Start with a result you can measure
Choose an AI project by asking which change would move the business toward its current goal. Then check whether AI addresses the cause of the constraint. The best candidate may be a process change, a tool you already own, or a focused build.
“Use AI across the business” doesn’t tell your team what to prioritize. “Complete more client work with the same team” gives them something to investigate. Define the unit of progress: completed matters, accepted analyses, qualified opportunities, or another result the business needs.
Write down the baseline, target, time period, and owner. Keep the first version simple enough that someone can challenge it.
Find where the work waits
Follow one real piece of work from request to completion. Separate time spent doing the task from time spent waiting. Ask what prevents the next piece of work from moving forward.
If drafts wait four days for a partner’s review, generating drafts twice as fast may make the queue larger. The useful intervention could be clearer review criteria, a different approval rule, or better support for the reviewer. Draft generation is a candidate only if it changes the limiting step.
- Where does unfinished work accumulate?
- Whose availability controls the pace?
- What sends work backward for correction?
- What would happen downstream if this step became faster?
These questions turn a vague complaint into something you can observe. They also stop a local time saving from masquerading as business growth.
Compare the ways to change it
Once you understand the constraint, compare three paths: change the process, configure a tool you already have, or build something new. Each path needs an owner and a way to test whether it works.
A new build makes sense when the workflow needs capabilities the other options can’t provide within the relevant access, quality, and operating requirements. It comes with its own cost: implementation, maintenance, adoption, and ongoing review.
For a practical comparison, see our build-or-buy guide. The decision should follow the workflow, rather than a preference for a particular model or platform.
Define what would count as progress
Choose one business outcome and a small number of supporting measures before work starts. A useful measurement plan distinguishes throughput, quality, adoption, and cost. Don’t combine them into one claim without explaining the relationship.
In our AM100 law firm engagement, use in the corporate practice rose from 32% to 72% after the mandate was lifted. That matters because people chose to use the system. It still isn’t a revenue figure. The scope and measurement method belong beside the result.
Write the decision rule in advance: if the result improves and the quality checks hold, continue; if it doesn’t, investigate or stop. Keep the first test small enough to learn from without committing the whole business.
Use a one-page decision brief
A useful project brief answers six questions:
- What business result are we trying to change?
- What evidence identifies the limiting step?
- What are the plausible ways to change it?
- What information, access, and people does each option need?
- How will we measure the result and check quality?
- Who will own the workflow after the project ends?
If the answers are weak, the next step is investigation. If they’re strong, you have a basis for choosing the work and setting its scope.
Find your next useful step.
Share your goal. We’ll investigate the likely bottleneck and recommend what to do next.
Start your free assessment