycle are now shipping in days or hours. Deployment frequency has climbed across the industry. The code is moving faster. The rest of the organization is not.
Customer Experience sits at the end of that pipeline. When engineering accelerates without a corresponding change in how CX operates, Support teams absorb the difference in the form of tickets, confusion, outdated documentation, and customers who encounter features the team was never trained to support. This is not a future risk. It is the current operating condition for most Support organizations.
The Old Model Assumed Predictable Handoffs
The prior model worked because it was built on shared assumptions about time. Engineering showcased completed work on the last Friday of a sprint. Product signed off the following week. CX had the early part of the next week to update help articles, revise internal wikis, build or edit macros, and schedule team training before the feature hit production.
That window gave CX leaders a reasonable chance to keep the team current. Articles reflected the product. Support Specialists could answer questions about new features because they had been trained on them. When a company ships multiple times per week or per day, that preparation window collapses. CX does not fail to prepare because the team is not trying. CX fails to prepare because there is no longer a defined preparation window to work within.
AI Has Not Made Products More Stable
AI tools amplify the output of existing engineering practices. Strong engineering cultures ship more reliably with AI assistance. Organizations with weak testing, unclear ownership, or thin documentation now ship those problems faster and at higher volume. The assumption that faster engineering produces more reliable software does not account for the quality of the practices accelerating it.
Bug prioritization compounds the problem. New features attract new customers and new revenue. Bug fixes do not. That incentive structure, which existed well before AI entered the development cycle, now operates at a higher tempo. Known issues persist in production while new features continue to ship, and Support Specialists are left explaining both.
The Knowledge Base Is Always Behind
The single most visible symptom of this gap is documentation lag. When product changes land faster than CX can document them, the knowledge base stops being a reliable source of truth. Support Specialists begin working around it, relying on Slack messages, institutional memory, or direct escalations to product contacts to answer questions the help center should already cover. This is a structural problem, not a documentation team failure.
AI-powered knowledge tools can help close this gap. They can draft article updates, flag outdated content, and generate first-pass documentation from release notes. But those tools require clean inputs. If CX does not have visibility into what engineering is building before it ships, no documentation tool can compensate. The documentation problem is upstream of technology. It is a visibility problem.
What CX Needs to Rebuild
There is no version of this problem that resolves without deliberate structural change.
Visibility into the development pipeline. CX leaders need direct access to the system engineering uses to manage work, whether that is a Jira board, a Linear workspace, or an equivalent tool. Waiting for announcement emails or release notes is not a workable operating model when engineering ships continuously. CX needs to see what is in development, what is moving toward QA, and what is scheduled for production, early enough to act on that information.
A recurring working session with Product. Not a quarterly review. A structured, weekly session where CX brings Voice of the Customer data and Product brings what is in the upcoming queue. This session gives CX advance notice of what is coming. It gives Product a structured mechanism to hear what is actually happening in production from the customer perspective, which is the signal that most AI-accelerated development processes are losing access to.
A categorized, maintained data structure for customer signals. Bugs, feature requests, and customer friction points need to be captured in separate, clearly structured queues, not a single backlog of things customers mentioned. This is the data CX brings to the weekly Product session. When that data is organized, it is usable. When it is not, it stays inside tickets and goes nowhere. AI tooling can help surface patterns, but only if the underlying data is captured with enough structure to work from.
Defined ownership of the cross-functional operating model. The most overlooked element is accountability. Someone in CX needs to own the relationship with Product and Engineering as an operational responsibility, not an informal one. That means maintaining the feedback data structure, attending the weekly Product session, and monitoring what is moving through the development pipeline. In smaller organizations, this may be the Support Leader directly. In larger ones, it likely requires a dedicated role.
The Real Cost of the Gap
Support Specialists who are not trained on new features give inconsistent answers. Some escalate. Some improvise. The customer experience becomes a function of which Specialist happened to pick up the ticket, rather than what the organization actually knows. AI-handled conversations require an accurate, current knowledge base to resolve queries correctly. When the knowledge base is outdated, deflection rates and resolution quality fall together.
The merging of engineering and product roles that AI is accelerating also carries a risk for customer relationships. As companies reduce headcount and consolidate functions, the number of people with direct customer contact decreases. CX has historically served as the primary translator of customer experience into product language. As organizations move faster and rely more heavily on usage data and AI-surfaced signals, that translation function becomes less visible. Usage data tells you what customers do. CX tells you what they mean by it. Both are necessary. Right now, most organizations are investing heavily in one and reducing investment in the other.
Starting Points
The full redesign of how CX integrates with engineering and product is a sustained effort. But several things can begin this week without a budget request or an organizational restructuring.
Request read access to the tool engineering uses to manage their work. Frame it as an operational need: you need to see what is coming before it arrives.
Audit the current knowledge base for accuracy. Identify which articles cover features released in the last 90 days and measure how many of them reflect current product behavior. That number is your documentation debt.
Pick one customer signal category, bugs, feature requests, or friction points, and build a consistent intake process for it. The goal is to move from anecdotal signal to structured data before the next Product conversation.
Request a recurring one-hour weekly block with your Product counterpart. Bring data. Ask about the queue.
If you are working through how to restructure your CX operation to keep pace with your engineering organization, schedule a discovery call at rushtoresolution.com/contact.
Kustomer Launches Architect for Goals-Driven AI Workflows
Kustomer, a customer experience platform, announced Kustomer Architect, a workflow design environment that lets CX operators build and deploy AI-powered support experiences without engineering involvement. The platform uses a goals-driven model: operators define the outcome they want, such as retaining a customer, resolving an issue, or protecting a high-value relationship, and Architect orchestrates the combination of AI, workflows, and human handoffs required to reach it. The tool is positioned as a direct shift away from the deflection-rate era of CX measurement, toward outcomes that appear in a P&L rather than a support dashboard.
Architect gives CX operators direct control over how AI behaves inside the platform: what it handles autonomously, where it escalates to a Support Specialist, and how it responds to edge cases, without writing code or filing an engineering ticket. Every AI decision is grounded in Kustomer’s full data layer, including customer history, conversation context, existing workflows, and knowledge content. The platform also supports MCP connections, so teams can extend AI into external systems such as order management or returns platforms without being restricted to Kustomer’s native data model. Kustomer cites HexClad as an early customer that achieved reduced cost-to-serve alongside maintained CSAT, though specific figures were not released.
Architect is a Kustomer platform feature. Teams not already using Kustomer do not have access to it. The goals-driven configuration model is more complex than rule-based automation and requires teams to have already defined what successful customer outcomes look like in concrete, measurable terms. Many support organizations have not done that definitional work, and the platform cannot supply it. Human-in-the-loop controls, including confidence thresholds and escalation rules, require calibration before they reflect accurate operational judgment. MCP connectivity is a genuine capability, but it depends on whether the external systems a team wants to connect actually expose MCP servers. Many enterprise systems do not yet.
Assembled Brings MCP to Contact Center Operations
Assembled, a workforce management and AI agent platform, announced Assembled MCP, the first MCP server built for contact center workforce management. The server connects Assembled’s live and historical contact center data to the AI tool of the operator’s choice, including Claude, ChatGPT, and Gemini, allowing support leaders to query forecasts, adjust scheduling, monitor SLA performance, and manage intraday operations through natural language rather than manual data exports. Assembled MCP is now available to all current Assembled customers.
The practical change for WFM teams is access speed. Reports that previously required analysts to pull data from multiple systems, format it, and return results have become conversational queries. A leader managing a volume surge can ask whether current staffing is sufficient to hold a response time SLA and receive an answer grounded in live Assembled data, with time to act on it. Cross-system synthesis is also available: leaders can pull Salesforce revenue data alongside Assembled contact volume in a single conversation rather than across separate exports. Checkr’s Senior Manager of Operations Tools and Systems described it as a tool that democratizes WFM data across the business, giving leaders access to operational answers where they already work.
Assembled MCP is not a standalone tool. It requires an active Assembled subscription and assumes Assembled is already the system of record for workforce data. The AI-agnostic design is the right architectural call for enterprise environments where teams already use multiple AI assistants for different purposes, but output quality depends on how precisely leaders structure their queries. Teams without prior experience using conversational AI for operational analysis will need time to develop that skill. OAuth authentication scopes access to individual accounts, which addresses the core data privacy concern, but teams in regulated industries should verify which data categories flow through the MCP connection before enabling it broadly. The forecasting and scheduling examples Assembled demonstrates also assume accurate underlying data. WFM outputs are only as reliable as the contact volume history and scheduling records feeding into them.
Decagon Adds Guided Discovery to AI Agent Capabilities
Decagon, an AI customer service platform, announced Guided Discovery, a capability that lets AI agents handle open-ended, exploratory conversations without relying on predefined decision-tree workflows. Where standard agentic workflows require a clear problem to follow a mapped path, Guided Discovery allows agents to ask follow-up questions, gather context as the conversation unfolds, and reason toward the right outcome for the customer. The feature targets use cases such as product discovery, retention, and expansion, where the customer’s need emerges through dialogue rather than arriving fully formed.
Guided Discovery is built on Decagon’s next-generation Agent Operating Procedures engine, which lets teams define what an agent should understand and reason toward using natural language instructions rather than flowchart logic. Agents maintain context across both the current session and prior customer history via Decagon’s user memory feature and can call external tools mid-conversation. Decagon’s Duet interface allows teams to write these goal-oriented instructions, generate test scenarios, and surface improvement trends from live conversations without building through code.
Goal-oriented configuration is a different skill set than building rule-based workflows. Teams transitioning from step-by-step automation will need time to write effective natural language instructions and test how agents reason under varying inputs. Guided Discovery also depends on accurate underlying data: for product discovery use cases, the quality of recommendations is a direct function of how well the product catalog and customer history are structured and accessible to the agent. Teams without clean, organized inputs should expect a gap between what the feature can do and what it does in production.
Fin / Intercom Lets Admins Hide CSAT Scores from Agents
Fin, a customer communications platform, added a workspace-level control that allows Admins to hide CSAT ratings from the conversation stream visible to Support Specialists. All CSAT data continues flowing to reporting as normal. The setting is available in Settings → General.
The update gives teams a direct way to separate the measurement of satisfaction from the handling of individual conversations. When agents can see a prior CSAT score while handling a new conversation, it can shape how they respond, whether toward the customer’s actual need or toward protecting their own metric. Removing scores from the conversation view while keeping them in reporting gives team leads data for coaching and QA without that data affecting in-conversation behavior.
The setting is workspace-wide and not scoped to specific teams or queues, so enabling it removes CSAT visibility for all agents across the workspace. Organizations that use visible CSAT scores as a signal for prioritizing high-risk conversations should evaluate that workflow before switching it on. The feature does not change how CSAT surveys are sent, how scores are captured, or how they surface in reporting.
Gorgias Adds Self-Serve Message Content Export
Gorgias, a customer support platform for e-commerce, added a self-serve export option that lets teams pull raw ticket message content directly from reporting drill-downs. Previously, accessing message bodies required either building an API integration or submitting a support request for a manual CSV export. The new “Export with message content” option appears alongside the existing metadata-only export in the standard ticket export flow and covers up to one month of data.
Teams running quality reviews, auditing agent responses, or building internal conversation datasets no longer need to route requests through Gorgias support or maintain API connections to access message content. The export is available within the reporting workflow directly, which reduces the time between identifying a conversation set worth reviewing and having the data to work with.
The message content export is capped at one month. Teams that need more than 30 days of conversation bodies for a single export will still require the API. The metadata-only export retains its 2.5-year window for historical volume and resolution analysis. For organizations that handle sensitive customer data, exporting raw message bodies introduces a data handling consideration the metadata export does not. Teams should confirm that the environments where these exports are processed and stored align with their data retention and privacy policies before making this a routine practice.
Linear Auto-Creates Slack Channels for New Projects
Linear, a software project management platform, added automatic Slack channel creation for new projects. When a project is created in Linear, a dedicated Slack channel is spun up, all project members are added, and project updates post to that channel by default. The release also includes Linear Asks, which lets anyone in a Slack workspace file a request in Linear by mentioning @Linear Asks, with the agent selecting the appropriate template and collecting any missing fields before creating the issue.
For CX teams working to build the pipeline visibility described in the main story, this update creates a ready mechanism. If Engineering creates projects in Linear, those projects now automatically generate Slack channels that CX leaders can request access to. Linear Asks offers a complementary path: Support Specialists and leads can log bugs and feature requests from Slack directly into the engineering pipeline without navigating Linear or knowing which team handles a given issue type. Both lower the coordination overhead that typically slows cross-functional feedback loops.
Automatic channel creation requires admin setup in Linear settings and an active Slack integration. Linear Asks is available on Business and Enterprise plans only. For CX teams, the practical value depends entirely on whether Engineering already uses Linear and whether the workspace is configured to grant CX meaningful access. The feature removes a logistical barrier to pipeline visibility, but it does not resolve the organizational one. Requesting access to project channels still requires buy-in from whoever manages Linear workspace membership on the Engineering side.
Guru and Notion Ship Table Cell Merging
Guru and Notion, both widely used for team knowledge management, each shipped table cell merging within the same week. Guru’s update adds Merge Right, Merge Down, and Split controls to its table editor inside Cards and Pages, with merged cell formatting preserved when pasting from Google Docs. Notion’s release enables cell merging in its simple table block, allowing users to create headers that span multiple columns or group related rows.
For CX and support teams that maintain structured knowledge bases in either platform, cell merging enables table layouts that were previously only achievable through workarounds. Escalation routing matrices, product comparison tables, tiered response guides, and troubleshooting decision layouts can now carry spanning headers and grouped sections that match how the content is actually organized, rather than being flattened into rows of equal-width cells. Teams that import documentation from Google Docs into Guru will also benefit from merged cell formatting surviving the paste.
Cell merging is a formatting capability and does not change how either platform indexes, searches, or surfaces content through AI features. Teams should use merged cells for readability and structure, not to signal hierarchy to search or AI tools that parse article text. In Notion, the feature applies to simple tables and not to database views. Teams using Notion databases for structured content should confirm which table type they are working in before expecting merging to be available. In Guru, it applies to Cards and Pages.
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.
