Business cases for legal technology tend to fail in the same meeting. The slide says reviews get faster and lawyers get more productive; the CFO asks where the numbers came from; the answer is another company, another contract mix, another year. This post gives you no numbers. It gives you a model, the measurements that fill it and the questions finance will ask, so the case is built from your own legal function.

In short

  • Measure before you model. A two-week log of where counsel time goes on each contract type is worth more than any benchmark, including any we could give you.
  • Value comes from four places. Counsel hours on preparation, cycle time to signature, overflow sent to outside counsel, and the leakage after signing: renewals that roll over unnoticed and obligations nobody chases.
  • Cost is more than the licence. Encoding the playbook, connecting the contract repository, change management and the running cost of models belong in the case from the start.
  • The sensitive variable is review effort. How long a lawyer spends checking a prepared matter, rather than preparing it, decides most of the outcome. Test the case against a pessimistic value.

Start with a time sample

Pick a representative fortnight. For a sample of matters of each common type — NDAs, supplier MSAs, SaaS order forms, statements of work — ask counsel to log minutes against a few coarse activities. Honesty matters more than precision.

ActivityPreparation or judgement?Measure
Triage: type, paper, personal data, what is missingPreparationMinutes per matter
Finding and reading the clausesPreparationMinutes per draft
Comparing with the company’s positionsMostly preparationMinutes per draft
Cross-checks: exhibits, DPA, parent agreementPreparationMinutes per draft, and how often something is missing
Drafting redlines and the cover notePreparationMinutes per draft
Re-reading revised draftsPreparationMinutes per round, and rounds per matter
Deciding what to accept; escalating to the General CounselJudgementMinutes per matter
ApprovalJudgementMinutes per matter, and days waiting

Alongside the sample, pull four numbers from your systems: matters opened per year by type, median days from request to signature, the share of matters sent to outside counsel and what that costs, and the number of signed contracts that auto-renew. These are the baseline the case is measured against.

The model

Write the case as formulas with named inputs, so finance can change any input and see the effect. For each contract type t:

# inputs you measure
N_t        matters per year of type t
P_t        preparation minutes per matter today      # all rounds
J_t        judgement minutes per matter today
R_t        minutes to check a prepared matter       # estimate, then measure in a pilot
C          fully loaded cost per counsel minute

# counsel capacity released
minutes_today    = Σ_t N_t × (P_t + J_t)
minutes_after    = Σ_t N_t × (R_t + J_t)
capacity_value   = (minutes_today − minutes_after) × C

# overflow brought back in house (only if capacity is actually redeployed)
O          outside counsel spend on routine review today
share_back share of that work the released capacity can absorb
overflow_value   = O × share_back

# after signing: renewals and obligations
A          auto-renewing contracts per year
miss       share whose notice deadline passed without a decision   # from a look-back
would_end  share of those the owner would have ended or renegotiated
V_r        average annual value of such a contract
renewal_value    = A × miss × would_end × V_r × share_recoverable

annual_value     = capacity_value + overflow_value + renewal_value

Every input is yours. Leave out any line you can’t measure; the case should survive without it.

Where each input comes from

Two-week
time sample

P_t, J_t

Matter and
request systems

N_t, cycle time

Finance:
outside counsel spend

O

Signed repository
look-back

A, miss, V_r

Pilot on
real matters

R_t, share_back

The model

The one input you can’t measure today, R_t, is what the pilot is for.

The renewal look-back

Take last year’s auto-renewing contracts and ask, for each, whether anyone made a decision before the notice deadline. The deadline is the end of the term less the notice period stated in the executed copy; on a twelve-month term with ninety days’ notice it falls three months before the end. For the ones that renewed by default, ask the owner whether they would have ended or renegotiated. This is the only honest source for miss and would_end.

Cycle time

Faster signature has value when revenue or an operational start date waits on the contract — a sales order form, a supplier the plant needs. Model it only for those types, and only with a figure the business owner will stand behind.

The cost side

CostWhat to include
Licence and hostingThe platform and its hosting
Encoding the playbookCounsel time to write positions, fallback language and approvers for the contract types in scope, and to test them against recent drafts
Connecting sourcesThe intake queue, the contract repository and its executed copies
Change managementTraining, a period of dual running, and updating approval policy into controls
Model running costPer-agent spend is recorded on each step; with the model off the agents run on rules, so this line can be zero
Ongoing ownershipSomeone in legal operations keeps the playbook and the controls current

Run a pilot that measures R

A pilot designed to fill the modelFour to six weeks on one or two contract types
1
Choose the typesLegal operations

High-volume, mostly counterparty paper: supplier MSAs and SaaS order forms are typical.

2
Encode the playbook for themSenior counsel

Positions, fallback language, severity, approver.

3
Run live matters through itCounsel

The agents prepare; counsel checks and decides as usual.

4
Log the checking timeCounsel

Minutes from opening the prepared matter to a decision: this is R_t.

5
Record agreementThe platform

How often counsel accepts what the agents prepared, by agent and action.

6
Compare and decideGeneral Counsel and finance

Re-run the model with measured values, including a pessimistic R_t.

The agent activity page: what the agents did across matters and what waits on a person
What the pilot reads from: every agent step on every matter, and which prepared decisions were answered. Demo data on a fictional company.

For the General Counsel

The case rests on your own time sample and look-back, so you can defend every input. The renewal line is often the one a CFO understands first.

Keep the risk benefits — consistent positions, DPA checks, four-eyes approval — out of the arithmetic and state them plainly beside it.

For legal operations

You own the measurements: the time sample, the pilot log and the repository look-back.

For the board

The question is not whether contract review can be faster but what the legal team does with the time, and what the company stops losing after signing.

No borrowed numbers

We deliberately give no benchmark figures here. A saving measured at another company, on another contract mix, will not survive your CFO’s first question; a saving measured on your own matters will.

Build the case from your own matters, and let the pilot fill in the one number you can’t know yet.

Preparation minutes, judgement minutes, outside counsel overflow and renewals missed: all measurable before you buy anything. Checking time is measurable in a month.

Keep reading