← Back to Resources

Article · Contract Intelligence

The Model Will Change. Your Playbook Won’t.

Don’t wait for your CLM to figure out AI

November 19, 2026 · George Pratscher

TL;DR

AI models now change faster than any software roadmap can follow. A meaningful jump in capability arrives roughly every quarter, and a typical frontier model is retired within 12 to 18 months. CLM vendors are rebuilding to keep up, and many are struggling.

Whatever tool you end up with will need the same two things from you, a written playbook that says what acceptable looks like for your company and a contract portfolio you understand. Neither depends on which model is current. Build them now with Revver Contract Intelligence and carry them into any CLM or agent you choose later.

It’s a conversation I’m hearing more often. A procurement team has paused its contracting technology plans because its CLM vendor is rebuilding for AI and can’t yet say what the new product will do or when. So the team waits.

I understand the instinct, but I think it’s backwards. The question to ask isn’t which tool will win. It’s which parts of the work will still matter when one does.

The Ground Is Moving Under Every Roadmap

Start with the pace. Somewhere in the industry, a frontier lab ships a new premium AI model every few days. Not every release is a leap, and many repackage an earlier advance at a new price. But a large step in capability lands about once a quarter, and the length of task an AI agent can finish on its own keeps doubling.

Then there’s retirement. Providers now retire models on a rolling schedule, a typical model lasts 12 to 18 months, and notice can be as short as two weeks. Changing the model name in a configuration file takes an afternoon. Revalidating every prompt and every output built on it does not.

Now picture building a contract lifecycle product on that ground. Every design decision is made against a model that will be superseded before the product ships and retired not long after. The figures below are as of November 2026.

The pace of change in AI models as of November 2026, and what each figure means for a CLM roadmap
MeasureFigureWhat It Means for a CLM Roadmap
Industry release paceA premium model from some frontier lab every 4.5 days, and from a typical leading lab every six to 12 weeksThere’s always a newer model than the one your design assumes
One lab, acceleratingAnthropic went from a model every 46 days in the first half of 2026 to every 26 days in the secondEvaluating release by release can’t keep up
Large capability stepsA median gap of about 3.5 months between large stepsA jump worth a redesign arrives roughly every quarter
Agent capabilityTask length doubling about every seven months over six years, and in a little over four months for models since 2023Agent features scoped today look modest within a year
Model lifespanTypically 12 to 18 months before retirementA CLM release can outlive the model it was built on
Retirement noticeAt least 60 days from Anthropic, and about two weeks when GPT-4o left ChatGPTRevalidation is forced on someone else’s schedule
Not always betterAbout one in six measured releases in 2026 scored worse than the model it replacedYou need a fixed standard to notice

The CLM Market Is Feeling It

This isn’t hypothetical. As early as May 2025, an article on Law Insider reported that many CLM vendors were struggling, that some couldn’t raise new capital, and that others were “frantically rebuilding their platforms to survive in an AI-first world.” The models have only moved faster since.

The question is being asked out loud. That article ran under the headline “Is CLM Dead?”, and its author, a chief legal officer, argued the category is being unbundled, with the stages of a contract’s life handled by different tools. PwC’s roadmap has agents coordinating work across CLM, CRM, sourcing and ERP systems. If that’s where this goes, the system that stores your contracts matters less than what you bring to it.

None of that is a criticism of the vendors. It’s what happens when the foundation of a product changes every quarter. It does mean that a buyer waiting for the dust to settle may wait a long time. The model will change before they ship.

What Every Tool Will Need From You

Here’s the part that doesn’t change. Every approach to AI contracting, whether a CLM agent, a copilot or something built in-house, depends on the same two inputs.

  • Your standards. A written statement of what acceptable looks like for your company. It sets the floor you won’t go below, what you ask for first, what you’ll settle for, and the language you’ll always strike. No model upgrade produces your risk appetite. Humans decide these.
  • Your contracts. A portfolio you understand. You know what you’ve already signed, where it falls short of your standards, and what you decided to do about it.

PwC makes the same point in its 2026 work on AI agents in contract lifecycle management. Before any agent is introduced, its recommended path has a company identify the business units and processes affected, clarify ownership across legal, procurement, sales, risk and operations, and prepare its contract data and repositories for intelligent use. Only after that groundwork does it bring agents in. The foundation isn’t a technology choice. It’s preparation, and it’s yours to do.

Build the Durable Parts Now

This is the work Revver Contract Intelligence (RCI) was built for. RCI reviews your vendor agreements against your standards, together with their order forms, policies and addenda, and pins every finding to the exact language it rests on. It drafts the remediation and the negotiating position. Each review records every standard it checked, clean results included, and your decisions stay in the workspace, becoming part of your playbook only when someone deliberately promotes them.

The output is written in plain language, not tuned to one model’s prompt format. That’s the point. When your CLM vendor ships its new product, or you choose a different one, or you build agents of your own, you arrive with a ratified playbook and your contracts in order. The new tool has something real to work from on day one. You may also find you need less of a CLM than you were about to buy.

A Portfolio Audit is the fastest way to start. It’s a fixed-scope, fixed-price review of the contracts you already hold, and you keep the workspace and the decision record afterward. When it’s time to go back to your vendors, our Advisors can negotiate on your behalf or alongside your team.

A Fixed Standard Is How You Notice a Worse Model

One more reason to write your standards down. Newer isn’t always better. The same tracker found that about one in six of the 2026 releases it measured performed worse than the model it replaced.

If your only definition of good contract review lives inside a vendor’s prompts, you won’t know when a model change made it worse. If it lives in a playbook your people ratified, you will.

We run this way ourselves. RCI doesn’t use one model for everything. A full review runs on a larger model set to reason at length. A quick first read, and the jobs that pull dates, duties and dollar amounts out of a contract, run on a faster one. Each pairing of model and settings is chosen by testing it against a fixed set of contracts where we already know the right answers, including clean clauses that should be left alone. The set includes contracts read for a buyer and contracts read for a seller, and each side is scored on its own.

This month we put a newer model through that test for the full review. The longer list it produced wasn’t a better one. It marked clauses that already met the standard as falling short, and it added minor points an experienced negotiator wouldn’t raise. On a two-page staffing agreement it returned 19 items to fix where our answer key has 12. On a seller’s own contract it recommended a change that would have helped the customer. It also cost more to run. We kept the model we had.

Without the fixed standard, we’d have seen a newer model and a longer list and called it an improvement.

Models are becoming a commodity that changes every quarter. Your judgment about what you’ll sign is not. Capture it, and every tool you adopt afterward starts out ahead.

Sources

Revver Contract Intelligence is an AI-native platform that evaluates a company’s contracts from its own position, turning the judgment behind every decision into institutional knowledge that compounds over time.