The Four Questions to Ask Before Automating Anything

I’ve spent the better part of two years showing enterprise legal teams what AI can do in discovery. Our demos keep getting better. What I can put in front of a room today would have looked like science fiction eighteen months ago. What’s changed is the reaction. It’s closer to polite exhaustion now. Somebody watches an agent do something remarkable, nods, tells me it’s impressive, and then goes back to their desk to reformat a privilege log export because opposing counsel wants the date fields a different way.

These are not slow organizations afraid of technology. Some of the sharpest thinking about AI in legal is happening in those rooms. Those same people answer to model risk, validation, security, data processing agreements, retention rules, old outside counsel guidelines, and procurement cycles measured in trimesters. Behind it sits a regulator who may eventually ask what you did and why.

And the docket keeps growing anyway. More matters, more data, more sources that did not exist five years ago, and flat budgets in the good years. Legal teams need to move faster inside organizations built to make new technology move slowly. That does not mean they have to wait eleven months to solve problems today.

Four questions

The instinct is to go after the biggest pain point. Resist it. If the last article was about turning playbooks into skills, this is about choosing the first piece of the playbook worth automating. The biggest pain point is usually the hardest judgment call in the matter, which makes it the worst place to start. Use four questions instead: Do I do this in every matter of this type? Could I write down how to do it correctly in about two pages? Does doing it right actually require my license? Do I need access to production systems to do this?

Do I do this in every matter of this type?

You only get paid back on work you do over and over. The thing you did once, in the awful case, stays manual no matter how much it hurt. Pull your last five matters of one type and write down what your team did in all five. The better candidate is usually the small task somebody does forty times a year without anyone noticing.

Could I write down how to do it correctly in about two pages?

If it takes less than a page, the task is probably trivial. If it runs past five, you have several tasks bundled together, or a judgment call sitting in the middle nobody has noticed. Two pages is useful because it forces the work into the open: the inputs, the rules, the exceptions, and the point where a person needs to decide. It also exposes fragility before you automate anything.

Does doing it right actually require my license?

Plenty of discovery work is difficult, consequential, and needs no law degree at all. A privilege log export is an obvious place to start. Fields get reconciled against the ESI protocol, dates are normalized across messy metadata, attorney names are highlighted, and the final version is cleaned up to match what opposing counsel demanded. No legal judgment is required. None of us went to law school to learn conditional formatting.

The same deliverable can hold both kinds of work. The export mechanics are formatting. The privilege narratives are not. Tools can help identify and log privileged documents, but that is probably not where you start if the goal is a safe, repeatable first workflow. It requires independent judgment and runs straight into the fourth question.

Do I need access to production systems to do this?

Production systems deserve the highest scrutiny. They hold the record, trigger downstream consequences, and sit closest to client data, audit obligations, and legal work product. We can do far more than most production tools expose today, but in high-risk, highly regulated environments, restraint is not a flaw. Vendors implement AI deliberately, often a few clicks behind the frontier, because those workflows are on rails to keep people out of trouble.

But rails are not where most tinkering happens. If you are figuring out what should be automated, how the workflow should run, what exceptions matter, or whether the process saves time, the production system may be the wrong first place to learn. A lot of critical work happens one layer out: spreadsheets, exports, trackers, checklists, Teams threads, SharePoint folders, and the documents people already use to move the matter forward. If the task can be tested with approved tools already in front of you, start there. Prove it works, then bring that evidence back to the systems that need more control.

The first step is smaller than the demo

The big version of AI in discovery will always be there, but that doesn’t mean you have to wait for it. This framework lets you attack the less glamorous parts of your job.  It may not be exciting, but it is where legal teams lose hours every week. Automate it safely, measure it clearly, and bring that proof back when the organization is ready for the next layer.