What to automate first in your business: a five-question scoring method

Quick answer: what to automate first in your business? Score each repeated task on five questions: how often it happens, what a mistake costs, whether it follows fixed steps, whether the data behind it is clean, and how hard it is to build. A frequent task that is costly to get wrong, rule based and fed by clean data usually wins. Below are the five questions, six invented examples scored, and a worked example with the hours and cost it would save.

Owners rarely lack ideas for what AI could help with. The hard part is choosing which repeated task to fix first, out of a long list that all feel urgent. A short scoring method turns that choice into numbers you can compare, instead of a guess based on whichever task annoyed you most this morning.

Most owners can list ten repeated tasks without thinking hard. The list rarely shrinks on its own, because new tools and busy weeks keep adding to it faster than anyone finds time to fix the old ones. Scoring the list, instead of picking whichever task is loudest this week, is what turns a backlog into a short, ordered plan you can actually work through.

What five questions decide what to automate first?

These are the same five questions our own AI audit asks about every manual task it finds, so you can use them before you ever book a call.

• check_circleHow often does it happen, and how long does it take each time?
• check_circleWhat does a mistake cost, and who notices it?
• check_circleDoes it follow the same steps each time, or does it need judgment?
• check_circleIs the information it needs somewhere you can reach and trust?
• check_circleHow hard is it to build, and what does it cost to run?

A task that is frequent, rule based and fed by clean data usually ranks high. A rare task, a task that depends on judgment, or a task fed by messy data usually ranks low. The fourth question, about trustworthy data, matches a wider point made by InterTradeIreland's seven AI readiness questions for SMEs: AI can work quickly with information, but it cannot tell you that a spreadsheet is out of date.

How do you turn the five questions into a score you can compare?

Rate each task 1 to 3 on every question, then add the five numbers for a total out of 15. Question 5 is reversed: a cheap, easy build scores 3, and a hard, expensive one scores 1. The shape of this, plotting payoff against ease of implementation, borrows from the PICK chart, a Lean Six Sigma tool originally built at Lockheed Martin to sort process improvement ideas into what to implement, what is merely possible, what to challenge and what to kill.

A task scoring 13 to 15 is a strong first pick. A task scoring 5 to 8 usually belongs on a list of things not to build yet. If two tasks tie, pick the one that saves more hours a week, using the arithmetic in the next section.

Six example tasks, scored (the figures are invented)

Here is the method applied to six common business tasks, with made-up figures to show how the scoring works. Replace the tasks and the numbers with your own.

The follow-up email task comes out on top: it happens often, the data (who was quoted, and when) already sits in one place, and a drafted reply is cheap to build. Deciding which lead to call first scores lower overall because it still needs judgment, even though it happens often and a slow response is costly.

What does the top task save? A worked example

Turn the winning task into a time and a cost, so the choice is about numbers you can check, not a feeling. Here is the method with invented figures for the follow-up email task. Replace them with yours.

Run the same arithmetic for your own top-scoring task before you build anything. If the saving does not clearly beat what the fix would cost to build and run, the task is not ready yet, whatever it scored. The AI audit does this measurement with your real numbers, across every task at once, and ends in a ranked, written plan. Run the scoring again every few months: new tools change what counts as easy to build, and a task that scored low can move up once the data behind it gets cleaner.

Traps that make you automate the wrong task first

A high score does not guarantee a good result. These four traps turn a well-scored task into wasted work, even when the arithmetic above looks convincing.

• check_circleA task nobody owns. If no one checks the output or notices when it breaks, automating it just hides a problem faster.
• check_circleMessy or scattered data. A high score on frequency does not help if the information the task needs lives in three systems that disagree with each other.
• check_circleA rare task with a high score everywhere else. A once-a-month task rarely pays back the cost of building and maintaining anything.
• check_circleA task that needs judgment every time, forced into a fixed process. That tends to raise risk instead of cutting it, whatever the other four questions say.

If you are not sure whether your business has the groundwork for any of this yet, the AI readiness assessment is a five-minute self-check. For the difference between this scoring exercise and a full engagement, read what an AI audit is. When you have a ranked list and want a second opinion on it, book a free call.

Frequently asked questions