AI consulting we'd never recommend

If a consultant tries to sell you “AI strategy” before they’ve asked what’s broken, they’re selling you their availability, not their thinking.

Here are five engagements we wouldn’t take, and five questions we’d ask before we’d quote you anything.

Five engagements we won’t take

1. “Build us a chatbot.” 90% of chatbots get installed and ignored. The real problem is usually a knowledge base that doesn’t exist. We’d fix that first.

2. “Train AI on our entire internal wiki.” Most wikis are 60% outdated. Training AI on bad data makes the bad data more accessible, not less.

3. “Replace our salespeople with AI.” AI is excellent at the first 80% of qualification. It’s terrible at the last 20%, where deals close. Hand it the triage, let humans close.

4. “Generate all our marketing with AI.” SMB buyers can spot synthesized voice in two paragraphs. If your competitor has a real voice, you can’t compete with synthetic.

5. “Build us an AI strategy.” Without a specific business problem, “AI strategy” is shelf-art. We’d want to know what’s broken first.

Five questions we’d ask before quoting

1. What did the person doing this work say when you mentioned AI? If they’re scared, the engagement won’t stick. If they’re relieved, you’ve found the right problem.

2. What does this workflow cost you in real dollars per week? If you can’t math it, it isn’t bad enough to fix.

3. What have you already tried? Tells us whether the problem is “AI” or “the person who tried it last time.”

4. Who owns this when we leave? Workflows without owners don’t survive 90 days. Period.

5. What would “this didn’t work” look like in 30 days? If you can’t define failure, you can’t measure success.

The reason most AI consulting fails isn’t bad AI. It’s good AI applied to the wrong problem, with the wrong owner, on a workflow that should’ve been deleted — paid for through a contract that didn’t measure anything. We’ve watched it happen at scale. We’re not going to let it happen at yours.