Path one: do the work yourself

The first path is to do the work yourself. It comes in two flavours.

The first flavour is to pivot to selling outcomes directly, the way the autopilot-natives do. Crosby is the cleanest example: a $60 million Series B on the back of selling NDAs at around $400 a contract.1 These firms are doing real work and will keep taking share of a class of work that used to sit with law firms and in-house teams. But that route is structurally closed to the platforms in market, for the reason set out last week. To sell outcomes is to compete with the firms and in-house teams the platform sells to. A CLM vendor reselling contract review as a service is selling the thing that replaces its own customers’ contract review function. The channel walks.

The second flavour is to buy your way there. Eudia is the cleanest example: a $105 million Series A, two ALSP acquisitions inside a year, a captive law firm, and a platform partnership, assembling services, software, ALSP, and firm under one roof.2 It is a coherent strategy, well executed. It is also not one most platforms can copy, because most do not have nine figures to deploy, the board appetite to acquire and integrate a services business, or the cultural capacity to operate one once they have it.

Path two: hand it to an integrator

The second path is to hand the work to someone else, which in practice means the systems integrators and the Big Four. The numbers in that category are remarkable; Accenture’s generative AI bookings reached roughly $5.9 billion across its 2025 financial year, nearly double the year before, and the major firms are on similar curves.3 The question is not whether the SI model works. It does, very well, for the work it is built for. The question is whether legal AI platform activation is that kind of work. Usually it is not. The deal sizes are wrong, the unit economics are wrong, the customer relationship ends up owned by the integrator rather than the platform, and the cultural defaults of the SI model do not fit the agentic delivery economics that activation pricing needs. Having spent two years inside the Big Four, I can say this without sounding ungrateful: they will keep doing brilliant work on the engagements they are built for, and platform activation will mostly go elsewhere.

So a platform CEO is left with the discourse on offer: pivot to autopilot and fire your channel, buy a captive ALSP and hope you can integrate it, or hand your customer relationship to an SI. None of those is the move most platforms will make.

The path that works

The third path is the one that does not require the platform to give up its customer relationship, compete with its own channel, or bet the company on M&A integration. Keep selling the platform, and add an outcomes layer beneath it that a partner operates. The vendor stays the customer-facing entity. The platform stays in the stack. The activation work, configuration, playbook build-out, change management, ongoing managed delivery, runs through an operated layer the vendor co-brands or white-labels. The customer buys through a single commercial motion, whether vendor-led, co-branded or white-labelled, rather than negotiating a separate implementation SOW months later. The vendor reaches the work budget alongside the software budget, the services budget for getting the work done, captured by operating the customer’s own instance rather than by competing with the platform’s channel. The activation gap closes, the channel stays intact, and the data flywheel keeps running through the platform.

This is the path the existing platforms are best placed to take, because it protects the asset that is hardest for any new entrant to replicate: the platform itself. What it asks of the partner is specific, and it is the reason a new kind of firm is needed to operate it.

What it asks of the partner

The partner has to deliver agentically from day one, not bolt AI onto a billable base. Outcome pricing only works if delivery is genuinely AI-leveraged, with human supervision built in as an exception layer rather than the default. A firm running humans-bill-hours economics with AI on the side will never reach the price points the market wants.

The outcome is not “perfect legal judgement delivered by machine.” It is a bounded operational output: review to playbook, issue identification, fallback application, escalation pack, cycle-time reduction or throughput increase, with lawyers reserved for judgement and exceptions.

The partner has to field teams that pair deep functional legal expertise with applied AI fluency, in the same teams. Domain credibility is existential; buyers can tell the difference inside a single conversation. The economics also need a delivery footprint the larger incumbents cannot easily match. Our hubs in the UK, South Africa, and Argentina are built for exactly that profile.

The partner has to align commercially with the vendor rather than compete for the same budget. The platform keeps the seat budget; the outcomes layer earns the work budget. Neither cannibalises the other.

And the partner has to give the vendor confidence on IP. Platform-specific implementations belong to the vendor. Generalised know-how and cross-vendor capability belong to the partner. Customer work data belongs to the customer and is used to improve the customer’s outcomes. The vendor’s flywheel stays intact and stays inside the platform.

This is what we have built Telon to be: an AI-native legal services firm built around those four requirements, entering through the activation layer the third path requires. The systems integrators are doing a different job. The autopilot-natives compete with the platforms’ customers. The consolidators are running a play most boards will not approve. Telon is built for the role those categories leave open.

Anthropic’s launch of Claude for Legal has only sharpened the question every platform CEO is now being asked.4 The platform sale used to be the work. Activation is becoming the work above it. The platforms that operate both will compound; the ones that operate neither will be commoditised, acquired, or replaced.

If your board is asking what you are going to do about Anthropic, this is one answer worth a conversation. If you are a GC eighteen months into a deployment that has not moved your backlog, it is worth one too.

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.