Take two accounting firms with the same headcount, in the same town. One does monthly bookkeeping for a few dozen small businesses. The other does tax returns and year-end planning, and most of its year happens between January and April.
Hand both owners the same list of "AI ideas for accounting firms" and they'll circle different items. They should. The first firm's week is a steady cycle of chasing bank statements and closing books. The second firm's year is a pile-up of returns that all need the same documents at the same time.
Same industry. Different work. Different first move.
That's why an industry list is a good place to start and a poor place to stop. The better question is how work actually moves through your business, because that decides where it waits, and the wait is usually where AI pays off first.
Industry tells you what you sell. Delivery tells you where work waits.
Most owner-led service businesses deliver their work in one of four ways. Many run two, but one usually dominates the week.
- Appointments. Work happens in scheduled sessions: practices, advisors, coaches, consultants who bill by the meeting.
- Projects. Work has a start, a scope, and a finish: agencies, production firms, PR, implementation and IT projects.
- Recurring cycles. The same work repeats every week or month for the same clients: managed IT, bookkeeping, payroll and HR support, retainers.
- Crews in the field. Work happens at the customer's location and someone has to route people to it: trades, cleaning, inspections, home services.
Each one has a predictable place where work stalls. Find yours and the shortlist for a first AI workflow gets very short.
Appointment-based work: it waits before and after the session
In an appointment business, the session itself is the part you're good at. The trouble sits on either side of it.
Before the session, someone collects forms, confirms the time, reschedules, and tries to fill the gap when someone cancels. After it, someone writes up notes, sends a follow-up, books the next visit, and handles billing questions.
Good first candidates:
- Collecting intake information before the visit, so the session doesn't start with paperwork.
- Reminders and rescheduling that run without a person re-keying calendars.
- A first draft of the session summary or follow-up email, which the professional reviews and sends.
The test is simple: does work sit waiting for a form, a confirmation, or a write-up? If yes, start there. One caution: anything touching health, financial, or legal information needs a privacy check on the tools before anyone types a client's details into them.
Project work: it waits at the edges of scope
Project firms rarely lose time in the middle of a project. They lose it at the start, the handoffs, and every time the scope changes.
Proposals get rebuilt from old ones. Kickoff information gets asked for twice. Status updates eat Friday afternoon. A client asks for "one small change," and nobody writes down that it's out of scope until the invoice conversation.
Good first candidates:
- Assembling first-draft proposals and scopes from your past proposals and a short intake call.
- Weekly client status updates drafted from your project tool, reviewed by the project lead.
- Flagging requests that look like scope changes, so someone decides on purpose instead of by accident.
Notice that the thinking stays with your people. Pricing, scope, and the creative work are judgment calls. AI handles the assembly around them.
Recurring work: it waits on missing pieces
Recurring work looks easy to automate because it repeats. But the repetition is usually not the slow part. The slow part is waiting on things: a client who hasn't sent statements, a ticket nobody has looked at, a monthly report someone rebuilds by hand.
Good first candidates:
- Chasing missing documents and information on a schedule, with a person handling the clients who still don't respond.
- Sorting and routing incoming requests or tickets so the right person sees them first.
- Drafting the monthly client report from data you already have.
If your team says "we're waiting on the client" more than any other sentence, the first workflow is probably the chase, not the work itself.
Field work: it waits between the office and the job
When the work happens at a customer's location, the expensive gaps are between the office and the crew. A job gets booked without the right details. Someone drives out without the part. Notes from the job don't make it back in time to invoice.
Good first candidates:
- Capturing complete job details at booking, so the crew shows up with what they need.
- Turning a tech's job notes or photos into a clean work summary and draft invoice.
- Following up on open quotes that never got a yes or a no.
Dispatch and routing decisions can come later. The first win is usually making sure information arrives complete, on both ends.
Find your wait points in an afternoon
You don't need a consultant to see your own pattern. Try this:
- Pick three jobs you finished last month. Choose ordinary ones, not the disasters.
- Trace each from first contact to paid. Write every step in plain language, including the ones that feel too small to mention.
- Mark every place the work waited. Note what it was waiting on, whether that was a person, a document, an approval, or a reply, and roughly how long.
- Circle the wait that shows up in all three. That's your strongest candidate.
- Check it against the basics. Does it happen often, is it repeatable, and does the delay cost money or customers? We walk through that ranking in where to put your first AI lever.
If the wait that keeps showing up is "waiting on the owner," that's a different problem, and adding a tool won't fix it. Our piece on growing without every decision landing on your desk covers that one.
A hypothetical, to make it concrete
Picture a 20-person managed IT firm. It's a made-up example, but a typical shape. Most revenue is recurring support contracts, with a few projects a year.
An industry list for IT services would point at a dozen things, from automated security reports to AI chat for end users. The trace tells a narrower story. On all three jobs, the slowest stretch is the same: a ticket comes in, sits unassigned, then gets bounced because it went to the wrong tech.
So the first workflow is sorting and routing incoming tickets, with a dispatcher approving the assignments for the first few weeks. It's not glamorous. It's measurable from day one, using time to first response and tickets reassigned, and it touches nearly every client every week.
The numbers that matter here would be your own. Count them before you build anything. Then keep the returns separate: faster response is a service and capacity gain, and it only becomes saved cost or added revenue if you act on it. More on that distinction in AI ROI for small businesses.
Before you build it
Choosing the workflow is step one. Before anything gets automated, check that the process underneath is consistent, the data is where you think it is, and someone owns the result. Before You Automate walks through those checks, and From Industry AI Ideas to Your Business Case covers what to measure so you'll know if it worked.
Questions to ask this week
- Which of the four delivery patterns describes most of our week?
- Where did our last three jobs wait, and what were they waiting on?
- Which of those waits costs us money or customers, not just patience?
- Who would own the first workflow, and how would we know in a month if it helped?
Where to start
If you want a head start, grab the free AI Leverage Report for your industry. It shows where AI tends to pay off in businesses like yours. Then run the three-job trace to see which of those ideas fits how your firm actually works.
And if you'd like a second set of eyes on what you find, the free 30-minute AI Leverage Call is built for exactly that. Bring the three jobs. We'll help you pick the first workflow worth building.
