26Point Checklist

The framework · Technology & Future

23 Automation & AI Leverage

12 sub-categories. The three numbers on each are the typical opportunity in that area across businesses — not your result.

Time9.3/10Financial7.8/10Automation9.3/10

These three numbers are researched cross-business benchmarks for the area — the typical opportunity available in it. They are not your score, not a forecast, and not a promise. What is actually critical depends on your business, which is what the audit is for.

23.1

Repetitive work

Time10/10Financial8/10Automation10/10
What this is
Work that is done the same way, repeatedly, by a person.
Why it matters
Ten for time and ten for automation. Repetitive work is where automation pays back fastest and most predictably, and every business has more of it than it thinks.
You have a problem here if
Someone can describe their task as a sequence of steps with no judgement in it.
What to automate
List the repetitive tasks, multiply each by frequency and time, and rank by total hours. Automate top-down. Simplify or eliminate the task before automating it — automating waste just makes it faster.
23.2

Manual administration

Time10/10Financial7/10Automation10/10
What this is
Data entry, filing, scheduling, chasing, reconciling and re-keying.
Why it matters
Ten out of ten for time. Administration is the largest single consumer of owner and staff hours in most small businesses and produces no value the customer can see.
You have a problem here if
Several hours a week are spent moving information between systems.
What to automate
Attack it in order: eliminate the need, then integrate so it happens once, then automate what remains. Integration usually removes more admin than any dedicated automation tool.
23.3

Workflow automation

Time10/10Financial8/10Automation10/10
What this is
Multi-step processes that run themselves, with people only where judgement is required.
Why it matters
Ten for time and ten for automation. Workflow automation is what allows a business to grow throughput without growing headcount proportionally.
You have a problem here if
Processes stall waiting for someone to notice it is their turn.
What to automate
Automate hand-offs, notifications, task creation and status. Model the process honestly first, including the exceptions — automation that only handles the ideal path pushes all the difficulty onto the people handling the rest.
23.4

System-to-system integration

Time10/10Financial8/10Automation10/10
What this is
Systems exchanging data directly rather than through a person.
Why it matters
Ten and ten. This is the plumbing that makes everything else in the category possible, and its absence is the reason most automation projects stall.
You have a problem here if
Your automation ideas all require someone to export a file first.
What to automate
Prioritise integration ahead of tooling. A connected stack with modest tools outperforms a disconnected stack with excellent ones.
23.5

AI assistance

Time9/10Financial8/10Automation10/10
What this is
Using AI to help people work faster — drafting, summarising, analysing, answering.
Why it matters
Nine for time, ten for automation. Assistance is the lowest-risk form of AI adoption because a person remains in the loop and accountable.
You have a problem here if
Staff spend significant time drafting routine documents and summarising information.
What to automate
Start here rather than with autonomous systems: drafting replies, summarising calls and documents, preparing first-pass analysis. Set a clear policy on confidential data, and verify anything factual before it leaves the business.
23.6

AI/agent opportunities

Time9/10Financial8/10Automation10/10
What this is
Work that could be handed to an AI agent to run rather than merely assist with.
Why it matters
Nine and ten. Agentic automation extends reach substantially, and it also extends the consequences of a mistake, which is why it needs governance rather than enthusiasm.
You have a problem here if
You have processes that are fully rule-based and still consume a person.
What to automate
Choose tasks that are reversible, observable and bounded. Require an audit trail. Run in shadow mode against human output before handing over. Never give an agent an irreversible action — payment, deletion, external commitment — without an approval step.
23.7

Knowledge retrieval

Time10/10Financial7/10Automation10/10
What this is
Finding information the business already holds, when it is needed.
Why it matters
Ten for time, ten for automation. Search is invisible time: minutes at a stretch, dozens of times a day, and never recorded anywhere.
You have a problem here if
People ask a colleague for information that exists in a document somewhere.
What to automate
A well-scoped retrieval system over your own documents, procedures and history is one of the highest-return AI applications for a small business. It requires the content to exist first — which is why documentation is worth doing.
23.8

Automated reporting

Time10/10Financial7/10Automation10/10
What this is
Reports that produce and distribute themselves.
Why it matters
Ten and ten. Manual reporting is recurring, entirely mechanical work whose only value is in being read.
You have a problem here if
A person builds the same report every week.
What to automate
Automate assembly, generation and distribution. Add an interpretation layer only once the numbers are trusted — commentary generated on top of doubtful data spreads the doubt rather than resolving it.
23.9

Human approval requirements

Time8/10Financial6/10Automation7/10
What this is
Deciding deliberately where a human must approve before something proceeds.
Why it matters
Seven out of ten and lower than the rest of the category, which is the point. Approvals are the control that makes automation safe, and each one is a deliberate trade of speed for assurance.
You have a problem here if
Either everything needs approval, or nothing does.
What to automate
Set thresholds by risk and reversibility rather than by habit: automate the routine, require approval for the material, and always for anything irreversible or externally visible. Review the thresholds as confidence grows — an approval gate nobody ever declines is theatre.
23.10

Exception handling

Time8/10Financial7/10Automation8/10
What this is
What happens when an automated process meets a case it cannot handle.
Why it matters
Exceptions determine whether automation actually saves time. A system that automates 80% and dumps the rest, unlabelled, on a person can leave the workload unchanged.
You have a problem here if
Automated processes fail silently and are discovered later.
What to automate
Design exception handling with the happy path, not after it: detection, clear routing to a person, enough context to resolve it, and reporting on exception rates by cause. Falling exception rates are the real measure of an automation maturing.
23.11

Productivity improvement

Time10/10Financial10/10Automation9/10
What this is
The actual improvement in output per person from automation to date.
Why it matters
Ten for time and ten financially. This is the outcome the whole category exists to produce, and it is the honest test of whether the tooling worked.
You have a problem here if
You have automated several things and cannot say whether anyone has more time.
What to automate
Measure before and after: hours on the task, throughput, error rate. Automation that does not release time or capacity has usually just moved the work somewhere less visible.
23.12

Automation ROI

Time8/10Financial10/10Automation8/10
What this is
What each automation cost and what it returned.
Why it matters
Ten out of ten financially. Tracking return is what stops automation becoming a technology habit rather than a business decision.
You have a problem here if
Automation spend has never been evaluated after the fact.
What to automate
Record expected benefit before building, measure actual after, and review. Kill the ones that did not work — including, when the evidence says so, ones you were personally attached to.

Category totals

Time 112/120 · Financial 94/120 · Automation 112/120. The sum is the official roll-up; the averages above exist so categories of different sizes can be compared.