Should Firms Bill for AI Agent Supervised Work

Anthropic's Project Deal pilot completed 186 transactions across 69 employees on April 24, 2026. The agents did the per-transaction work. Lawyers weren't in the room. So when corporate clients deploy this architecture at scale, what should firms bill for? The naive answer is nothing. The AI handled it. The sophisticated answer is firms should bill for architecture, not labor. Engagement letters, authority envelopes, audit-log specifications, dispute-resolution clauses: those are productizable deliverables. The fee structure that survives 2027 looks different from the billable-hour grind that defined 2024. Here's the math, the practice-shape, and the procurement-side incentives that will reshape mid-market and BigLaw economics for transactional work.

The unbundling: what the agent does vs what the lawyer does

In a Project Deal-style flow, the agent handles listing creation, discovery, negotiation, transaction execution, and routine dispute resolution. The lawyer handles framework architecture: which transactions trigger review, what authority envelope applies, how the audit log is configured, what the dispute-resolution clause stack says, how cross-border data handling works.

That's an unbundling. The traditional commercial-transactional practice billed for per-deal review at the hourly rate, plus a partner premium for novel structures. Agent flows compress the per-deal review into the agent layer. The lawyer's value moves up the stack to the architecture.

The second-order effect: the architecture is reusable across deployments. Write the engagement letter template once, customize per client. Each marginal client deployment is high-margin because the template carries most of the work. That's a productized practice shape, not a labor practice shape.

The third-order effect: firms that resist the unbundling and try to bill hourly for transactions the agent already did will lose to firms that bill productized fees on the architecture. Procurement teams will notice. They already do for SaaS contract reviews and vendor procurement playbooks.

Fixed-fee architecture: the deliverable is the document stack

A defensible fixed-fee structure for agent-supervision practice has three tiers:

Tier 1, Initial deployment framework. One-time fixed fee per principal client. Deliverables: master engagement letter, deployment authority envelope template, audit log architecture specification, disclosure language for transaction documents, dispute-resolution clause stack, antitrust compliance protocol, privilege-and-confidentiality posture. Estimated price band: $5,000-$25,000 depending on client complexity and regulated-industry overlay.

Tier 2, Quarterly audit retainer. Recurring fixed fee per quarter. Deliverables: review of deployment audit logs against authority envelope, compliance-protocol effectiveness check, regulatory-update overlay (state bar opinions, FTC guidance, court rulings on agent flows), engagement letter updates as needed. Estimated price band: $2,500-$15,000 per quarter.

Tier 3, Per-incident escalation. Fixed fee or hourly for specific dispute escalations exceeding the authority envelope. Deliverables: review of agent transcript, application of legal judgment to disputed transaction, drafting of resolution position, communication with opposing counsel or arbitral forum. Estimated price band: $1,500-$10,000 per escalation depending on complexity.

This structure aligns with how procurement teams buy professional services in 2026: fixed fees on deliverables, predictable margins on quarterly retainers, transparent escalation pricing. It also aligns with the agent supervision rules deep-dive which establishes engagement letter as the work product.

The mid-market opportunity: predictable margin, scalable practice shape

For a 25-person mid-market firm, agent-supervision practice has a different economic shape than commodity transactional work. Three deployments per quarter at $15,000 fixed fee plus $7,500 quarterly retainer per client equals $90,000 quarterly run rate from agent-supervision retainer alone, growing as the client base compounds.

That's $360,000 annual run rate from a practice area that didn't exist 18 months ago, with template-based deliverables that become more efficient over time. By year three, the same firm with 30 retained clients across 12 quarterly retainers each could run a $1M-$2M agent-supervision book with two-to-three lawyers dedicated to the practice.

The second-order economics: the practice doesn't compete on price with commodity transactional work because procurement teams don't perceive it as a substitute. It's a category that emerges because the underlying technology shifts. Early movers own pricing power.

The third-order point: the firms that win are the ones with a written template stack and documented competence. Firms that hand-wave the practice ("we'll figure it out per matter") lose the procurement RFP because corporate counsel can't defend an unstructured purchase to their board.

BigLaw economics: the playbook is the asset

For BigLaw firms, the opportunity isn't the per-matter retainer fee. It's the playbook. Anthropic, OpenAI, Google, and the agent-deployment vendors will need a small number of firms to build the canonical agent-transaction-governance framework. The first firm to publish a defensible whitepaper plus engagement-letter template plus state-by-state supervision analysis owns the next decade of inbound from corporates running these pilots.

Freshfields' Anthropic co-build is the precedent: co-build with a foundation model rather than wait for the spec. The firms that will be quoted in *Above the Law* and *American Lawyer* on agent supervision in 2027 are the ones publishing whitepapers in 2026.

The second-order BigLaw effect: associates who develop deep agent-supervision competence become high-leverage practitioners. The work scales because templates scale. Associate hours per matter are lower than traditional transactional work, but matter count per partner is higher. The economics support a different staffing pyramid: fewer associates per partner, but each associate handles more matters.

The third-order effect: in-house legal teams at corporates running 50+ agent deployments will start hiring senior agent-supervision specialists. The talent pipeline runs through the firms that build the practice early.

What corporates will pay for vs what they won't

Procurement teams will pay for productized agent-supervision deliverables when the deliverable is documented, the scope is clear, and the price is fixed. They won't pay for hand-wave "we'll review it" engagements at hourly rates.

The productized deliverables that procurement will buy:

- Documented engagement letter and authority envelope (one-time fixed fee). - Quarterly audit retainer with documented scope (recurring fixed fee). - Specific deliverables tied to deployment milestones, pre-deployment review, post-deployment audit, and regulatory-update overlay (fixed fee per deliverable).

What procurement won't pay for:

- Open-ended "we'll be available" retainers without documented scope. - Hourly billing for transactions the agent handled, without architecture-level value-add. - Legal opinions that don't address the specific deployment configuration.

The procurement signal: corporates running agent pilots are sophisticated buyers. They're already buying SaaS, cloud infrastructure, and vendor management products with productized scopes. They expect legal services on agent supervision to follow the same pattern. Firms that meet that expectation win the engagement.

The escrow and arbitration framework covers the dispute-resolution architecture that supports the per-incident escalation tier of the fee structure.

The Bottom Line: My take: Don't bill hourly for transactions the agent handled. Bill fixed fees on the architecture: engagement letters, authority envelopes, audit-log specifications, dispute-resolution clauses. Mid-market firms can build a $1-2M agent-supervision book within three years on template-based deliverables. BigLaw's bigger play is the playbook itself: publish the whitepaper, ship the template, own the next decade of inbound. Procurement teams will buy productized deliverables. They won't buy hand-wave hourly retainers.