free playbook · read it here · no email required

The AI Opportunity Playbook for Small Business.

Where AI actually creates value in a small business, industry by industry, and the questions to ask before you spend a dollar. Written by the engineer who builds these systems. Read it, print it, send it to a friend who owns a business. You will not be added to a list.

in this playbook

part one

Before you buy anything.

Three numbers explain the whole small-business AI market. 58% of small businesses already use AI in some form, so access is not the problem. Only about 5% of company AI projects ever make it past a pilot into real daily use, so follow-through is. And businesses that bring in outside help get there roughly twice as often as the ones going it alone.

That gap is not about picking the wrong tool. It is about the unglamorous work of wiring AI into the places where the work actually happens, and keeping it running when something goes sideways. Keep that in mind every time a product demo looks like magic: the demo is the easy 5%.

The six questions. Before any AI spend, ask: Will it increase revenue? Will it reduce operating costs? Will it improve your customers’ experience? Will it make your people more productive? Will the return be measurable? Is it practical for a business your size? If the answers are no, do not buy it, no matter how impressive the demo. This is the same gate we apply to our own recommendations.

The data rule. Decide where your data is allowed to go before deciding what AI does with it. Consumer chatbot accounts are not a place for client records, patient data, or financials. The legitimate paths are business tiers with real data agreements, or AI running on hardware you own. The full argument is in is your business data safe with ChatGPT and frontier vs private AI.

Buy or build. If a well-supported product already does the job, buy it. Custom work earns its keep where your process is specific, where tools need to talk to each other, or where the data cannot leave. Anyone selling you a build without checking the off-the-shelf answer first is selling, not advising.

part two

The opportunity, industry by industry.

Each chapter follows the same shape: the problems every owner in the industry recognizes, where AI creates value, one workflow made concrete, and the questions to ask before investing.

Home services & trades

the common problems

Quotes go out during the busy season and follow-up dies because everyone is in the field. Invoicing slips to the weekend. Maintenance agreements and seasonal contracts lapse because nobody is watching the renewal dates in the middle of a heat wave.

where AI creates value

Written estimates from a tech’s photos and field notes
Follow-up on every open quote, approved by you before it sends
Renewal tracking and chasing for agreements and seasonal contracts
Same-day invoicing and aging reminders
Review responses and appointment reminders in your voice

one workflow, made concrete

A tech finishes a water heater estimate visit and sends three photos and two sentences to the office. The system drafts a written quote with your pricing for approval, sends it, then follows up at day 3, day 8, and day 15 unless the customer books or declines. Every send waits for your ok.

questions to ask before investing

How many open quotes died from silence last quarter?
How many days pass between job done and invoice sent?
Who is watching renewal dates in your busiest month?

Where to start: Start with follow-up on open quotes. It is money you already half earned. Full write-up: AI for home services.

Medical & dental practices

the common problems

Patients quietly go overdue for hygiene, follow-ups, and accepted treatment plans. Claim denials sit unappealed because appeals take time nobody has. And every fix is governed by HIPAA, which most off-the-shelf AI tools do not survive.

where AI creates value

Recall and reactivation tracking with outreach drafted for staff approval
Insurance denial appeals assembled from the record for the biller’s review
Intake and verification paperwork prepared before the visit
Review responses that never discuss care

one workflow, made concrete

Each week the system lists every patient overdue for hygiene or a stranded treatment plan, drafts the outreach in the practice’s voice, and queues it for front-desk approval. Patient data stays inside systems covered by a business associate agreement, or on hardware inside the practice.

questions to ask before investing

Do we have a business associate agreement with every AI vendor touching patient data?
How many denials went unappealed last year, and what were they worth?
How many patients are overdue right now?

Where to start: Settle the data question first: business tier with a BAA, or on-premise. Then start with recalls. Full write-up: AI for medical & dental offices.

Accounting firms

the common problems

Client documents arrive in every format and someone retypes them. The numbers live in tools that do not talk to each other. Season multiplies routine client back-and-forth until the team is answering email instead of doing the work clients pay for.

where AI creates value

Numbers pulled off documents and receipts automatically
Reconciliation across systems that do not integrate
Client intake and requests sorted and routed
Routine replies drafted for review, sent in your name

one workflow, made concrete

A client uploads a shoebox of PDFs. The system extracts the figures, flags the ones it is unsure about, reconciles against the ledger, and leaves a short exception list for a human. The reply to the client is drafted and waiting. Nothing posts to the books without sign-off.

questions to ask before investing

How many staff hours went to document entry last season?
Where do client financials travel when staff use AI tools today?
What share of client emails are the same twenty questions?

Where to start: Start with document extraction on one document type you see every day. Full write-up: AI for accounting firms.

Restaurants

the common problems

Food cost creeps invoice by invoice and gets noticed at month-end, if at all. Reviews decide who walks in and most go unanswered. Catering quotes sit half sold because follow-up never happens between services.

where AI creates value

Supplier invoices matched against orders, price creep flagged weekly
Every review answered within a day, in your voice, for your approval
Catering and event quotes chased until they close
Schedules drafted, menu copy kept consistent everywhere online

one workflow, made concrete

Invoices get read as they arrive and matched against what was ordered and last week’s prices. When the produce supplier raises a line item 9%, it shows up in Monday’s summary instead of the month-end P&L, with the email to the rep already drafted.

questions to ask before investing

When did you last catch a price increase the week it happened?
What share of your reviews got a response this month?
How many catering inquiries went quiet without a decision?

Where to start: Start with invoice matching. It pays for itself out of leaks you cannot currently see. Full write-up: AI for restaurants.

Auto repair shops

the common problems

The biggest pile of pre-sold revenue in the building is declined work, and nobody follows up on it. Reminders go out late or never. Estimates and invoices lag behind the bays.

where AI creates value

Declined-work follow-up, timed and worded for your approval
Service and inspection reminders that actually go out on schedule
Estimates drafted from the tech’s notes and photos
Invoices out the day the car leaves, review responses in your voice

one workflow, made concrete

A customer declines rear brakes at 4mm. Ninety days later the system drafts a check-in referencing the actual measurement and the miles since, and queues it for the service writer’s ok. The customer books, and the job that walked out the door walks back in.

questions to ask before investing

What was declined in your shop last quarter, and who followed up?
Are inspection and service reminders going out on time today?
How long after the car leaves does the invoice go out?

Where to start: Start with declined-work follow-up. The list already exists in your shop software. Full write-up: AI for auto repair shops.

Property managers

the common problems

The maintenance inbox eats more staff time than anything else. Lease renewals sneak up and become vacancies. Owner reports consume days every month that were supposed to go to growing the portfolio.

where AI creates value

Every maintenance request read, classified, and routed with the right urgency
Lease renewals tracked and surfaced months early
Monthly owner reports drafted from data you already have
Listing descriptions written, rent reminders sent before balances grow

one workflow, made concrete

A tenant emails "the sink is leaking again" at 9pm. The system reads it, checks the unit’s history, sees two prior visits, classifies it as recurring plumbing, and routes it to the right vendor queue with the history attached, flagged for morning review.

questions to ask before investing

How many renewals surprised you into a vacancy last year?
How many staff hours go to reading and routing requests?
Do owner reports go out on time every month?

Where to start: Start with maintenance triage. It frees the most staff time the fastest. Full write-up: AI for property managers.

Real estate agents

the common problems

Speed to lead decides who gets the client. The past-client base is where most future business lives, and it goes cold the moment you get busy, which is exactly when it matters most.

where AI creates value

New leads answered within minutes with a real, specific response
Past clients kept warm with personal touches drafted for approval
Listing descriptions written from the property details
Transaction deadlines tracked so nothing slips

one workflow, made concrete

A portal lead arrives while you are at a showing. Within two minutes the inquirer gets a real answer about the property and a question back. You get a summary and the drafted follow-up when you are out, instead of a cold lead three hours later.

questions to ask before investing

How fast did your last ten leads get a real reply?
When did each past client last hear from you personally?
What is tracking your deadlines during the busiest weeks?

Where to start: Start with speed to lead. Research has shown for over a decade that response within the hour changes outcomes. Full write-up: AI for real estate agents.

part three

Your first 30 days.

Week one: find the leak. Pick the single most repetitive money leak in your business. Your chapter above names the usual suspect. Write down what it costs you now, even roughly. That number is your baseline, and without it you will never know whether anything worked.

Week two: settle the data question. Decide what data the system is allowed to touch and where it is allowed to go. If your industry is regulated, this decision comes before any tool, not after.

Weeks three and four: run one thing, with approval loops. Automate the one leak, and keep a person approving what goes out until it has earned trust. One process, done properly and measured against your baseline, beats five pilots every time. That is exactly the 5% trap from part one.

You can do this yourself, and our guide on DIY vs bringing in help is honest about when that makes sense. If you want a second pair of eyes, the fixed-price assessment produces this exact plan against your actual numbers, and the free call before it costs nothing.

Book the free call → See your industry →

sources

US Chamber of Commerce, Empowering Small Business (2025): small business AI adoption.
MIT, The GenAI Divide: State of AI in Business (2025): pilot-to-production rates and the effect of outside help.
Harvard Business Review, The Short Life of Online Sales Leads (2011): lead response timing.

Updated July 2026. Free to read and share. No email, no list, no gate.