Quick note before the notes. My kids started back at school, and whatever spare time I thought I had went with them, so these posts stopped for five weeks. September 7 through today are all published now, and I'm back on the weekly schedule... probably.
On to this week. Fourteen items, no deprecations, and one quiet change to a metric definition that will move numbers in reports nobody is going to re-read.
TL;DR
- The workitem Answered metric is renamed Answer Attempted, and a new Answered metric counts only workitems an agent accepted. Your existing workitem reports change meaning.
- Journey Management adds an effort score dashboard that infers customer effort from behavior rather than survey responses.
- Agent Copilot can summarize long email threads on demand.
- Agentic virtual agents get preset conversation hooks for repetition, tool input preconditions, and knowledge tools gated on prior tool output.
Analytics and Reporting
Start with the workitem metric change, because it's the item most likely to cause a confusing meeting. In the Agent and Queue Workitems Performance views, the metric previously called Answered becomes Answer Attempted, and it keeps counting what it always counted: workitems an agent tried to handle, including ones that were declined or timed out. A new metric named Answered now counts only workitems an agent accepted.
That's the correct definition. It was wrong before, and calling a declined workitem "answered" was always going to mislead someone. But the correction lands on a name people already use. Any dashboard, scheduled report, or quarterly deck built on the old Answered metric now reports a different number, probably a smaller one, with no error and no warning. If your supervisors track workitem answer rates, tell them before they notice it themselves and conclude the team got worse.
Journey Management gets two upgrades. Charts can now compare two metrics at once, with independent aggregations and separate Y-axes so a count and an average can share a chart without one flattening the other. And there's a new Effort Score dashboard that calculates an organization-level effort score from observed interaction behavior, groups customers into effort segments, shows trends, and breaks down which factors contribute most.
The effort score is the more interesting one. Measuring customer effort from behavior instead of asking people to rate it after the fact gets you a number for every interaction rather than for the small, self-selected group who answer surveys. Treat it as directional rather than absolute, and spend your time on the factor breakdown instead of the headline score. The score tells you how you're doing; the factors tell you what to fix.
Agentic Virtual Agents
Three items here and they all address the same complaint, which is that agentic VAs are hard to control.
Preset conversation hooks let authors inject turn-specific instructions automatically when the platform detects the agent or the user repeating themselves, with customizable instruction text per hook. Repetition loops are the most common way an agentic VA visibly fails, and until now the fix was to keep adding defensive language to a prompt and hope. Handling it as a platform-detected event is a better design.
AI Studio also supports preconditions for tool inputs, so a tool only runs when defined conditions are met and invalid inputs don't get executed. And knowledge tools can be linked to virtual input variables even when the underlying API doesn't require them, which means you can gate a knowledge lookup on something like a successful authentication step.
That last one sounds obscure and isn't. If your VA could answer account-specific questions from knowledge before the caller proved who they were, you had a problem you may not have noticed. Now you can enforce the order.
Agent Experience
Agent Copilot can generate on-demand summaries of email interactions, including multi-message and multi-agent threads, and regenerate them as the conversation develops. Email is where context goes to hide, especially on a thread that has passed through three agents, so this is a good fit for summarization. Spot-check the summaries for a week before you let agents treat them as a substitute for reading the thread on anything consequential.
The Associate mobile app now handles SMS ACD interactions with automatic acceptance, threaded history, MMS images, outbound SMS, and wrap-up. Voicemail transcriptions also become visible for personal and group voicemails across web, desktop, and embedded experiences for Associate and Communicate users, controlled by the Voicemail Transcriptions org setting, which now covers ACD, personal, and group voicemail together.
Check what that org setting is currently set to. It governs more than it used to.
Routing and Digital Channels
Customer-first callbacks get configurable agent reservation, so you choose whether an agent is held before the outbound call is placed or only after the customer answers. Reserve early for high-value interactions where you can't risk nobody being available, reserve late when you care more about agent utilization. It's a real trade-off and it's good to be able to pick a side per use case.
Mobile Messenger supports list pickers in Architect digital bot flows through the Ask for Slot action, so mobile users select from predefined options instead of typing. Free-text responses are where bot conversations go wrong, and every place you can replace typing with tapping removes a class of misunderstanding.
Integrations and Platform
The ServiceNow integration can capture and persist metadata from the first inbound email of an interaction, including subject, body, and the From, To, and CC fields, and make it available to Architect flows for priority and skill-based routing. It can also auto-acknowledge inbound emails in CSM Workspace. Subject-keyword routing for email is a long-standing request and this is the plumbing that enables it.
Operating Schedule Group and Site settings now pull from an expanded set of IANA time zone identifiers, with new entries becoming available without waiting for a Genesys product update. Existing configurations are unaffected. Nothing to do here unless you have ever tried to configure a site in a region whose time zone identifier changed, in which case you know why this matters.
Workforce Engagement
Agents can attach a note to a shift trade request explaining the reason or what they'd accept in return, and admins can enable or disable the field. Shift trades fail most often because the person approving them has no idea why the trade was requested, so a free-text field is the cheap fix.
What this means for your contact center
The workitem metric change is the one with a real cost if you ignore it. Find every report and dashboard referencing the old Answered metric and decide whether it should move to Answer Attempted or stay on the new, stricter Answered. Doing that deliberately takes an hour. Discovering it in a quarterly review takes longer and starts from a wrong conclusion.
The agentic VA improvements are worth revisiting even if you evaluated agentic VAs earlier in the year and found them too unpredictable. Repetition hooks and tool preconditions address control, which was the honest objection, and the platform looks different now than it did in the spring.
Everything else is incremental quality-of-life work. Good incremental work, and none of it needs a project plan.
What to check
- Audit workitem reports and dashboards for the old Answered metric and repoint them deliberately.
- Brief supervisors on the Answer Attempted rename before they see the numbers change.
- Review the Voicemail Transcriptions org setting now that it covers personal and group voicemail.
- Enable preset conversation hooks on any agentic VA that has shown repetition loops.
- Gate knowledge tools behind authentication where your VA answers account-specific questions.
- Decide per callback use case whether to reserve the agent before or after the customer answers.
- Replace free-text slots with list pickers in Mobile Messenger bot flows where the options are fixed.
- Spot-check Agent Copilot email summaries against the source threads before agents rely on them.
Full details are in the official Genesys Cloud release notes for October 12, 2026. Questions about how any of this lands in your environment? Email me.
Tracking a deadline? Every announced Genesys Cloud deprecation is on the deprecation tracker.