AI on Contingency, the "Bessemer Model"
Value Billing - One Very Good Future Leading Modality for AI/Law
Something we are accustoming to seeing in other businesses, including commercial leasing, restaurant businesses, and other business tie ups — is customer revenue-based pricing — is entering the space of AI for Law.
And I predict it will not only grow but become a constitutive component of future AI companies financial modeling but LAW FIRM FINANCIAL MODELING as well.
Its not entirely foreign to law.
We have the well known tradition of Contingency fee billing, where attorneys fee is a portion of that recovered by attorney. On the other end of the pecuniary (and sometimes ethical?) scale is client finders and catch-and-kill or percentage of fees referral arrangements for sourcing legal business.
The Rise of Revenue-Share Pricing in AI: Why Companies Are Betting on a Cut of the Value They Create
Law pricing models are undergoing a major revolution (for survival of law firms).
Billable hour major declines.
Flat fee major advances.
Subscriptions (pricing like “Maintenance Contracts” in Software) sure we did since 2007.
On the code side, the traditional SaaS model has always relied on monthly subscriptions or per-user seats. [Harvey and Legora do this. Harvey may do more, read on.]
Yet more and more, the early AI tools for law are leaning in heavily to usage-based pricing—think tokens for large language models or API calls.
[We did this with our model of I n t e r s t i t i a l, which billed by minute.]
Yes, a growing number of AI companies are flipping the script entirely: they’re charging a percentage of the revenue, or savings, or direct value their systems directly generate.
This “revenue-share” or “outcome-based” approach—where the AI vendor only gets paid (or gets paid more) when the customer sees measurable results—aligns incentives in ways that feel like kismet.
No more paying for seats that sit idle or tokens that don’t move the needle.
A vendor who bets on your success, and the lawyer bets on them by signing up and entrusting them with their lawware.
In 2026, this model is exploding in vertical AI applications where ROI is clear and trackable: chargeback recovery, sales automation, customer support resolution, debt collection, and legal outcomes.
Industry analysts at Bessemer Venture Partners and others now explicitly recommend it for AI products that can prove tangible business impact.
Running AI infrastructure is expensive.
Inference costs can eat 20-50% [and in “streaks” where a product “takes off” even much more] of revenue for many AI-first companies, making traditional flat-fee models risky for both sides.
At the same time, customers have grown wary of “AI hype tax”—paying upfront for tools that may or may not deliver.
Outcome-based pricing solves two problems at once:
LOWER ADOPTION COST/RISK - Customers pay only for results. Doesn’t touch present revenues.
VENDOR HUNDER - Vendors are motivated to over-deliver because their revenue scales directly with customer success. AI company obsesses over accuracy, attribution, and continuous improvement.
= Making for near perfect incentive alignment. In essence the two are “investing together.”
The Bessemer’s 2026 AI Pricing Playbook LINK notes that companies using outcome-based models are shifting from “selling access” to “selling outcomes,” with fees tied to resolved tickets, revenue recovered, deals closed, or cost savings delivered.
Real-World Examples
Chargeflow (Chargeback Recovery AI)
The purest example in the market. Chargeflow uses AI to automate and win chargeback disputes for e-commerce merchants. Their pricing is simple and radical: 25% of every successfully recovered chargeback amount. No monthly fees, no setup costs, no contracts. You only pay when money hits your account.
The model has fueled explosive growth—Chargeflow has helped recover over $100 million in disputed revenue for 15,000+ merchants and tripled revenue year-over-year.Debt Collection - AI-Powered Revenue Recovery
Multiple platforms in fraud prevention and collections charge 20-40% of recovered funds. FlyCode, for instance, uses a “Revenue Boost-Based Pricing” model where fees are tied directly to the incremental revenue recovered. Customers pay nothing if the AI doesn’t deliver measurable lifts.Sales - AI Sales Agents & Lead-Generation Platforms
Some AI sales tools and agencies have moved to true commission-style pricing: 10-20% of revenue from closed deals or qualified leads attributed to the AI. Conversica and certain performance-based AI SDR platforms offer options where the vendor takes a cut of incremental sales generated. This mirrors how human salespeople are compensated and is increasingly discussed as the gold standard for agentic AI.[Legal AI (Harvey AI)]
Harvey AI primarily uses high-value enterprise subscriptions (hitting $190M ARR by early 2026), IT IS RUMORED FROM A SOURCE TO AI COUNSEL THAT certain legal AI contracts incorporate success fees tied to case outcomes, settlements, or value recovered—effectively a percentage-based component in high-stakes matters that Harvey earns from law firms.Customer Support - CSR AI Agents
While most (Intercom’s Fin at $0.99 per resolved ticket, Zendesk at $1.50 per successful resolution) use flat per-outcome fees rather than pure percentages, the principle is the same: pay only when the AI delivers a measurable win.
Pros, Cons, and the Road Ahead
Advantages:
Dramatically higher customer willingness to pay (they see direct ROI).
Lower churn—vendors are incentivized to make the product better every day.
Easier sales cycles (proof is baked into the contract).
Challenges:
Attribution can be complex (how do you prove the AI caused the revenue?).
Revenue for the AI company becomes lumpy and harder to forecast.
Requires sophisticated tracking and clean contracts.
FINALLY LAWYERS, I KNOW WHAT YOU ARE THINKING — CAN LAWYER LAW FIRM “SHARE FEES” WITH A NON-LAWYER??? This cat has not yet been skinned but will be.
Bottom Line
Despite the hurdles, the momentum is clear. As agentic AI (autonomous systems that complete entire workflows) matures, more companies are expected to adopt hybrid models: a modest base fee plus a meaningful percentage of value created.
“We’ll take our cut only after we’ve put money in your pocket.”
If you are still pricing your AI for Law on months, look out!
And I do think this aligns much better with the model of Law, and the FUTURE model of law where our customers do not care how many hours or weeks or how many months — a great leap forward toward RESULTS BASED BILLING (FOR AI) AND FOR LAW.
MOVING LAW FROM TIME METRIC TO VALUE METRIC IS A MANIFEST GOOD.