Where Do I Even Start With AI?
Pick a real problem. Ship something small. Build from there.
You already know you should be doing something
I hear a version of this almost every week.
A managing partner or a COO sits down and says it out loud. We know we should be doing something with AI. We just don't know where to start.
They have sat through the demos. Someone on the team has a chat window open all day. A competitor dropped AI into a proposal. The board asked about it once, and nobody had a clean answer.
If you run an accounting firm, an architecture practice, an engineering company, a staffing firm, or an advisory shop, that feeling is familiar. You are busy. The work is already complicated. Adding "an AI initiative" on top of it does not make the first step any clearer.
The firms getting real value are not the ones with the fanciest tools. They picked a real problem, shipped something small, and built from there.
I have put more than 300 AI systems into production for professional services firms. The ones that paid off started the same way. AI is a business decision. The software comes after you know which problem you are solving.
The wrong first question
The first question I get is almost always this one. What AI tool should we buy?
It feels like the responsible move. You can compare vendors. You can put a recommendation in front of the partners. You can tell the board you did something.
Then the licenses sit half used. The work still lives in dozens of systems that don't talk to each other. Six months later, nobody can point to a change a client or a staff member would notice.
A tool purchase is a line item. It doesn't tell you which problem is worth solving first.
A tool is only one of four things companies get sold. Dev shops sell a build. Trainers sell a workshop. Strategy firms sell a plan. Each one can be the right piece. None of them owns whether the business actually runs differently.
What we did at a 200-person architecture firm
A 200-person architecture firm came to me with that exact request. Do something with AI. There was real interest in the room. There was no clear starting point. Project files, drawings, email, and HR documents were spread across more than 50 disconnected systems.
I didn't open with a platform or a strategy deck. I asked where people were losing time, and what frustrated them every week.
One of the first things we looked at was software crashes. We went through about 180 crashes across 42 users. Outdated software versions were behind roughly 20% of them.
180
crashes reviewed
42
users in the sample
20%
tied to old versions
Nobody would put that on a conference stage. It gave people time back. Designers stopped losing work to a crash they could have avoided. And it built trust fast, because the finding was specific and the fix was obvious.
From there we built three more tools. An internal AI chat tool the team could use on firm knowledge. An automated drawing check that runs every few minutes. An HR policy assistant, so people could get a straight answer without hunting through a handbook.
Four working tools in about 90 days.
After that, leadership approved a 12-month AI roadmap and kept me on to lead it. The roadmap followed the proof. It didn't come first.
The right first question
The better question is smaller. What is annoying, slow, or expensive right now?
Not what AI might do next year. What already costs your people hours, rework, or money this month.
In a professional services firm, it usually sounds ordinary. A proposal that takes three days to pull together. A drawing set checked by hand late at night. A policy question that lands in the same inbox every Tuesday. A report rebuilt from five exports. A software problem everyone has learned to live with.
Those are starting points. They belong to a person. In a few weeks, you can tell whether they got better.
Four things you can do this month
You don't need a task force or a vendor shortlist to find the first win.
- 1
Ask the people doing the work
Sit with two or three of them for an hour. Ask what they redo, what they wait on, and what they complain about on a normal Tuesday. Write it down in their words.
- 2
Pick one problem you can count
Crashes. Hours. Turnaround time. How often the same question gets asked. If you cannot tell whether it improved, it is the wrong first project.
- 3
Ship something small
A check that runs on a schedule. A search box over the policies people already ask about. A short note each week that flags the same failure. Small enough that a skeptical partner can try it without a rollout plan.
- 4
Show the result, then pick the next one
Tell the team what you found and what changed. The second project is easier once people have watched one thing work on their own work.
The first win does not need to be impressive
It needs to be real.
Leaders skip the small problems because they don't sound like an AI initiative. That is how a firm stays in the "we should do something" conversation for another year. They wait for a use case big enough to announce, and they never learn how their own work actually runs.
The architecture firm did not earn a roadmap by picking the flashiest idea in the room. They earned it by fixing crashes people felt every week. Once that landed, the chat tool, the drawing check, and the policy assistant had a reason to exist. People already trusted that the work was pointed at their problems.
Real means someone got time back, a number moved, or a recurring frustration got smaller. If you can say that in a staff meeting, you have started. The impressive version can wait.
Book 30 minutes
If you are sitting with that same blank page, bring the problem you are stuck on, the tool someone wants you to buy, or the question your partners keep asking.
Or email anthony@anthonywhitaker.com
Work With Me
I work as a fractional AI lead for a limited number of small and mid-sized businesses at a time. Tell me about your situation and I'll be in touch within 24 hours.
Retainers and the 90-day AI Accelerator available.
What to expect:
- ✓ 20-minute conversation to understand your context
- ✓ Quick assessment of AI opportunities in your operations
- ✓ Honest take on what's worth pursuing (and what's not)
- ✓ No obligation, just clarity on your next steps