What Zendesk's big bet on autonomous AI signals to midmarket teams

Between the Forethought acquisition and the Relate 2026 roadmap, Zendesk has made its direction clear. For midmarket support leaders, that direction is worth reading closely.

Written by

Chris Vavra

Last month I wrote about what Zendesk's AI actually costs — where the charges land, how they stack, how they scale. That's the tactical question, and it matters when you're budgeting. But it isn't the most important thing that happened to Zendesk this year. The most important thing is what the company decided to become. If you run support for a midmarket team, that decision is worth understanding, because it tells you where your platform's attention is going — and whether it's going toward you.

What Zendesk actually announced

In March, Zendesk acquired Forethought, an agentic AI company founded in 2018. It was Zendesk's largest acquisition in nearly two decades, and the company said it would accelerate its own AI roadmap by more than a year. Two months later at Relate 2026, its annual conference, Zendesk showed what that acquisition was for.

The headline was a vision Zendesk calls the "Autonomous Service Workforce" — a single platform meant to run service operations with, in the company's framing, as much autonomy as you choose to grant it. Forethought's technology now underpins the platform's AI Agents layer and its self-improving "Resolution Learning Loop," in which agents detect gaps in a workflow, generate new procedures to close them, and test those changes before deploying. Around that, Zendesk announced a stack of capabilities aimed squarely at complexity: Agent Builder for constructing custom agents in natural language, Action Flows for running workflows across external systems, a new Model Context Protocol layer for connecting agents to outside tools and models, and specialized agents reaching into middle- and back-office work. The stated ambition, repeated across the event, is that 2026 will be the year autonomous AI resolves more service interactions than humans do.

This is a repositioning, not an update

Forethought is the capstone of a buying spree — by one count, eight acquisitions in three years — and the pattern points in one direction. Zendesk CEO Tom Eggemeier put it plainly at the event: "Resolution is our identity." A company that built its name on ticketing is now defining itself by autonomous outcomes, and the numbers show it's serious. Zendesk has said its AI-specific annual recurring revenue exited its first fiscal quarter around $250 million and is targeted to reach roughly $400 million by the start of 2027. Industry analysts covering the Forethought deal read it the same way — as Zendesk's clearest move yet away from its ticketing roots and toward enterprise-grade autonomous resolution.

For large, complex operations, this is a credible and even impressive direction. Self-improving agents that reason across systems and handle multi-step workflows are genuinely hard to build, and Zendesk is assembling the pieces at scale. But it raises an important question: what does that mean for a midmarket support team?

What the roadmap signals for midmarket

Read the Relate announcements as a midmarket support leader and a pattern emerges. The center of gravity is autonomy, self-improvement, cross-system orchestration, and middle-office complexity — the concerns of large enterprises with sprawling systems and the staff to govern them. Take Forethought itself: Zendesk is running it as a cross-platform, enterprise-grade product — sold through a sales conversation with no public pricing, and offered to Zendesk and non-Zendesk customers alike, so it can win large, complex operations wherever they already run their tickets, no migration required. That's an upmarket play, not a midmarket one.

The AI you actually meet inside Zendesk Support is packaged to match. The autonomous AI agents now bundled into every plan are metered per resolution, on an allowance most teams exhaust quickly. The agent-assist layer, Copilot, sits behind a plan-tier upgrade and a per-agent add-on, with its deepest features reserved for the top tier. I walked through those economics in detail separately; the short version is that reaching the parts that matter means some combination of tier upgrades, per-agent add-ons, and per-resolution charges that climb as you grow.

None of this makes Zendesk a poor platform. It makes Zendesk a platform optimizing for someone specific — and for a midmarket team, that's useful clarity rather than bad news. It tells you not to expect the roadmap to bend toward your starting point. The capabilities you most want to begin with — reliable help on everyday tickets, a knowledge base that stays current — are available, but they sit behind the packaging built for a larger, more complex buyer.

The choice for midmarket teams

So what about midmarket teams? None of Zendesk's three AI paths fits them cleanly. Forethought targets complex, autonomous use cases, and it's shaped for enterprise in both implementation and cost. The autonomous agents bundled into Zendesk's plans carry per-resolution fees that climb with volume. And even the agent-assist use cases — often the most attractive entry point for AI in support, because they're lower-risk and quicker to show value — sit behind higher plan tiers and per-agent fees. For a team that values affordable, predictable pricing and simple deployment, none of these is a comfortable starting point.

The good news is that the choice isn't between overpaying and going without. You don't have to migrate to a new ticketing system to get AI, and you don't have to buy the enterprise-shaped version to get value. The most sensible path may well be to keep the Zendesk that's served you well and add purpose-built, third-party AI from the Zendesk ecosystem — agent-assist on your everyday tickets, on pricing built for your size.

The takeaway

The Forethought era is real, and for enterprises building toward autonomous service at scale, Zendesk has made a serious, well-funded bet. The smart move for a midmarket team isn't to dismiss it — the benefits of AI in customer support are real. It's to read the signal accurately: know where your platform's attention is heading, and choose your own entry point on purpose. Start where the economics are predictable and the value shows up fastest — usually building a more effective self-service knowledge base and adding agent-assist on the tickets you handle every day.

That's the difference between adopting a vendor's thesis and making your own decision. For a midmarket support team, the second one tends to pay off sooner.

Chris Vavra is the founder of Flexivity AI, which brings affordable agent-assist and continuous knowledge-base gap-closing to the Zendesk you already run. The views here are his own read of the research.

Figures as of July 2026, drawn from Zendesk's published announcements, newsroom, and pricing page; some capabilities are in early access and subject to change. Zendesk and Forethought are trademarks of Zendesk, Inc.