The 3-5-4 Method

Find Your First AI Use Case
in 45 Minutes.

Three signs of friction. Five questions that map the workflow. Four decisions about what to automate. You are describing your own job, and that is the whole skill.

Get the Worksheet → See the Method

Why It Works

Most of Us Cannot See Our Own Repetition

Because we are inside it. Writing it down is what makes it visible.

62%

of the average working day goes on repetitive tasks.

ProcessMaker, 2024

45

minutes, start to finish. Five minutes to find the friction, twenty to map it, fifteen to decide, five to name the build.

0

technical knowledge required. You will not be asked to build anything. You are the only person who knows how your work really flows.

A Real Example

"Everyone here assumed this was simply the cost of doing business."

A manufacturing team, a task everybody hated, and nobody had ever written it down. Five questions found otherwise. The step that ate the hours was not the difficult part of the job. It was the copying.

The Method

Three. Five. Four.

Pick one repeating piece of your work, not the whole job. "Putting a client pack together" is the right size. "Looking after clients" is not.

Step 1

Find the Friction

3 signs 5 minutes

Think of one task you do again and again. Tick the signs it shows. Two out of three and it is worth going further.

"We do this every single day"

Same pattern each time, little variation.

"This takes forever but it's simple"

Clear rules, but it eats hours.

"We're copying from here to there"

Moving the same information between systems.

Write it down. Take 60 seconds now and name the one task. It becomes your working example the whole way through.

Step 2

Map the Workflow

5 questions 20 minutes

Most teams have never written this down. The writing down is where the answer appears.

Trigger

What starts it?

Something always does, even if it is you remembering.

Steps

What happens, in order?

Open this, copy that, paste there, save, send. Ten to fifteen lines is normal. The boring detail is the valuable bit.

Handoffs

Where does it pass to someone else, or another system?

System to system counts, and it is usually the bigger number.

Time

How long does each step take?

Per step, not the total. Measure it, do not estimate.

Pain point

Where does it stick, or go wrong?

What makes people groan? Do not be polite about it.

Step 3

Decide What to Automate

4 decisions 15 minutes

Go back through your steps. Put each one into a box.

1

Must stay human

Judgement, exceptions, relationships. Leave these alone.

2

Pure data transformation

Information in, clear rules, information out.

This box is the one that matters.

3

Just waiting time

Waiting on a reply, an approval, someone else. Pure waste.

4

The 80% case

The standard version. Automate that, handle the odd one yourself. Perfection is the enemy of progress.

Step 4

Your First Build

5 minutes

One problem. One pilot. One measurable outcome.

Look at box 2. The step that is pure data transformation, happens often, and annoys you most.

The task

What you want to build first.

The outcome

What good looks like when it works.

The baseline

How many minutes it takes today, and how many times a week.

Bring this to your build session. You describe what you want, we build it together.

The Objections

Three Things People Ask

What if my job is mostly judgement?

Then most of it belongs in box 1, and that is a good answer. Almost every job has admin wrapped around the judgement. The wrapper is what we are after; the judgement is what we protect.

What if I pick the wrong task?

You cannot really. It takes 45 minutes and you learn how your own work flows either way. If it turns out to be a poor candidate, that shows up fast and you pick another.

Is there a deadline?

No. Everyone moves at their own pace. If you get stuck halfway, bring the half you have done.

Free Worksheet

The 3-5-4 Method: Find Your First AI Use Case in 45 Minutes

The printable process-mapping worksheet we use in build sessions. Three signs, five questions, four decisions, and the one-page plan you bring to the table. Write in it, scribble on it, it is yours.

Download the worksheet

Three signs, five questions, four decisions. One page, no jargon.

Then We Build It With You

The worksheet is not the deliverable. The build is.

You bring the map. We sit down together and build the first thing on it, in your own tools, against your own work. No demos, no sandbox, no pilot that quietly gets abandoned. One problem, one pilot, one measurable outcome, and then the next one.

Book a Build Session →
Open the interactive tool →

Work the method in your browser, at 354.igniteaisolutions.co.uk

Common questions

What is the 3-5-4 Method?

The 3-5-4 Method is a 45-minute exercise for finding your first AI use case without buying anything first. It has three steps, named for the number of things you do at each: 3 signs that tell you a task is worth automating, 5 questions that map how the work actually flows, and 4 boxes you sort each step into. The output is one specific, evidenced candidate task rather than a strategy document.

How long does the 3-5-4 Method take?

Forty-five minutes in total, split across the three steps: 5 minutes finding the friction, 20 minutes mapping the workflow, and 15 minutes deciding what to automate. Most teams have never written a workflow down, and the writing down is where the answer usually appears.

What if my job is mostly judgement?

Then most of it belongs in box 1, and that is a good answer. Almost every job has admin wrapped around the judgement. The wrapper is what we are after; the judgement is what we protect.

What if I pick the wrong task?

You cannot really. It takes 45 minutes and you learn how your own work flows either way. If it turns out to be a poor candidate, that shows up fast and you pick another.

What happens after the 45 minutes?

You have a mapped workflow, a timed set of steps and one candidate task sorted into the box that matters. From there you either build it yourself or bring it to us and we build it with you. There is no deadline and no obligation; if you get stuck halfway, bring the half you have done.

Chris Duffy, founder of Ignite AI Solutions

About the author

Chris Duffy Founder and Chief AI Officer, Ignite AI Solutions

Chris is the founder of Ignite AI Solutions and a Certified Chief AI Officer. He's a Claude Partner, was named a Top 20 AI Leader in 2026, is a regular expert source for Forbes on AI and cybersecurity, a Responsible AI Awards nominee, and nominated for 3 awards at the National AI Awards 2026. Ex UK Special Forces, he now helps UK SMEs adopt AI safely.

More about Chris · Chris in Forbes · LinkedIn · Substack

Chat on WhatsApp