The Process Audit: How to Find the Three Things Worth Automating First

The hardest part of AI adoption isn't the technology. It's choosing where to start.

Most companies either pick the flashiest opportunity (usually too complex, fails in 90 days) or the safest one (usually too trivial, nobody cares). The goal is to find the three processes that sit in between: meaningful enough to matter, contained enough to ship.

Here's the audit framework I use to find them.

Step 1: Map the Pain, Not the Potential

Before you think about what AI can do, document what's actually causing friction. Spend a week asking every team lead one question: What does your team do regularly that feels like a waste of their skills?

Write everything down without filtering. You're looking for patterns. Common categories:

• Data assembly — pulling numbers from multiple sources to build a report someone could act on
• Status communication — sending updates that require no judgment, just current information
• Routing and triage — deciding who or what to send something to based on known rules
• First-draft creation — producing an initial version of something that always gets revised anyway
• Lookup and retrieval — finding information that exists somewhere but takes time to locate

These categories are where automation consistently wins. If you can place a painful process into one of these buckets, you have a candidate.

Step 2: Score Each Candidate on Three Dimensions

For each process you've identified, assign a score from 1-3 on each dimension:

Frequency × Volume

How often does this happen, and how much of it is there? A task that happens 20 times a day is a better automation candidate than one that happens twice a quarter, even if the quarterly one takes longer. Frequency is what makes automation pay off quickly.

Rule Clarity

Can you write down the decision rules for this process in plain language? "If X, do Y. If Z, do W." The more clearly the logic can be stated, the more reliably it can be automated. Processes that require tacit knowledge, relationship context, or judgment calls are harder. Don't start there.

Blast Radius

What happens when automation gets it wrong? Low blast radius (internal, recoverable, low stakes) means you can learn fast. High blast radius (customer-facing, financial, compliance-adjacent) means you need more safeguards. Score your safest candidates highest here.

Add the three scores. Your top three scorers are your starting candidates.

Step 3: Validate with the Person Doing the Work

Before you commit to automating anything, sit down with the person who actually does that task today. Two things you need to know:

1. What do you know about this process that isn't written down anywhere? There's always something. Edge cases, exceptions, informal rules that evolved over time. These are the things that break naive automations. Document them before you build.
2. What would make you trust an automated version of this? This question tells you what oversight and verification the system needs to have. It also builds buy-in. People who helped design the system are more likely to use it and less likely to route around it.

Step 4: Define the Win Before You Build

For each of the three processes you're moving forward with, write one sentence that defines success. Specific, measurable, time-bound. Not "improve the process" — "reduce time-to-first-response from 4 hours to under 30 minutes by end of Q2."

This sentence does three things: it keeps the build scoped, it gives you an obvious stopping point, and it makes it easy to declare victory and move on to the next thing.

The Three You're Looking For

After running this audit, you want to end up with:

• One quick win — high frequency, crystal-clear rules, low blast radius. Something you can automate in two weeks and show results within a month.
• One meaningful improvement — moderate complexity, meaningful time savings, affects a process your team genuinely cares about. Takes 4-8 weeks to do well.
• One longer play — higher complexity, bigger payoff, requires more careful design. Start scoping it now so it's ready when the first two are running.

Three parallel tracks. Different timelines. Different risk profiles. This structure gives you early wins to build confidence while the harder work gets done right.

The Real Purpose of the Audit

The audit isn't just a prioritization tool. It's an organizational alignment tool. When you run it properly — talking to actual people doing actual work — you come out of it with a shared understanding of where the friction is and a credible plan for addressing it.

That shared understanding is worth more than any specific process you automate. It's the foundation that makes the second round of automation easier, faster, and better-received than the first.

Start there. The technology is the easy part.