Going into October 2026, the customer experience news tells one story. Service software is now being priced like labor, and the work left for people is getting harder. Salesforce closed its purchase of the company that made per-resolution pricing popular. Zendesk says the generic chatbot is finished. The most useful number of the month came from a survey, not a keynote: only 9% of customer conversations are handled entirely by AI.
If you have spent years building a service team your customers trust, that’s worth protecting. Below is what happened, why it matters, and what we would do about it if we ran a business rather than a contact center.
Salesforce Buys the Resolution Meter
Salesforce completed its roughly $3.6 billion acquisition of Fin on September 10, 2026. Fin is the company formerly called Intercom, which renamed itself after its AI agent in May. A week later at Dreamforce, Salesforce showed a lineup of job-specific agents, including Casey and Fin for service and customer experience.
The deal size isn’t the interesting part. The pricing is. Fin charges $0.99 per outcome, and a resolution can be counted two ways. A confirmed resolution means the customer said the answer helped. An assumed resolution means the customer left after Fin’s last reply without asking for more help. Some of those customers were satisfied. Some gave up.
So what: when you pay per resolution, the definition of a resolution sets both your cost and your customer experience. Before you sign anything priced this way, pull a sample of assumed resolutions and check how many of those customers came back within a week through another channel. If you can’t run that check, the problem is your measurement, not your vendor. That is exactly the gap our analytics and RevOps work is built to close.
Zendesk Declares the Generalist Bot Dead
On September 14, Zendesk launched Specialized AI Agents, which it says can automate up to 80% of workflows. They come in two types. Industry Agents launch first in commerce, connecting to systems like Shopify, Narvar, Stripe, and Riskified to handle orders, returns, and refunds. Custom Agents are built in Zendesk’s no-code Agent Builder around a company’s own policies. Analysts at Futurum read the launch as Zendesk competing on agent quality rather than platform lock-in, because the agents can also run inside other service environments.
So what: the words “up to” are carrying most of that 80%. A specialist agent can only be as good as the policies it’s given. If your returns policy lives in three people’s heads and a spreadsheet from 2019, no agent will fix that. It will just apply the confusion faster. Write the rules down first, then automate them. That order is how we approach AI integration.
Nine Percent: The Number the Keynotes Skipped
This month’s vendor events leaned hard into autonomy. At Xperience in Las Vegas, Genesys framed its roadmap as a move from an AI-assisted operating model to an agentic one. It also reported Genesys Cloud AI annual recurring revenue above $400 million. Those are Genesys’s own figures.
Then there’s the operator data. Cresta’s 2026 CX Workforce Report, a survey of 300 U.S. CX decision-makers, found that only 9% of conversations are handled entirely by AI. Most are handled by people and AI together. (Cresta’s own pages give that shared figure as 76% in some places and 78% in others, so treat it as roughly three in four.) The same research found that 93% of leaders say calls handled by people are getting more complex. On Cresta’s report page, 81% name integration complexity as the biggest barrier, and only 7% can easily get to their own conversation data.
Even the flagship deployments hedge. Telstra, which is rolling out Agentforce in one of Australia’s largest contact centers, stresses maintaining strong human accountability and governance.
So what: as AI takes the easy tickets, the people left handle your most upset, highest-value customers. Cutting headcount based on deflection numbers while the remaining work gets harder is how service quality slips without anyone noticing until renewals do. Plan staffing, training, and pay for the queue you will actually have. Our AI-enabled marketing operations work starts from that assumption.
The CX Scoreboard Moved. Don’t Frame It Yet.
Forrester’s 2026 CX Index showed the first year-over-year improvement in U.S. CX quality since 2021. Of 225 brands measured in both years, 26% improved significantly and 7% declined. Forrester itself warns that scores are still slightly below 2024 levels.
The more useful finding is inside the company. In Forrester’s new Employee Experience Index, 37% of 150 U.S. brands landed in the negative category and only 25% were positive. Forrester also reports that U.S. retailers with a strong total experience see a 3.8x revenue lift from retention and enrichment.
So what: your employees deliver your customer experience. A tired, under-equipped frontline will show up in your customer scores eventually, whatever the brand campaign says. If you survey customers, survey the people serving them too, and read the two sets of results side by side. We build that into our NPS and survey programs.
Qualtrics Wants to Predict You. We Want to See the Scorecard.
On September 9, Qualtrics previewed its XM Data & AI Platform, which is meant to simulate outcomes, predict what individual customers will do, and act in the moment. It isn’t scheduled to ship until 2027. Meanwhile the company is still digesting its Press Ganey Forsta acquisition. A September WARN filing confirmed 117 Seattle roles cut, which Qualtrics attributed to duplicate functions from the merger.
There’s a practical reason predictive tools appeal to buyers: most customers never tell you anything. Qualtrics’s own consumer research found that only three in ten customers give direct feedback.
So what: prediction is worth having. Calibrated prediction is worth paying for. When any vendor, Qualtrics included, pitches predictive CX, ask one question: where are last year’s predictions next to what actually happened? A forecast nobody scores is just a guess. Talking directly to your customers is still the cheapest way to fill in what the models can’t see. That is the core of our ICP and customer intelligence work.
Europe Makes “Is This a Bot?” a Legal Question
As of August 2, 2026, the EU AI Act’s transparency rules are in force. Users must be clearly informed when they are dealing with an AI system rather than a person, whether that’s a chatbot, an AI agent, or an avatar. According to law firm Travers Smith, a disclosure buried in terms and conditions doesn’t count, and systems already on the market have until December 2, 2026. CMSWire adds that a vendor’s claim of compliance does not prove your specific use case complies.
So what: a U.S. company with European customers is in scope, and the December deadline for existing systems falls in your Q4. Even if you have no EU exposure, clear disclosure is good service. Customers forgive a bot. They don’t forgive feeling tricked by one. Go through your own customer experience as a customer would and check whether the disclosure actually shows up.
Loyalty Programs Are Being Asked for Receipts
Open Loyalty’s 2026 trends report describes tighter budgets and teams now having to prove incrementality, not just engagement, with customer lifetime value and retention ranked above short-term revenue lifts. On the consumer side, Attentive surveyed 600 U.S. shoppers and found that 81% find it motivating to see their progress toward rewards.
So what: a loyalty program that discounts customers who would have bought anyway is leaking margin. The only way to know is a holdout group: members who get the program compared against similar customers who don’t. If the difference doesn’t cover the cost of the rewards, you have a very expensive thank-you note. Visible progress toward a reward is cheap to build and customers clearly want it. Measuring what the program actually adds is where our Revenue Marketing System starts.
Our Take on the October CX News
This month’s news comes down to one idea: automation is now bought by the outcome, so defining the outcome has become the most important job in customer experience.
Salesforce and Zendesk are selling resolutions. Genesys is selling orchestration. Qualtrics is selling predictions it won’t ship until next year. All of it may be worth buying. But each pitch assumes you already know what a resolved customer is worth, which conversations need a person, and whether your frontline is set up to handle what’s left. Cresta’s 9% and Forrester’s employee data point the same way. For now, people still handle nearly all the work that matters, and more of it lands on them each quarter.
The companies that come out ahead won’t be the ones that bought the most agents. They’ll be the ones that measured before they automated, kept a person accountable for the customer, and reported the results in dollars. That is the order we’d recommend: diagnose, then prescribe.
October CX Events
- CCW Europe Summit, October 5-7, 2026, Amsterdam, Netherlands. Customer Contact Week’s European event for customer service and CX leaders, focused on contact center operations and experience management. A good read on how European teams are handling the new AI Act disclosure rules in practice. thecxlead.com
- CCW Nashville, October 7-9, 2026, Omni Nashville, Nashville, TN. The organizer lists 1,000+ attendees and 50+ sessions on AI adoption, workforce strategy, and self-service, with financial services, retail, and healthcare the top industries represented. The one to attend if you want operators rather than vendors. customermanagementpractice.com
- Engage Customer Summit, October 7-8, 2026, London, UK. A practitioner-focused CX summit that is free for practitioners, which makes it an easy yes for UK-based teams. hellocustomer.com
- CX Summit Boston (Digital Transformation & Customer Experience Summit), October 21-22, 2026, DoubleTree Suites by Hilton, Boston, MA. A senior-level summit on customer service modernization, contact center transformation, and AI-powered service operations. conferenziaworld.com
- Support Driven New York Summit, October 27-29, 2026 (dates tentative), New York, NY. A leadership event for support directors, team leads, and ops managers at fast-growing companies. Worth it if your support leader is building the team that handles the harder tickets. gainsight.com