The 30-Day Shippable AI Scorecard
Most AI sold to small businesses is priced and timelined like enterprise AI — a 6-month pilot, a six-figure SOW, a discovery phase before any shipped work. For a $5M–$50M business with 20–100 employees, that math doesn't work and it never did.
What does work: pick one specific workflow that wastes a person's time every week, ship something that fixes just that, in 30 days or less, for the cost of a mid-level contractor for a month. No pilot. No SOW. If it doesn't work in 30 days, you paid for one month of a senior person and you own the artifact.
The hard part is picking the right workflow. Pick wrong, and 30 days isn't enough. Pick right, and 30 days is plenty. This scorecard is a 10-minute worksheet that tells you whether a workflow you have in mind is in the shippable zone.
The five tests
A workflow fits the 30-day shippable profile when it scores yes on all five tests below. Score it like a referee — partial counts as no for purposes of picking. You're looking for one workflow that's a clean yes-yes-yes-yes-yes.
If nothing in your business scores all five, the answer isn't "AI later." The answer is that the upstream work — figuring out which workflow is shippable — is where to spend the first conversation.
Test 1 — Is it ONE workflow, with a name?
Not "the back office." Not "operations." Not "customer service." A named workflow is something a specific person does on a specific cadence, where you could describe what they do in one sentence and what triggers it in another.
Passes: "the dispatcher routes service tickets every morning between 7 and 9am" / "the office manager chases unpaid invoices Monday afternoons" / "the estimator writes quotes from intake calls within 48 hours of the call."
Fails: "we want to automate operations" / "customer service could use AI" / "the back office takes too long."
Test 2 — Does it eat 5+ hours a week of one specific named person?
Not "it eats time." A specific person, a specific weekly hour count. If you can't name the person and the hours, the workflow isn't measured well enough to ship against.
Passes: "my office manager spends 6–8 hours/week on invoice follow-up" / "my service dispatcher spends 60–90 minutes every morning on routing, longer in heating season" / "I personally spend a full Saturday morning every week writing estimates."
Fails: "a few hours here and there" / "the team spends a lot of time on it" / "it's hard to say exactly."
Test 3 — Is the work mostly the same shape every time?
The workflow involves the same kind of decision, applied to a stream of similar inputs. Not identical — similar. AI is good at "this looks like the last hundred of these and the answer was X." AI is bad at "this is a brand-new situation that requires judgment we've never applied before."
Passes: invoice follow-up (same email, different customer) / quote generation (same fields, different specs) / service ticket routing (same dispatch rules, different jobs) / appointment confirmation (same script, different patient).
Fails: customer dispute resolution (every dispute is unique) / new-product strategy (judgment-heavy, low-volume) / onboarding a new senior hire (high-stakes, one-off).
Test 4 — Does the output go somewhere measurable?
After the AI does the work, the output lands somewhere you can audit — an email sent, a quote written, a ticket routed, a calendar entry created. If the output is "advice the person reads and decides what to do with," that's harder to ship in 30 days because measuring whether the AI is helping requires watching the person's behavior change.
Passes: email drafted and queued for review / quote generated and saved to CRM / calendar entry created and confirmed / invoice flagged for follow-up.
Fails: "AI gives the office manager suggestions" / "AI summarizes the situation for me" / "AI helps the team think about it."
Test 5 — Are you OK with the AI being right ~85% of the time, with the rest reviewed?
Not 100%. The shippable profile assumes a human reviews the AI's output before it goes out the door — at least for the first few months. If the workflow requires the AI to be 100% right with no human in the loop, that's a different scope (and a different price tag, and a different timeline). 30-day shippable means the AI takes the work from "0% to 85%" and a human takes it the last 15%.
Passes: AI drafts the invoice follow-up, office manager reviews and sends. AI generates the quote, you check the numbers and send. AI routes the ticket, dispatcher confirms.
Fails: AI directly debits customer accounts. AI signs off on regulatory filings. AI sends emails to clients without review.
What a yes-yes-yes-yes-yes tells you
If a workflow passes all five tests, it's in the 30-day shippable zone. The next conversation is about how — what the AI does specifically, what the human reviews, where the output lands, how you'll measure whether it's working. That conversation is short. Picking the workflow was the hard part.
If a workflow passes 4 of 5, you may be one tweak away. Often the failed test is Test 1 (the workflow is too broadly defined) or Test 2 (the time cost isn't measured yet). Sharpen and re-score.
If you're at 3 or fewer, the workflow isn't shippable in 30 days. That's fine — it just means a different conversation. Sometimes the right move is to re-define the workflow more narrowly until it passes. Sometimes the workflow genuinely isn't AI-tractable yet. Either way, the scorecard tells you which conversation you're in.
What the scorecard isn't
It isn't a vendor-deck. It isn't a sales pipeline. It isn't a way to qualify you as a buyer. The scorecard exists because most local businesses I talk to have at least one workflow that scores cleanly — and they didn't realize it because they'd been told AI was a six-month pilot project. The 30-day frame is real. Most owner-operators just need a tool that helps them see which workflow fits.
Fill it out for yourself. Or fill it out with your office manager, dispatcher, or estimator — whoever's hours are at stake. The first conversation worth having is the one where the named person on the named workflow is in the room.