From saved hours to ROI: the question law firms still cannot answer about AI

Two years after the great wave of generative AI adoption in Spanish law firms, the sector has a pending account: almost nobody can say, with numbers on the table, what the return on the investment is. The conversations with managing partners and innovation directors all begin with the same figure - saved hours - and end in the same silence when the question changes to profit.

The metric everyone uses and what it hides

Saved hours are the sector's comfort metric. It is easy to measure: a lawyer estimates how long a task took before and how long it takes now with the assistant, and the difference is multiplied by the volume of tasks. The problem is what happens next with those hours. If the work was billed by the hour, the saved hour is not a gain: it is revenue that disappears. If the work was billed at a fixed price, the saved hour improves the margin, but only if the price held.

The real return depends on a chain of decisions that most firms have not made: repricing the service, redeploying the freed lawyer to billable work, or reducing the cost structure. Without that second step, the saved hour is an accounting illusion.

The invisible costs of the equation

On the other side of the balance, the costs are clearer than ever and larger than the licence invoice. Training is the obvious one: every tool demands hours of learning that nobody bills. Supervision is the hidden one: the output of a generative model must be reviewed by a qualified lawyer, and that review time, in complex matters, can eat a significant part of the saving. And governance - confidentiality protocols, client approvals, audits - adds a layer of fixed cost that did not exist before.

Firms that have done the full exercise privately acknowledge returns that are positive but modest, and very far from the transformation the vendors' narrative promised.

The pricing knot

Everything points to the same knot: the billable hour. As long as the dominant pricing model charges time, efficiency is a business problem, not a solution. The firms that are converting AI into margin are those that have moved the affected services - contract review, due diligence, document production - to fixed prices or subscriptions, where the efficiency gain stays at home.

The paradox is that AI, presented as a technology, is acting in practice as the most powerful argument the sector has ever had for changing its pricing model. Clients, informed of the tools their firms use, increasingly ask why they should pay for hours that a machine performs.

What comes next

The sector consensus is forming around an uncomfortable idea: the ROI of legal AI will not be measured in the profit and loss account of one year, but in the competitive position of the next five. Firms that industrialise AI-assisted services - as ECIJA attempts with The Garage, its new innovation centre in Madrid - are not buying efficiency: they are buying the capacity to offer, in three years, services that their competitors will not be able to match in price or speed.

The firms that answer the ROI question honestly will discover that the technology was never the hard part; the business model was.

More analysis in legal AI and business development.