EngagementAI for legal

AI operating model for the legal department

A fixed-scope engagement in which HOVR Legal draws up an operating model for how the department uses AI, whatever the tool. When the engagement ends, you have the operating model in a document the department owns and can run itself.

Who it is for

Legal departments that already have one or more AI tools in use and notice that use is uneven: some use it for everything, others not at all, and nobody knows what is good enough to rely on. Also departments that are about to choose a tool and want to know what to require of it.

What you contribute

  • A contact person with a mandate.
  • Time for interviews: one hour per lawyer and half an hour per person instructing legal.
  • Access to the tool you have.
  • Decisions within the agreed time in step two.

What is not included

The data protection documentation, which is a separate package under AI for legal. Configuration of the tool, which for Claude is part of the setup. Choice of tool, see legal technology selection. Legal advice on the department's matters.

What it is

Data protection answers whether a tool may be used. The operating model answers how:

  • Which workflows go through the tool, and which do not?
  • Who may use it for what?
  • How is the output reviewed before it is used?
  • Who owns the playbooks, templates and settings?
  • How do new staff learn it, and how does the department know whether it is worth it?

A tool that is right 98 times out of 100 sounds good. Run the work through five such steps and it is right all the way through about 90 times out of 100. That is why governance and review are needed.

The engagement ends in a document of ten to fifteen pages that the department owns and can change itself. It does not depend on the tool.

Step by step

Step What happens What you have afterwards
1 Mapping: interviews with the lawyers and three to five of the people who instruct legal A picture of the department's workflows with current AI use marked
2 Decisions with the department on where AI is used, with review and responsibility per workflow A decision per workflow and the list of matters that stay out of the tool
3 The operating model is written up and reviewed with the general counsel The operating model, a twelve-month plan and a date for the next review

What is included

  1. Mapping, step one. HOVR Legal interviews the lawyers and three to five of the people in the business who instruct legal. Where does the time go, which matter types recur, where is AI already used and with what result? The result is a picture of the department's workflows with current AI use marked.
  2. Decisions, step two. Together with the department, HOVR Legal decides in which workflows AI is used by default, in which it is allowed but not the default, and in which it is not used. For each workflow: what goes in, what review is required before the answer is used, and who is responsible. This is where the list of matters that stay out of the tool belongs, if it exists, and otherwise it is drawn up.
  3. Governance and roles. Who owns the tool in the department, who changes playbooks and settings, and how are decisions on new uses made? How does it fit with the company's wider AI governance? Roles are written down with names.
  4. Skills. What each role needs to know, how it is taught, and how it is documented as a measure under Article 4 of the AI Act [2]. A two-hour training session is included.
  5. Measurement. Three to five measures that can be produced without new systems: lead time per matter type, share of matters that go through the standard workflow, number of review deviations, time in the tool. And a date for the first reading.
  6. The operating model, step three. Everything above in one document, reviewed with the general counsel, with a plan for the next twelve months and the next review.

Add-on: a reading after ninety days

The operating model sets the measures and the baseline. Ninety days later HOVR Legal goes through, if you want, what the measures show: which workflows work as intended, where use is low and why, which rules need changing, and whether a tool or a flow should be discontinued. You get a two-page reading with the adjustments made to the operating model and the decisions the general counsel needs to take. The add-on can be done with the engagement or later.

What you get

  • The operating model as a document in your template, with workflows, roles, review rules and measures.
  • The list of matters that stay out of the tool, if it did not already exist.
  • Training material and documentation of the training.
  • A twelve-month plan with prioritised next steps.

Time

Scope and time are agreed in the first conversation.

Next step

One hour about how the department uses AI today and whether the engagement fits. Write to hello@hovrlegal.com.

Read also AI literacy training for lawyers and an AI policy for the legal department.

Ask a question or choose a time for a conversation.

Get in touch

Back to AI for legal

More under AI for legal