DEEP DIVE
Most AI Audits Are Backwards
Don't start with where AI could fit. Start with where the business leaks time, information, and revenue. Sometimes the fix isn't AI at all.
A lot of AI audits start with the technology. That almost guarantees the wrong conversation.
The consultant walks in and asks, "Where could we use AI?" Then comes the list. AI chatbot. AI meeting notes. AI email drafting. AI reporting. AI agent. AI whatever.
You can find dozens of places to jam AI into any company. That doesn't mean you should, and it definitely doesn't mean the owner just bought a plan.
Drew
The Happy Path
THE DEMO
Department โ AI opportunity โ estimated time savings โ roadmap
It looks strategic because the deck is long.
Interview the team. Find 47 opportunities. Estimate hours saved, multiply by a loaded hourly rate, and present an "AI Transformation Roadmap" with an ROI number big enough to justify the audit fee.
Invoice paid. Everybody goes home.
Then Reality Hits
Now someone actually has to build it.
REALITY
The real constraint might be a broken process, duplicate software, unclear ownership, bad data, or a task that shouldn't exist at all.
Adding AI to a bad process makes the bad process faster.
The CRM data is garbage. The API doesn't support what the audit promised. Three people do the same process three different ways. There's no clean trigger. The owner changes the rules every other Tuesday. And opportunity #31, the AI quote assistant, turns out to be useless because pricing actually lives in the owner's head.
Meanwhile the stuff that's really costing money didn't make the deck, because it didn't sound like AI. Leads nobody calls back. Invoices that go out ten days late. The owner approving every purchase order over $500.
The business pays twice: once for the audit, and again in the months the team spends discovering what the audit missed.
I'm not against audits. A good one is worth a lot. But the deliverable shouldn't be a list of places AI could go. It should be a short list of problems worth solving, what each one costs today, and a realistic plan for who builds it and who owns it after.
The Play
Find the expensive friction first. Pick the technology second.
Start with operational pain:
Where does work stall?
Where do people copy data by hand?
Where are customers waiting?
Where does the owner get pulled in?
Where does someone rebuild the same report every Monday?
Where does revenue leak because follow-up depends on memory?
Then pick the mechanism, roughly in this order: process redesign, software configuration, integration, deterministic automation, AI assistance, human approval, full autonomy.
Sometimes the answer is a better HubSpot setup. Sometimes it's deleting three tools and keeping one. Sometimes it's a policy change. And sometimes AI is exactly right. Earn that conclusion.
Three Takeaways
Ideation is cheap. Finding 40 possible automations is easy. Finding the five that matter is the work. Most decks never get past that first part.
Price the build, not the idea. An opportunity without a delivery estimate is a wish.
"Not AI" is a valid finding. If the best fix is a config change, the audit did its job.
Steal This
STEAL THIS
The Better AI Audit
About 2 to 3 hours with the owner and one person from each department.
Find repeated work that eats meaningful time every week.
Find handoffs where information gets lost or delayed.
Find decisions that wait on the owner but shouldn't.
Find systems that disagree or force double entry.
Put a dollar or hour cost on each one.
For each, pick the fix: process, software, integration, automation, AI, or some mix.
Build only where the expected value beats the complexity.
Take the top five from step 5. Park the rest for a quarter.
Close
Finding the five things that matter, designing them around the real business, building them, and making sure people actually use them is hard. That's where the value should be. Delivery beats ideation.
Got an AI audit deck sitting in a drawer? Hit reply and tell me how many of the opportunities actually got built. I have a guess, and I read every reply.
BUILD SYSTEMS THAT SURVIVE REALITY
If one workflow keeps breaking inside the real business (lead follow-up, onboarding, invoicing, reporting), SMB Ops will scope it, build it, and hand it over. Bigger than one workflow? Same place, bigger conversation.
Drew Reynolds
The SMB Operator