Tessera

Your firm is already using AI. Tessera tells you what it cost, and which client it was for.

Every request tagged to a matter, priced to the penny, and recorded in an audit trail your compliance officer can export. Installed in your own environment, using your own provider accounts.

View the demo

A worked demonstration using fictional data for a 46-person firm.

The problem

Fee-earners are already using it

On personal accounts, outside the firm's control, with client documents. A firm-wide policy prohibiting AI has not stopped this anywhere.

Nobody can attribute the cost

Subscriptions sit on partners' expenses. No one can say which client's work consumed what.

There is nothing to show an insurer

When your PII renewal asks how AI is governed, a policy document is not evidence of practice.

AI coverage
AI coverage for the demo firm, read from a month of firewall logs the firm already held.
Demo figure

38 of 51

Identities reaching an AI service in the demo firm's own firewall logs, and how many of them are people working inside Tessera. The rest are the onboarding list.

What it does

Every request tagged to a matter

A matter code is required before anything is sent. Untagged usage is visible, by fee-earner, as a percentage.

Cost recorded to the penny

Priced at the moment of the request against the rate in force, so historic figures never shift. Export by matter for your billing period.

Matters / COMM/2026/0085 — Ravenscraig Logistics Ltd
One commercial matter, with every request charged to it and priced at the rate in force that day.
Demo figure

£0.004821

What one request cost: 1,704 tokens of prompt and 350 in reply, at the rate in force that day. The ledger holds six decimal places; the screens round to four.

Compliance
The compliance export, and the requests served above or below the tier the rules suggested.

Routed to an appropriate model

Routine extraction does not need your most expensive model. Tessera suggests a tier, shows the fee-earner the cost before they run it, and records every override.

A complete audit trail

Append-only. Who ran what, against which matter, on which model, in which region, and what was blocked. Exportable by date range.

How it runs

Where Tessera sits in relation to your firmA boundary is drawn around the firm’s own Microsoft tenant. Inside it sit three things: the fee-earner, Tessera, and the firm’s documents together with the record of every request. Outside the boundary sit the AI providers, reached on the firm’s own accounts. A single line crosses the boundary, from Tessera out to the providers and back, carrying the request itself. Nothing else crosses: the record of what was asked, what it cost and which matter it belonged to stays inside the tenant.YOUR MICROSOFT TENANTFee-earnerTesseraYour documents, and the record of every requestAI providersyour own accountsthe request
Tessera and the record both sit inside your tenant. Only the request crosses to the provider, on your own account with them.

Installed in your firm's own Microsoft environment. Your data does not pass through us and we hold no copy of it.

It uses your firm's own accounts with the AI providers. We do not resell model access and your inference spend never appears on our invoice.

Your provider contracts already commit to not training on your data. Tessera's job is to enforce which providers are permitted and to prove what actually happened.

Demo figure

7 years

How long the demo firm keeps the prompt and output behind each request. The period is the firm's own setting, and the ledger entry outlives the text either way.

Who it is for

UK law firms of roughly 20 to 100 fee-earners, with a practice manager or COLP who needs to answer questions about AI use and currently cannot.

What it is not

Not a legal research tool. Not a drafting assistant. Not a replacement for Copilot or the tools your fee-earners already use. Tessera sits underneath them and records what they cost.

View the demo

Built in Scotland.