Resources · Small business AI
This is the page I wished existed when I started talking to owner-operators about AI two years ago. No vendor pitches. No reselling. Just the questions an owner has to answer before spending money on AI tools, and the math behind each answer.
01
Probably yes for one or two specific workflows. Probably no as a general "we are an AI company now" posture. The shape of the question is workflow-by-workflow, not company-wide.
The three workflows where AI most often earns its keep in a small services business:
If your business has none of those shapes, AI is probably not your highest-leverage next move. The honest answer might be hiring the person you were already going to hire, or fixing a process that has been broken for two years.
02
The single repetitive workflow that costs you the most hours. Not the most exciting one. The most expensive one. If you cannot pick one, the audit format exists for that.
Time three real instances of the workflow before you install any tool. Write the numbers down. The most common cause of "AI did not work" is no baseline to compare against.
Pick one tool that targets the workflow. Use it on three new instances. Time those too. Two weeks is enough to see whether the tool fits or fights the workflow.
If the time saved is meaningful, expand. If not, cancel. Do not skip this step with hope replacing data. The 90-day mark is where most firms quietly let unused tools renew.
Assign one person to own the tool, the workflow, and the measurement. Without an owner, the system rots within 90 days. The owner does not have to be the founder, but it has to be one person.
03
A typical AI stack for a small services firm runs between $200 and $2,500 per month depending on size and workflow mix. Below is the math from the CPA buyer guide as a worked example. The shape applies broadly to dental, insurance, and other services firms with similar headcount.
| Layer | Tool | Monthly |
|---|---|---|
| Practice management + AI | Karbon Team, 15 seats | $885 |
| Document capture | Dext accountant plan | $300 |
| Tax software (amortized) | Drake Unlimited or ProConnect | $200 |
| 1099 filing (amortized) | Avalara 1099 / Tax1099 | $150 |
| Ad-hoc LLM | ChatGPT Team or Claude Team, 15 seats | $375 |
| Audit add-on (optional) | DataSnipper, ~5 seats | $250 |
| Total | $2,160 / mo · $26K / year |
Hour math: the peer-reviewed Stanford / MIT study found 3.5 hours per AI-using employee per 40-hour week. For 15 employees that is roughly 30 to 50 hours per week of capacity after a wait-and-see discount. At an internal cost of $75 / hour, that recovered capacity is worth $150K per year. At $200 / hour billable, $400K. The stack pays for itself if the firm redirects roughly 130 hours per year to billable work, about 2.5 hours per week firm-wide.
The risk is not break-even on paper. The risk is whether recovered hours actually become billable. Firms that absorb hours into earlier evenings rather than client work see no P&L benefit at all.
04
In-house works when
A consultant works when
The math usually favors a consultant for the first 90 days, then an internal owner for the long run. The consultant gets you from zero to one fast. The internal owner gets you from one to lasting integration. Both roles matter; they are different jobs.
05
Five burns that have cost real small firms real money in the last 18 months. Compressed here; the CPA buyer guide has the full citations and mitigations.
06
Print this. Walk it before you sign with any AI vendor.
07
Probably for one or two specific workflows, not as a general 'we're an AI company now' posture. The shape of the question is: which of my workflows are repetitive, well-defined, and have enough volume to justify integration cost. If you have those workflows, AI is worth your attention. If not, the answer is honestly no.
Pick one workflow. Run a two-week test with one tool. Measure before-and-after time on three real instances. Decide based on those numbers. Then expand or stop. Most owners try to start with five tools at once and end up with shelfware.
Stack cost runs between $200 and $2,500 per month for a typical 5 to 25-person services firm. For a 15-person CPA firm, an independent buyer guide priced a sensible stack at roughly $26,000 per year. The risk isn't the stack price; it's whether recovered hours become billable hours. If they get absorbed into earlier evenings instead of client work, the P&L doesn't move.
Yes if you have specific repetitive workflows AND a way to convert recovered time into either billable work or capacity for new clients. No if AI is going to recover hours that disappear into 'we just go home earlier now.' The answer lives in your operations, not in the tool.
The peer-reviewed Stanford/MIT study found 3.5 hours per week per AI-using employee. Vendor surveys claim 18-plus hours. Use the peer-reviewed number for planning. For a 15-person firm at 3.5 hours per person per week, that is 30 to 50 hours per week of capacity (after a wait-and-see discount). At a $75 internal cost or $200 billable rate, that pays back the stack quickly if the hours redirect to chargeable work.
If you have someone on staff who can spend two days a week on this and won't be pulled into client emergencies, build in-house. Most small firms don't. A consultant runs $999 for an audit (DomeWorks pricing) up to $15K for a fixed-scope build. The math usually favors a consultant for the first 90 days. Past that, an internal owner of the system is what makes it stick.
When you've tried two or three tools, none of them stuck, and you can't pinpoint why. Or when the workflows you're considering touch regulated data (PHI, attorney-client, confidential client records) and you want someone to flag where AI is the wrong answer before you ship it. The audit format is faster and cheaper than a paid pilot of the wrong tool.
Vendor concentration risk (signing with platforms that can shut down: Bench, Botkeeper). Believing the 5x time-savings overstatement (vendor numbers are about 5x the peer-reviewed independent figure). Trusting AI on regulated content without a human review gate. Tier-locked pricing surprises (the AI you saw demoed is usually one tier above the price quoted). Pasting client data into consumer chat tools.
Anything that requires legal or fiduciary judgment with personal liability attached. Anything where the input is a small number of high-value cases, not a high volume of similar ones. Anything involving regulated client data where you cannot guarantee enterprise-tier privacy. Broken processes, where AI just makes broken faster. The audit format flags these explicitly.
Two to four weeks for the first measurable workflow. Three to six months to install AI into the way the firm runs. Twelve months for cultural integration where AI is the default and not the experiment. The 90-day mark is where most firms decide whether they keep going or shelve it.
The DomeWorks AI Audit is a 45-minute discovery call plus a written action plan in 48 hours. Tells you which 3 to 7 tools fit your workflows and which workflows AI should not touch. $999 flat. Refund if it doesn't earn the price.