Cognitive Outsourcing™: Why Law Firm Support Costs Keep Rising

At a Glance

• Who this is for: Managing partners, COOs and finance directors at firms where the support function has grown alongside fee earner headcount and nobody owns the job of stopping that.
• The problem: Fee earner cost per head fell slightly in 2025. Support cost per fee earner rose. The cost line that gets the least management attention is the one moving in the wrong direction.
• What is changing: The Law Society's 2026 Financial Benchmarking Survey shows a sector growing at a 15-year high, with IT costs rising on software consolidation and AI spend, and AI now making up 12% of the total risk firms identify.
• The takeaway: Redesigning a support function with AI is capital work with a payback period. Fund it as capital work, or accept that pilots will keep proving the point and not moving the number.

Cognitive Outsourcing™ is a managed-services model in which a law firm transfers operational ownership of a back-office function to a provider, the provider funds the AI redesign of that function over the contract term, savings are shared, and the AI assets transfer back to the firm at the end. It exists because the redesign rarely gets funded any other way.

Why This Matters Now

Firms had a good year. The Law Society's Financial Benchmarking Survey 2026, written by the legal team at Hazlewoods LLP and published on 15 April 2026, draws on 121 firms in England and Wales with combined fee income above £1.2 billion. Median fee income rose 11.2%, close to double the 6.1% recorded in 2024, and 85% of participating firms grew year on year.

Productivity did the work. Median chargeable hours per fee earner rose to 807 from 756. Non-salary overheads fell as a share of fee income, down to 28.4% from 31%. Lock-up came down from 146 days to 134.

So this is not an article about a sector in trouble. It is an article about which cost line a good year hides.

The Line That Moved the Wrong Way

Two numbers in the same survey sit next to each other and point in opposite directions.

The median cost of a fee earner fell 0.4%, to £70,551 from £70,867. Firms held the line on their most expensive people.

The median spend on support staff, covering secretaries, reception, HR, finance and back-office functions, rose to £27,061 per fee earner from £25,655. That is a rise of about 5.5% in one year, on a base that scales with headcount.

The mechanism behind it is structural. Support cost is measured per fee earner, so growth carries it upward on its own. A firm growing fee income 11% and support cost 5.5% is still adding support cost every month it succeeds.

The survey's own account of rising IT cost is worth sitting with. It attributes the rise to consolidation in the software supplier market and to increasing AI spend. So AI is already in the cost base. It has not yet arrived in the support cost line as a reduction.

Why AI Pilots Have Not Moved That Number

Three things get in the way, and none of them is model capability.

The funding shape is wrong. Redesigning how a billing function or a document services team works is capital work with a payback period measured in years. It gets funded from an operating budget that is judged annually, by a partnership whose distributable profit is the same pot. A project that costs money this year and saves money in year three loses that argument most years, and it loses it again the next year.

Nobody owns the outcome. A pilot is owned by innovation. The function is owned by an operations director. The savings, if they ever appear, land in a finance line neither of them is measured on. The Law Society's separate April 2026 research on agentic AI in legal practice found solicitors carrying responsibility for outputs they cannot fully audit, which is the same problem in a different register: accountability sitting somewhere other than control.

The proof standard is asymmetric. A pilot that works still has to survive a partner asking what happens when it is wrong on a client matter. Without someone contractually on the hook for the service level, that question ends the conversation, and it should.

Notice that all three are commercial and organisational. A better model does not fix any of them.

What Cognitive Outsourcing™ Changes

Cognitive Outsourcing™ is a trademarked managed-services model developed by 3Rive, filed in the UK under UK00004278423. Four mechanics define it, and each one answers one of the blockers above.

The handback is the mechanic most people miss on a first read, and it is the one that changes the risk calculation for a managing partner. A firm that outsources a function and gets nothing durable back has bought a lower run rate. A firm that gets the redesigned process and the AI assets back at term end has bought a capability it did not have to fund up front.

The gain-share matters for a different reason. It puts the provider's revenue on the same side of the table as the firm's cost reduction, which is the opposite of a per-seat or per-hour arrangement, where the provider's revenue grows with the volume the firm is trying to shrink.

Where This Fits, and Where It Does Not

This model is not right for every function or every firm. The honest test is four questions.

1. Is the function definable? Billing, document services, conflicts, onboarding and finance operations have measurable throughput. A general "operations improvement" mandate does not.
2. Is there a real baseline? Gain-share is arithmetic on a baseline. A firm that cannot say what the function costs today cannot share a saving from it, and the negotiation will stall there.
3. Is the volume stable enough to redesign around? Redesigning a process that changes shape every quarter wastes the redesign.
4. Is there partner-level sponsorship for a multi-year term? The model only makes sense over a term long enough for the funded redesign to pay back. A two-year deal is a staffing arrangement with extra paperwork.

If the answer to any of those is no, LPO services or a scoped Tech Advisory engagement is the better starting point, and we will say so.

Common Pitfalls

• Treating gain-share as a discount. It is a share of a measured reduction, so it needs a measured baseline first.
• Scoping the function by department rather than by process, which puts work in scope that nobody can hand over cleanly.
• Signing a term shorter than the payback period, which guarantees the redesign is half done at renewal.
• Leaving the handback undefined, so "AI assets" turns out to mean documentation rather than working systems.
• Counting headcount reduction as the whole saving and ignoring what happens to error rates, rework and cycle time.
• Starting with the function that hurts most rather than the one that is most definable. The painful one is usually painful because it is ill-defined.

How 3Rive Approaches This

We run Cognitive Outsourcing™ as one of eight service lines, and most engagements start somewhere else. The pattern we see work is a scoped piece of delivery first, then a function transfer once both sides can price the baseline. Our legal operations work for a global law firm began as profitability planning and matter data work, not as an outsourcing deal. 3Rive is ISO/IEC 27001 certified and delivers from Australia, the UK, the UAE and Sri Lanka.

Key Takeaways

• The Law Society's 2026 Financial Benchmarking Survey reports median support staff spend at £27,061 per fee earner, up from £25,655, while the median cost of a fee earner fell 0.4%.
• Support cost is measured per fee earner, so it grows with the firm unless something changes the process itself.
• AI is already visible in law firm IT cost and not yet visible as a reduction in support cost.
• The obstacles to fixing that are commercial: annual funding against multi-year payback, unclear ownership, and no contractual party carrying the service level.
• Cognitive Outsourcing™ answers those with transferred ownership, provider-funded redesign, gain-share from Year 2, and AI assets returning to the firm at term end.