Your service vendors are being rebuilt around AI

Venture-backed firms are buying up the support, finance-ops and managed-services providers enterprises rely on and re-platforming them around AI agents — and the renewal that follows arrives priced per outcome, sold as your advantage. The acquisition-built structure and unproven stability create governance and continuity risks your vendor process isn’t sized for. Here’s how to keep the leverage on your side of the table.

The first time an AI-native services pitch crossed my desk, I nearly signed it. The savings were real, the agents demoed cleanly and the pricing was the kind procurement loves — per resolved case, not per seat. What I almost missed was who got to define the word “resolved.” On an earlier outsourced-support arrangement, back in my public-sector days, the vendor’s reported resolution rate looked excellent right up until we pulled the reopen numbers ourselves. Auto-closed tickets had been counted as wins. Users had quietly stopped logging issues at all. The dashboard was green; the service was not.

That gap — between the number on the contract and what your users actually live with — is the whole game now, and it is about to scale across your portfolio. Gartner’s 2026 CIO and Technology Executive Survey found only 17% of organizations have deployed AI agents, but more than 60% expect to within two years — the steepest adoption curve of any emerging technology it tracks. The providers running your services are moving first, and the contracts are changing faster than most of us can govern them.

The agents underneath these pitches are also nowhere near as reliable in production as they look in the room. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027 on cost, unclear value and weak risk controls, and reckons only about 130 of the thousands of self-described agentic vendors are the real thing — the rest are “agent washing” old chatbots and RPA. Treat any headline resolution rate the way you treat a vendor’s own uptime stats: Marketing, until you have seen it run on accounts like yours.

What’s behind the pitch

The challenger’s economics aren’t magic. They come from a reshaping of the services market: Venture-backed firms buying up fragmented, labor-heavy providers — support desks, contact centers, AP and finance ops, slices of managed IT — and re-platforming them around agents. Many of the companies pitching you are not one company at all but several acquired shops stitched onto a shared AI layer. That barely comes up in the sales meeting. It matters enormously once you are the customer.

The direction is independently corroborated, not vendor hype. Everest Group reports outcome-based pricing in business-process services moving from pilots to scaled adoption as AI makes outcomes measurable enough to contract on, with the binding constraint now governance and verified baselines. HFS Research tracks the same “services-to-software” shift across consulting, IT, and operations providers. Here is the part worth holding onto: Most enterprises are early, so you have a little time. But the first vendors to reprice you this way will be the small, single-source ones in the long tail of your portfolio — which, if your stack looks anything like mine, is most of it.

Don’t assume the incumbent is the safe choice

And don’t kid yourself that renewing with the familiar name keeps you clear of this. The big integrators are pulling labor out of their own delivery just as fast — Accenture cut tens of thousands of roles and rehired against an AI-skills filter — and rewriting deals around a share of savings instead of time and materials. Outcome pricing is becoming the default everywhere. There is no version of this where you sit it out.

Two risks your vendor process won’t catch

The first is governance, and the roll-up structure makes it worse than the usual AI-vendor worry. The company you are contracting with isn’t one system. It is several acquired firms with different data practices and security postures, with an AI layer dropped on top at speed. Your customer records, invoices and support transcripts flow into agents whose decisions you often can’t trace, across entities that were never built to one standard. The numbers here aren’t comforting: IBM’s 2025 Cost of a Data Breach Report found 63% of breached organizations had no AI governance policy at all, and 97% of those that suffered an AI-related breach lacked basic AI access controls — AI adoption, IBM concluded, is outpacing both security and governance. When an agent botches a dispute or misroutes regulated data, the regulator and the customer come looking for you, not the platform. And here is the organizational trap: The savings line is what your CFO signs; the provenance question is the one your audit committee won’t ask until after the incident. Nobody raises it for you.

The second is whether the provider will still be standing in three years. These platforms are new, built by acquisition, venture-funded and not one has run through a full contract term or a real downturn. With Gartner expecting more than 40% of agentic AI projects to be canceled by 2027, putting core operations on a young, agent-dependent vendor is a single point of failure dressed up as innovation. Write the exit and data-portability terms before you sign — while you still have leverage to.

Six questions for the renewal

When the challenger shows up — or your incumbent reprices to match — these are the questions I would put on the table. They come down to one thing: Making sure you, not the vendor, own the number.

1. Definition, baseline, guardrails. Make “resolved” mean what your users experience, measured against a baseline you have captured yourself, and tie it to metrics you own — first-contact resolution, reopen rate, time-to-resolution. Expect procurement to resist because pinning this down slows the deal. Hold the line; it is the whole ballgame.
2. Real agent, or agent washing. Gartner reckons only a sliver of self-described agentic vendors are the genuine article. Make them prove it: Production resolution on accounts like yours, and the human-escalation rate sitting behind that number. Not a demo.
3. Auditability, as a gate. SOC 2 at minimum, increasingly ISO 42001 or NIST AI RMF alignment, plus model cards, decision logs and an incident-response plan they have actually tested. If they can’t show how data is walled off between their acquired entities, or how an agent’s decision gets traced, they aren’t ready for anything regulated. This belongs in the shortlist criteria, not the post-mortem.
4. Where autonomy stops. Decide which actions an agent can take alone and which need a human, how it hands off with context and who is accountable when it acts on its own. Put names against it before go-live.
5. The exit. An embedded agent platform gets stickier than the staffed incumbent it replaced, faster than you would think. Lock down data portability, knowledge-base ownership and a way out while you are still the one with leverage.
6. Capacity, not just cost. The best outcome here often isn’t a smaller bill. It is the demand that your old service levels were quietly turning away. Nobody answered the tickets. The cases that aged out. Ask what fixing that is worth before you optimize purely for headcount.

The move

Outcome pricing is where this lands, and on balance, that is progress. But in the near term, it hands the advantage to whoever can measure the outcome — and in most shops, that isn’t the buyer. The edge isn’t picking the cleverest challenger or the safest incumbent. It is being able to hold any of them to a result, on your numbers. Look again at why Gartner thinks so many of these projects die: Not the technology — cost, fuzzy value, weak controls. Our side of the table. So, start there. Take one high-volume, measurable workflow, pilot it against a baseline you own, instrument it with your own metrics, and treat the muscle you build doing that as the real deliverable. Get it right and the pricing model stops mattering. Skip it, and you have just agreed to pay for someone else’s definition of done.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?