A legal AI tool’s monthly subscription may not tell the whole budget story. In a September 3, 2026 announcement, Gartner warned general counsel about a shift toward hybrid pricing that combines license fees with charges tied to AI consumption. For legal teams, the practical question is becoming: what will this workflow cost when a matter gets busy?
Gartner forecasts that consumption-based pricing will represent more than 35% of net new corporate legal technology spending with major vendors by 2028. That is a forecast about a defined spending category—not a measurement of today’s entire legal technology market, and not a prediction that every firm’s bill will rise by 35%.
What usage-based legal AI pricing changes
A per-user license gives a buyer one familiar budgeting unit: the number of people with access. A metered service adds another variable. The invoice can depend on what those people and their automated workflows actually do, subject to the vendor’s specific rules.
This is not only a theoretical pricing model. Microsoft’s Copilot Credits guidance describes both pay-as-you-go and prepaid purchasing options for eligible services. It explains that unused credits in the annual pre-purchase plan expire at the end of the term, and excess consumption may require additional credits or pay-as-you-go billing. These arrangements are product-specific; they should not be assumed to apply to every Microsoft license or legal AI vendor.
The useful lesson is to read the allowance, meter, overage, and renewal terms together. A generous-sounding credit balance has little planning value until the buyer understands how actual work consumes it.
Measure a completed task, not just a prompt
For litigation teams, an evaluation should follow a task from start to finish. Consider a hypothetical pilot that creates a first-pass chronology from a set of documents. Record the initial processing, follow-up questions, corrections, reruns, and the time a reviewer spends checking the result. This is an evaluation example, not a claim about a particular product’s billing behavior.
The finished, checked chronology is the useful unit of comparison. Counting only the first request can hide the work needed to turn an answer into something the team can rely on. Conversely, a higher software charge may still be worthwhile if the completed process is demonstrably more useful or dependable. Neither conclusion should be assumed before testing.
A practical checklist before expanding a legal AI pilot
The following steps are Webesco’s operational suggestions for evaluating a purchase, rather than reported findings from a vendor benchmark.
1. Ask for an itemized example
Give the vendor a nonconfidential sample workflow and ask it to explain which stages are included and which are metered. Ask whether retries, background activity, document processing, storage, or exports create separate charges. Request the explanation in writing so the evaluation does not depend on a sales demonstration.
2. Test a busy-matter scenario
Compare an ordinary workload with a substantially larger, hypothetical workload. Keep the assumptions visible: document volume, users, processing runs, and review time. The point is not to predict the next case perfectly. It is to discover which assumptions have the greatest effect on the estimate.
3. Separate an alert from a spending stop
Ask what happens when the allowance runs out. Does the service stop, become restricted, or continue charging? Who receives a warning, and who can authorize an exception? Confirm the answer for the actual plan under consideration rather than relying on a dashboard label.
4. Assign someone to review consumption
A small team needs a named owner, not necessarily a new department. During the pilot, schedule a short review of usage and completed work. Microsoft’s Copilot Credits report documentation, for example, describes reporting by user, agent, and billing policy. Reporting capabilities differ by product; ask for the equivalent visibility before committing.
5. Plan for interruption
Document how the team will access source material and finish urgent work if a quota is exhausted or an AI service is unavailable. Keep essential matter files in approved systems. A tool that helps prepare a deliverable should not become the only way to recover the underlying evidence or work product.
What this means for legal teams
The next useful step is a short, documented pilot—not an automatic upgrade or a blanket rejection of metered pricing. Choose one repeatable task, define its quality checks, capture the full cost, and decide in advance what would justify expanding it. Keep confidentiality, access permissions, and attorney review separate from the pricing decision; a cheaper workflow is not permission to weaken those controls.
For a related look at connected litigation workflows, see Webesco’s report on Relativity and Microsoft Copilot. Webesco’s legal technology services support practical preparation across trial presentation, video depositions, and litigation support. Clear scope and reliable human review matter just as much as the tools chosen.