The pressure from below and above

The incumbents here are the existing legal AI platforms: the CLM vendors, contract intelligence tools, e-billing systems, and document automation built over the last decade around the in-house function and the firms working with it. Most are well funded, well built, and led by capable people. None is in obvious trouble. Several are still growing quickly.

The pressure from below is the autopilot-native services category: firms pricing for the outcome rather than the seat and taking the lawyer out of the loop for a narrow band of work. By the platforms’ standards they look small and their output is occasionally rough. The pressure from above is the foundation model providers. Anthropic expanded Claude for Legal in May 2026 with practice-area plugins and a set of connectors into legal software, and OpenAI and Google are pushing in parallel.2 By the platforms’ standards those legal offerings are early and partial.

The classic incumbent response to both is that they are not really doing what we do. That is, for now, true. It is also more or less what Christensen says incumbents say shortly before the ground moves.

The reason the platforms cannot defend either flank is not that they are slow, or under-resourced, or blind to the threat. It is that the moves which would count as defending themselves are the same moves that would destroy the business paying the bills today.

Defending against the autopilot-natives means selling outcomes, which means selling work the platforms’ own customers currently do. The channel walks. The comp plans, the partner networks, and the success functions are all calibrated to the customer doing the existing work. To go that way, a platform has to be willing to compete with the people it sells to.

Why neither flank can be defended from inside

Defending against the foundation model providers means owning more of the operating layer beneath the model, which is exactly the activation work most platforms have left to others. The product was always a workflow on top of someone else’s intelligence; now the intelligence is starting to do the workflow too. Moving down into the operating layer is the defensive move. It is not where most platforms are built to sit today.

Each move is rational on its own. Together they ask a platform to do the thing almost no incumbent in Christensen’s casebook has done well: cannibalise the high-margin seat business that funds the company now, in order to stand up a lower-margin services business that might fund it in five years. Boards rarely approve that. Sales orgs rarely execute it. The investors who backed the current trajectory rarely reward it.

The honest caveat

I want to be careful not to overstate this. The Christensen framework is descriptive, not deterministic; patterns are patterns, not prophecies. There is a version of the next few years in which the autopilot-natives plateau, the model providers are pulled into larger markets, and the platforms ride the cycle out with their core intact.

But the question every platform CEO is now being asked by their board, in one form or another, is the same. What is our answer on the autopilot-natives? On Claude for Legal? On why NRR is not expanding and renewals are getting harder? The improvements most platforms can reach for, a sharper roadmap, a bigger success team, a tuned sales motion, are real, but they are not answers to a disruption pattern. The pattern asks for a different operating model, not a better version of the current one.

That is the trap. The companies that built the current legal AI market are, by construction, the least free to operate the next one, because operating it would mean becoming a different kind of company, and no company is naturally good at becoming a different kind of company.

The one move that remains

There is one move that does not require that. A platform can bring in a partner to operate the layer it cannot afford to build inside, on a commercial structure that aligns the partner’s economics with the platform’s rather than competing with them. That partner takes the work budget the autopilot-natives are reaching for, and the activation work the model providers will try to reach for, and runs both under the platform’s brand, in a single contract with the customer.

The platform CEOs thinking clearly about this are landing in the same place. The dilemma is not soluble from inside the platform alone. The answer has to come from outside, in a shape no current consultancy, ALSP, or services firm offers by default. Telon is built to be that partner. The next essay is about what that partner has to look like, and why none of the categories already in the market produces it.

For now the observation is simple enough. The platforms are facing the textbook innovator’s dilemma in real time. It is not soluble from inside the platform alone. The ones that find a way to bring the answer in from outside, before the pressure from below and above closes in, will be the ones that compound. The ones that do not risk becoming the case studies everyone later claims were obvious.

Lewis Bretts is the CEO and co-founder of Telon. Previously US Managing Director at LOD, Partner at PwC, and COO / Chief Legal Engineer at SYKE.