How to audit your business for AI leverage.

Most AI projects fail for the same reason most software projects fail: the team optimizes for the tool, not the workflow. They buy GPT licenses, run a workshop, and wait for ROI. Six months later, the only people using it are the early adopters who would have found leverage in any tool.

The audit framework below starts from the opposite end: which decisions, handoffs, or knowledge-retrieval steps in your business are slow, expensive, or error-prone? That's where AI belongs — applied to the system you already have, not bolted on as a new one.

Before any AI use case earns budget, it has to clear three filters:

1. Is this work repeatable? AI gets cheap when it runs hundreds of times. One-off creative work rarely justifies the integration cost.
2. Do we have the data? Garbage in, garbage out. If the inputs live in 4 systems with no clean source of truth, the AI is the wrong project — fix the data first.
3. Can a human still own the outcome? AI is an accelerator, not a delegate. If no one on the team owns the result, the workflow will quietly stop being used.

Eight times out of ten this disqualifies the use case you walked in with — and surfaces a better one.

The week-1 audit I run in client engagements has five steps. You can run it yourself:

1. List the top 10 recurring workflows in your business. By time spent, not importance.
2. For each, note: who owns it, how often it runs, how long it takes, where the friction is.
3. Mark which steps are information work (drafting, summarizing, classifying, searching).
4. Apply the 3 disqualifying questions to each information step.
5. Rank the survivors by weekly hours × people affected. That's your roadmap.

Once you've picked your first workflow, the pilot rules are simple:

• One workflow, one team, one owner. No "AI initiative."
• Define the kill switch. What specific metric must move by day 30?
• Measure baseline first. You can't claim a win without a before number.
• Weekly review. 30 minutes, every Friday, with the owner. Adjust or kill.

The hardest part isn't the first pilot — it's the second. The team that ran pilot #1 wants to keep iterating; the team that didn't doesn't see the relevance. Your job is to translate the pattern, not copy the tool. Run the same audit on a second team's workflows and you'll find a different #1 use case with the same characteristics.

That's the playbook. Audit → pilot → measure → translate. Most operators try to skip steps 1 and 3. Don't.