AI Opportunity Audit Worksheet

Find the repeated work in your business that is actually worth automating — before you spend a dollar on tools.

Outcome: After completing this worksheet, you can inventory your team's repeated tasks, score each one for value, feasibility, and risk, and name your top three automation candidates with reasons you can defend to your team.

Page 1 — About This Worksheet

What this is

Context: Most businesses don't have an "AI problem." They have a "where do we even start?" problem. Everyone is busy, several tasks feel painful, and every vendor claims their tool is the answer. This worksheet replaces guessing with a short, structured audit: you list the repeated work, score it honestly, and let the scores point to a starting place.

Explanation: You will do four things, in order:

1. Inventory — list every recurring task that takes meaningful time each week.
2. Score — rate each task for value, feasibility, and risk using the guides below.
3. Prioritize — place tasks on a simple matrix so the best candidates stand out.
4. Select — write up your top three candidates with a success measure for each.

Who this is for

Owners, office managers, practice managers, and operations leads at teams of roughly 2–50 people. No technical background needed.

When to use it

• Before you buy or subscribe to any AI or automation tool
• Before a Discovery & Roadmap conversation (it makes that meeting twice as productive)
• Once or twice a year, or whenever "we need to hire someone" might really mean "we need to stop doing this manually"

Warnings — read before filling it in

• This is a planning tool, not a promise of savings. Scores are structured judgment, not measurement. Validate the top candidates with real numbers before acting.
• Fill it in with the people who do the work. Scores assigned from the owner's chair alone are usually wrong, and surprise automation damages trust.
• Do not paste customer records into AI tools while doing this audit. You only need task names and rough counts — never names, account numbers, or health or payment details.
• High-sensitivity tasks are never "full automation" candidates, no matter how well they score. At most, they qualify for AI-drafted work that a human reviews and approves.

How the columns work

For each task, record:

Scoring method

Value (V), 1–5. Start from monthly hours, then adjust:

Adjust +1 if doing the task late or badly directly costs revenue or customers. Adjust −1 if nobody notices when it slips. Keep V between 1 and 5.

Feasibility (F), 1–5. How automatable is it today?

• 5 — Rules are clear, the data is already digital, the process rarely changes
• 3 — Mostly clear rules with regular exceptions a human must catch
• 1 — Judgment call every time, or the information lives on paper and in people's heads

Risk (R), 1–5. What happens if the automation gets it wrong?

• 1 — Low stakes, easy to undo, no sensitive data
• 3 — Customer-facing or money-adjacent; mistakes are embarrassing but recoverable
• 5 — Sensitive or regulated data (health, payroll, legal), or errors that are hard to reverse

Priority Score = V × F × (6 − R) — the range is 1 to 125.

Rule that overrides every score: any task with R of 4 or 5 is capped at "AI drafts, human approves" — it never runs unattended.

Prioritization matrix

Plot each task by Value (vertical) and Feasibility (horizontal). Flag any task with R ≥ 4 with a ⚠.

Page 2 — COMPLETED EXAMPLE (Fictional)

Summit Air Heating & Cooling — a fictional 8-person HVAC company: an owner, an office manager, a dispatcher, four field technicians, and a part-time bookkeeper. All names and numbers below are invented for illustration.

Step 1 & 2 — Task inventory

V notes: #2 and #3 adjusted +1 (missed after-hours calls lose emergency jobs to competitors; no-shows cost booked revenue). #9 stays low — helpful but nobody suffers when it slips a day.

Step 2 — Scores and bands

Step 3 — Matrix placement

• Quick wins: reminders & confirmations (3), after-hours calls (2), renewals (6), business-hours calls (1)
• Big bets: replacement quotes (4) ⚠ — high value, but judgment-heavy and pricing errors are costly; assist-only, later
• Fill-ins: review requests (9), invoice follow-up (5), timesheets (7)
• Skip for now: warranty paperwork (10); payroll (8) ⚠ stays with the bookkeeper and existing payroll software

Tie-break note: business-hours calls and after-hours calls both scored 80. Summit chose after-hours first — it changes nothing about the dispatcher's day, and every missed evening emergency call is revenue walking to a competitor. Business-hours support becomes phase two, after the after-hours agent proves itself.

Step 4 — Top-3 candidates worksheet

Candidate 1 — Appointment reminders & confirmations (score 100)

• Why it scored high: clear rules, digital calendar data, low risk, real no-show cost
• What "automated" looks like: system sends reminder 48h and 4h before each visit; customer can confirm or rebook by reply
• What stays human: rescheduling conflicts, upset customers, same-day emergencies
• Data touched: name, phone, appointment time (no payment or health data)
• Success measure: no-show rate — fictional baseline 12% of visits; target under 8% within 60 days; office time on confirmations, baseline ~10 hrs/month
• First step: export last month's no-show count so the baseline is real

Candidate 2 — After-hours & overflow call answering (score 80)

• Why: ~25 missed calls/week; emergencies booked by whoever answers first
• What "automated" looks like: a voice agent answers after 6 p.m., books standard visits, and routes true emergencies to the on-call tech immediately
• What stays human: emergency dispatch decision, pricing questions, angry callers
• Data touched: name, phone, address, problem description
• Success measure: answered-call rate after hours (fictional baseline ~20%; target over 90%) and booked jobs from evening calls
• First step: pull two weeks of voicemail logs to confirm volume

Candidate 3 — Maintenance-plan renewal outreach (score 75)

• Why: pure follow-the-list work with recurring revenue attached
• What "automated" looks like: drafts renewal reminders from the plan list; office manager reviews the batch and hits send
• What stays human: pricing exceptions, long-time customers who deserve a personal call
• Data touched: name, contact info, plan expiry date
• Success measure: renewal rate (fictional baseline 61%; target 70%+ this season)
• First step: clean the plan-expiry spreadsheet — automation can't fix missing dates

Page 3 — BLANK VERSION

Work with the people who actually do these tasks. Estimates are fine; honesty about sensitivity and rules-clarity matters more than precision.

Task inventory & scores

Priority = V × F × (6 − R). Bands: 60+ strong candidate · 30–59 worth exploring · under 30 not yet. Flag R ≥ 4 with ⚠ (assist-only).

Matrix

• Quick wins (high value, high feasibility): ______________________________
• Big bets (high value, low feasibility): ______________________________
• Fill-ins (low value, high feasibility): ______________________________
• Skip for now: ______________________________

Top-3 candidates (complete one block per candidate)

Candidate ___ — Task: ______________________ (score ____)

• Why it scored high: ______________________________
• What "automated" looks like: ______________________________
• What stays human: ______________________________
• Data touched: ______________________________
• Success measure (baseline → target, with a date): ______________________________
• First step this week: ______________________________

(Repeat for Candidates 2 and 3.)

Review Checklist — before acting on the results

• [ ] Every recurring task that takes 30+ minutes a week is on the list (including the "invisible" ones — chasing, re-typing, reminding)
• [ ] The people who do each task reviewed its scores
• [ ] Data sensitivity is honest — anything touching health, payroll, legal, or payment data is marked High
• [ ] No task with R ≥ 4 is slated for unattended automation
• [ ] Each top-3 candidate has a measurable baseline, not a guess
• [ ] Someone owns collecting real numbers for each baseline
• [ ] No customer records were pasted into any AI tool during this audit

Safety note: This audit surfaces where automation could go, not where it must. Keep humans in the loop for anything involving money, health, legal exposure, or an upset customer — AI supports judgment; it doesn't replace it. And roll out results transparently: your team should hear "here's what we found and why" from you, never discover an automation by surprise.

Practice (10 minutes): Before scheduling the full session, fill in one row right now for the task you complained about most this week. If scoring one row takes under three minutes, you're ready to run the whole audit.

Next step: Take your top three candidates to the AI ROI Estimator (/academy/demos/ai-roi-estimator) to put dollar and hour figures on each. If two or more land in the "strong candidate" band, that's exactly what a Discovery & Roadmap session is for — book a call at /contact and bring this worksheet.