A8gent
A8gent
·

Template

Free AI Automation Audit Template (2026)

A task-inventory audit template that scores every recurring task by frequency, minutes, and rule-based-ness, then ranks your top three automation candidates.

What is inside

6 sections · copy-paste ready
Task inventory
Scoring method
Automation score table
Top 3 shortlist
Savings estimate
Worked example
sample-prompt.mdPreview
ROLE: You are a support agent for {company}.
CONTEXT: Use only {trusted_source}. Never
  invent facts or policies.
TASK:
  1. Classify the request and its intent.
  2. Draft a grounded reply with citations.
  3. Flag anything risky for human review.
GUARDRAIL: If confidence < 80%, escalate.
OUTPUT: A ready-to-send draft + one-line why.

Use it to

Inventory recurring tasksScore automation potentialRank the top 3 candidatesEstimate hours and dollars saved
AI Agent Starter Kit cover

Free download

AI Agent Starter Kit

Workflow picker, ROI worksheet, 40-prompt pack, and the first-agent rollout playbook. The 1-hour version of the entire A8gent system.

Workflow picker
ROI worksheet
Prompt checklist
Team rollout plan
Open it now

No spam. Worksheet + prompt checklist + rollout plan delivered instantly.

Contents
  1. Fast answer
  2. Quick facts
  3. What this covers
  4. Why it matters
  5. Where to go next
  6. Outcomes
  7. Step by step
  8. The template: columns to copy into a spreadsheet
  9. Why frequency times minutes times rule-based-ness
  10. Worked example: a five-van HVAC company
  11. Turning the shortlist into three builds
  12. At a glance
  13. Mistakes to avoid
  14. FAQ
  15. Related guides

How to use it

Ship this template in three steps

It is built to be forked. Copy it, make it yours, and put it to work on a real task the same day.

  1. 01

    Copy the structure

    Drop the template into your own doc or workspace. Every section is laid out and ready to fill.

  2. 02

    Fill in your context

    Swap the placeholders for your real workflow, trusted data sources, and review rules.

  3. 03

    Ship and iterate

    Run it on one real task with a human in the loop, then expand the parts that work.

Get the full template

Free and copy-paste ready. Preview it before you download.

Open template
TL;DRAnswer-first

Fast answer

An AI automation audit template is a one-page task inventory with a scoring column. You list every recurring task in the business, record how often it happens and how many minutes it takes, rate how rule-based it is, and multiply those three numbers into an automation score. Frequency times minutes gives you the annual time cost of the task; the rule-based rating tells you how likely software can actually do it without a human deciding each case. Sort the sheet by score, take the top three rows, and you have a shortlist backed by arithmetic instead of enthusiasm. The whole exercise takes about an hour for a small business and usually surfaces one or two tasks nobody thought of as automatable, because they were never counted. Use it before you buy any tool: most automation projects fail on the choice of what to automate, not on the build.

Facts last verified 2026-08

Quick facts

Time to complete
About an hour once a two-week task list exists
Rows to aim for
20-40 discrete tasks, not departments
Formula
Times per week x minutes per instance x rule-based rating (1-5)
Annual hours
Times per week x minutes x 52 / 60
Hard filters
Data must be reachable; customer-facing actions need an approval step
Estimate haircut
Assume a first version covers roughly 70-80% of instances

On this page

What this page covers

A template visitor should know what to fill in, when to use it, and which tool or course turns it into a working workflow.

  1. 01Task inventory
  2. 02Scoring method
  3. 03Automation score table
  4. 04Top 3 shortlist
  5. 05Savings estimate
  6. 06Worked example

Why does this matter now?

Small businesses rarely lose money on the wrong automation platform. They lose it by automating something that happens four times a month while a fifteen-times-a-day task keeps eating an owner's evenings. Counting the tasks converts a vague feeling of being busy into a ranked list with an hours-per-year number attached to each line, which is also the number you need to judge whether a tool subscription or a build is worth paying for. The scoring column does a second job: it filters out tasks that look repetitive but actually require judgment on every instance, which are the ones that produce disappointing automations and burn the team's trust in the idea. An audit you can finish in an hour is worth more than a strategy deck, because it ends with three named candidates and an estimate you can defend.

Internal path

Where to go next from this page

These links are part of the A8gent learning and conversion path. Use them to move from concept, to diagnosis, to workflow build, to course.

Start with readiness

What you should be able to do after this

  • Inventory recurring tasks
  • Score automation potential
  • Rank the top 3 candidates
  • Estimate hours and dollars saved

How do you do it, step by step?

1. List every recurring task for two weeks

Write down each task that repeats on a schedule or a trigger: answering the same question, chasing a quote, rekeying a form into the CRM, sending reminders, building the weekly report. Ask each person on the team for their own list rather than guessing on their behalf, because owners systematically underestimate admin work they do not personally touch. Aim for 20 to 40 rows; a list of six means you have described departments, not tasks.

2. Record frequency and minutes honestly

For each row put the number of times it happens per week and the realistic minutes it takes including the context-switch and the follow-up, not the best-case minutes. Timing three real instances beats estimating from memory, which tends to run low on the tasks that annoy people most. Frequency times minutes times 52 is the annual time cost, and that single column reorders most people's priorities immediately.

3. Score how rule-based each task is

Rate every task 1 to 5 on whether a clear rule decides the outcome: a 5 means anyone following a written instruction would produce the same result every time, a 1 means the answer depends on judgment, relationship, or context that lives in someone's head. Missed-call text-backs and appointment reminders sit at 5; pricing a complicated job sits at 2. This is the column that separates real candidates from expensive disappointments.

4. Multiply into an automation score

Compute score equals weekly frequency times minutes times the rule-based rating. The multiplication is deliberately crude, and it is meant to rank rows against each other rather than to produce a number that means anything on its own. Sort the sheet descending and look at what floated to the top before you form an opinion about it.

5. Sanity-check the top rows

For each of the top five, ask three questions: is the data the task needs already in a system a tool can reach, is there an owner who can say what a correct output looks like, and what is the cost of getting it wrong. A task with a high score but no accessible data or no owner drops down the list, and a task where an error reaches a customer needs an approval step in the design. This check is where the audit stops being a spreadsheet exercise.

6. Pick the top three and estimate the savings

Take the three surviving rows and estimate hours returned per year, then multiply by a loaded hourly rate for whoever currently does the work. Be conservative: assume the automation handles 70 to 80 percent of instances and a human still handles the exceptions, because that is what a first version realistically does. An estimate you have hedged is one you can show a partner or an accountant without having to walk it back later.

7. Build one, measure it, then re-run the audit

Ship the single highest-scoring candidate and measure the real before-and-after for a month rather than starting three projects at once. The measured result tells you whether your minute estimates were any good, which recalibrates every other row on the sheet. Re-run the audit quarterly; the list changes as the business grows and as the first automation removes work from it.

The template: columns to copy into a spreadsheet

Open a blank sheet and create eight columns, in this order: Task, Who does it, Trigger, Times per week, Minutes per instance, Rule-based (1-5), Data reachable? (Y/N), Automation score. Everything else people add to audit spreadsheets is decoration. The eight columns above are enough to rank thirty tasks and defend the ranking to whoever is paying for the build.

Fill one row per discrete repeating task. Task is a verb phrase such as reply to after-hours missed call, not a department such as customer service. Who does it matters because the loaded hourly cost of an owner and of a part-time admin differ by a factor that changes the shortlist. Trigger is what starts the task, and it is the single most useful column for the build phase: a task with a clean trigger such as a missed call, a form submission, or a calendar event is far easier to automate than one that starts because somebody remembered.

Times per week and minutes per instance are the two columns people fudge. Use a real count from a real week, and time three actual instances including the interruption cost and any follow-up, rather than reporting the ideal-case duration. Rule-based is a 1 to 5 judgment: 5 means a written instruction produces an identical result every time, 3 means most cases are routine but a meaningful minority need judgment, 1 means each instance is a decision. Data reachable is a blunt yes or no on whether the information the task consumes already sits in a system with an API or an export, because a no makes the build fragile no matter how attractive the score is.

The last column is the arithmetic: Automation score equals Times per week times Minutes per instance times Rule-based rating. Annual hours equals Times per week times Minutes per instance times 52 divided by 60, and it is worth adding as a ninth column because it is the number non-technical stakeholders actually respond to. Sort descending by score, mark any row with Data reachable set to N, and read the top ten.

Why frequency times minutes times rule-based-ness

The formula is crude on purpose. Frequency times minutes is the only honest measure of what a task costs the business, and it routinely reorders priorities, because the task everyone complains about is often a monthly two-hour job while the invisible one is a four-minute job that happens eleven times a day. Ranking by irritation instead of by annual hours is the most common way a small business ends up automating the wrong thing first.

The rule-based multiplier exists to stop the sheet from recommending something automation cannot do. A high-volume task that depends on judgment, such as quoting a nonstandard job or handling an upset customer, has a large time cost and a small automation score, which is the correct result. Multiplying rather than adding the rating is deliberate: it pushes judgment-heavy work down the list hard, which matches what actually happens when you try to automate it.

Two filters sit outside the arithmetic and override it. If the data the task needs is not reachable by software, the real project is getting that data into a system first. And if a mistake reaches a customer or touches money, the design needs an approval step, which reduces the hours saved and should be reflected in the estimate rather than discovered after launch. Score first, then apply both filters, then shortlist.

Worked example: a five-van HVAC company

Here is the sheet a residential HVAC company with five vans, one office manager, and an owner who answers the phone at night produced in about ninety minutes. Row one: return after-hours missed calls, done by the owner, triggered by a missed call, 25 times per week, 6 minutes each, rule-based 5, data reachable yes, score 750, roughly 130 hours per year. Row two: follow up on outstanding quotes, done by the office manager, triggered by a quote sent with no reply, 18 times per week, 5 minutes each, rule-based 4, data reachable yes, score 360, roughly 78 hours per year.

Row three: remind customers about tomorrow's appointment, office manager, triggered by a calendar event, 40 times per week, 2 minutes each, rule-based 5, data reachable yes, score 400, roughly 69 hours per year. Row four: rekey the intake form into the CRM, office manager, triggered by a web form, 15 times per week, 4 minutes each, rule-based 5, data reachable yes, score 300, roughly 52 hours per year. Row five: price a nonstandard replacement job, owner, triggered by a site visit, 6 times per week, 35 minutes each, rule-based 2, data reachable no, score 420, roughly 182 hours per year.

Note what the scoring did to row five. It is by far the largest time cost on the sheet at around 182 hours a year, and it still does not make the shortlist, because a rule-based rating of 2 and unreachable data mean any automation would produce quotes the owner has to redo. That is the sheet doing its job. The shortlist came out as: after-hours missed calls first at a score of 750, appointment reminders second at 400, quote follow-up third at 360, with CRM intake a close fourth worth doing in the same build because it shares a trigger with the quote workflow.

The savings estimate was written conservatively. Missed-call text-back was assumed to handle about 75 percent of after-hours calls without the owner picking up, so roughly 95 to 100 of the 130 hours return, valued at the owner's opportunity cost rather than a wage. Reminders were assumed to still need a human for the small share of customers who reply with something unusual, leaving roughly 50 of the 69 hours. Quote follow-up was estimated at roughly 60 of the 78 hours because replies still route to a person. None of these numbers are promises; they are the arithmetic with an explicit haircut, written so the owner could challenge each assumption.

Turning the shortlist into three builds

The three candidates that surfaced in the HVAC example are the three that surface for most appointment-driven service businesses, which is why each has its own build guide on this site. After-hours missed calls become a missed call text-back workflow, covered step by step in the n8n missed call text-back guide including the Twilio A2P 10DLC registration that trips up most first attempts. Outstanding quotes become a scheduled follow-up cadence with stop-on-reply logic, covered in the automate quote follow-up guide with the message templates. Appointment reminders become a two-way confirm and reschedule flow, covered in the AI appointment reminders guide.

Build one, not three. The first automation teaches you whether your minutes column was honest, and that recalibrates every other row on the sheet. It also gives you a measured before-and-after to show the team, which matters more for adoption than any feature of the tool you picked.

If you would rather have the audit run for you, the free AI automation audit does exactly this on your task inventory and returns the ranked shortlist with the estimate. If the shortlist points at work you do not want to build in-house, the automation services page covers the done-for-you path. Either way the sequence is the same: count the tasks, score them, and only then choose a tool.

At a glance

AI automation audit template - copy these columns into a spreadsheet

ColumnWhat to enterWhy it matters
TaskOne repeating task as a verb phraseDepartments cannot be scored; tasks can
Who does itRole that performs it todaySets the hourly cost used in the savings estimate
TriggerThe event that starts itA clean trigger is what a workflow tool listens for
Times per weekReal count from a real weekHalf of the time cost, and the half people guess wrong
Minutes per instanceTimed, including interruption and follow-upBest-case minutes understate the true cost
Rule-based (1-5)5 = written instruction gives identical resultsFilters out judgment work that automates badly
Data reachable? (Y/N)Is the input already in a system with an API or exportAn N means the real first project is the data
Automation scoreTimes per week x minutes x rule-basedRanks rows against each other; sort descending
Annual hoursTimes per week x minutes x 52 / 60The number non-technical stakeholders respond to

What mistakes should you avoid?

  • Listing departments or goals instead of discrete repeating tasks
  • Estimating minutes from memory rather than timing a few real instances
  • Scoring a judgment-heavy task as rule-based because it feels repetitive
  • Ranking by annoyance instead of by frequency times minutes
  • Ignoring whether the data the task needs is actually reachable by software
  • Starting all three top candidates at once instead of proving one
  • Claiming full automation in the savings estimate when exceptions will always need a human

FAQ

What is an AI automation audit?
It is a structured inventory of a business's recurring tasks, scored so you can rank which ones are worth automating. You capture frequency, minutes per instance, and how rule-based the task is, multiply them into a score, and shortlist the top few. The output is a ranked list with an estimated hours-saved figure, not a technology recommendation.
How long does the audit take?
Most small businesses can complete it in about an hour once the task list exists, though the list itself is better if you collect it over a week or two of real work. Rushing the frequency and minutes columns is the main way the result goes wrong. Timing three real instances of your top candidates is worth the extra afternoon.
How do I score how rule-based a task is?
Ask whether a new hire following a written instruction would reach the same outcome every time. If yes, it scores 4 or 5; if the answer depends on judgment, relationships, or missing context, it scores 1 or 2. Tasks in the 4 to 5 band are where automation tends to work on the first attempt.
What counts as a good automation score?
The score is only meaningful relative to the other rows on your own sheet, so there is no universal threshold. In practice the top three rows of a 30-row inventory are the ones worth scoping, and everything below the top ten rarely justifies the build and maintenance effort. Compare rows, not numbers.
How should I estimate the savings?
Multiply hours returned per year by the loaded hourly cost of whoever does the task now, then discount it because a first version usually handles 70 to 80 percent of cases and a human still covers exceptions. State the assumption next to the number so it can be checked. Estimates that hide their assumptions are the ones that get discredited later.
What if the top task needs data we do not have in a system?
Then the real first project is getting that data into a system, not the automation itself. A task whose inputs live in someone's inbox or on paper will produce a fragile build and constant exceptions. Move it down the list and pick the highest-scoring row whose data is already reachable.
Can you run this audit for us?
Yes. The free AI automation audit walks your task inventory with you, scores it, and returns the shortlist plus an estimate, and you can also see the done-for-you build options on the automation services page. The template here is the same method we use, so running it yourself first is a perfectly good outcome.

Sources & further reading

Was this page helpful?