The APO method · AI Process Optimisation
First the process, then the AI. This is how the APO method works
APO maps which process steps AI can take over, what that delivers per year in hours and euros, and what it costs to build. Together with the people who do the work, using their own numbers. Not a presentation about what AI could do, but a calculation of what it delivers for you.
- Three numbers per process step
- 24 building blocks with fixed hour estimates
- The expected scenario leads
- InThe work as it runs todayProcess map in measurable steps
- HumanThree numbers per stepDuration, headcount, frequency
- AIAI fit and role per stepSeven questions, four roles
- AIBuild layer and hoursYourselves with AI, or NinA builds
- DoneBusiness case and roadmapExpected scenario, three phases
From gut feeling to numbers
01 / Why a method
Four things hold organizations back.
APO tackles all four in one working session, with the people who do the work at the table.
They don't know where to start
“Everyone has an opinion about AI. Nobody has a list.”
They can't back it up
“No number, no budget. No budget, no project.”
They overestimate and underestimate at the same time
“That's a quick bit of automation, we said. Six months later we were still building.”
They skip the preconditions
“The model wasn't the problem. The data simply wasn't there.”
02 / What it is
What is the APO method?
APO stands for AI Process Optimisation, the method of NinA AI. Together with the team, we map the most important processes in measurable steps. For each step we decide whether AI can take it over, what that realistically saves and what it costs to build. At the end there's a prioritised list, a well-founded business case and a plan in three phases.
The heart of the method is one simple rule: every step gets three numbers. How long does it take, with how many people, how often a month. "That takes a lot of time" becomes: forty minutes, two people, eighty times a month. That's 1,280 hours a year, and at a rate of € 45 per hour, € 57,600 worth of work. Only then comes the question of which part AI can take over.
The method used to be called APR, AI Process Redesign. We renamed it on purpose. Redesign suggests turning everything upside down. Optimisation describes what really happens: existing processes get better step by step, and AI only takes over what it can demonstrably do better or cheaper.
01Calculates live
Every input feeds straight into the business case. You see your own figures take shape, instead of getting a report afterwards.
02Separates doing it yourselves from having it built
Not every opportunity is a project. Your team handles part of it themselves with prompts and assistants.
03Honest about uncertainty
Opportunities without data or examples get a lower confidence. We advise on the expected scenario.
04Yours, not ours
The team names the opportunities, scores its own maturity and votes on priorities. That's what creates buy-in.
03 / Principles
Seven principles we never let go of.
01
Process first, technology later
We start with the work, not with AI. Reverse the order and you get solutions without a problem.
02
Everything in time and money
Duration, headcount, frequency. Without those three numbers, an opportunity doesn't exist in the model.
03
Make uncertainty visible
A range instead of a single number, a brake when data is missing, and the expected scenario as the guide.
04
Do it yourselves where you can
The bottom two build layers are yours. If you have everything built, you learn nothing and stay dependent.
05
The human stays in charge somewhere
Where creativity, physical presence or liability is involved, a person decides. Fully autonomous AI is the exception.
06
You talk, we steer
Asking over telling. The opportunities have to belong to the participants, otherwise there's no buy-in.
07
Buy-in counts
A quick win the team backs, with energy behind it, beats a bigger but heavier project.
In one sentence
The method in one sentence
APO maps which processes AI can take over, what that delivers per year and what it costs to build.
04 / Ten blocks
From welcome to plan, in ten blocks.
This is how a full session runs. In the AI Quick Scan we do this in one day at your site, with the people who do the work.
| Why we do it | What it delivers | |
|---|---|---|
| 1. Welcome and plan for the day | Everyone knows what's coming and why we do it together. | A team that takes an active part |
| 2. Intake and frameworks | So that everything we calculate later is right, and we know what is and isn't allowed with AI. | Context, financial framework, AI policy |
| 3. AI maturity | An honest picture of where you are and where you want to go. | Baseline on seven dimensions |
| 4. Collecting opportunities | A wall full of ideas. Anything goes, then we choose. | Top opportunities, voted on |
| 5. Process map | We map how the work really runs today. | Processes in measurable steps |
| 6. From manual to AI | Per step: what does AI take over, where does a person keep a check? | An AI version of each process, with the saving |
| 7. Classifying opportunities | Do you do it yourselves with AI, or does NinA build it? | Build layer and hour estimate per opportunity |
| 8. Prioritising | A lot of return for little effort wins. | A top 3 for the first phase |
| 9. Business case | What it delivers, what it costs and when it pays back. | Three scenarios, the expected one leads |
| 10. AI Future Plan | The plan in one place, the starting point of working together. | Roadmap in three phases |
05 / The baseline
Seven dimensions, now and in twelve months.
Strategy, data, technology, people, processes, governance and culture. Each on a scale from 1 (initial) to 5 (transformative), calibrated on behaviour you can see, not on gut feeling. The biggest jump between now and the goal decides where attention goes.
Profile · now vs. target
2.7
now
4.1
target after 12 months
Biggest jumpData readiness (2 to 4)
What we measure, on a scale of 1 to 5
Strategy and Vision+1 jump
How clear and shared is the AI vision at board level? Is the strategy written down and who holds the mandate?
now 3, target 4.Data readiness+2 jump
Is data centrally available, structured and accessible for automation and AI?
now 2, target 4.Technology and Infrastructure+1 jump
Which tools, cloud platforms and integrations are in place? How easily do you connect new systems?
now 3, target 4.People and Skills+2 jump
Does the team know AI tools, prompting and automation? Is there capacity and willingness to learn?
now 2, target 4.Processes and Adoption+1 jump
Are processes clearly described and actively optimised? How did earlier change programmes go?
now 3, target 4.Governance and Risk+2 jump
Is there policy for AI use, data ethics, privacy and security? How is risk weighed against speed?
now 2, target 4.Culture and Change+1 jump
How does the organisation deal with change? Is there room to experiment, fail and learn?
now 4, target 5.
06 / AI fit
Is this task suitable for AI? Take the check.
For each process step we ask seven questions, each answered with no, partly or yes. They're based on the work of Brynjolfsson and Mitchell on which tasks lend themselves to machine learning. Three more questions then decide whether the human stays in charge. Pick an example or fill in your own task.
Start with an example, or fill in your own task
AI fit: seven questions
01Is it clear what goes in and what should come out?
Can the person doing it explain: I get this, I deliver that?
02Is the information it needs digital?
In a system, a mailbox or a folder. Not in someone's head.
03Are there at least ten examples of good work?
Input and good output, digital. Shadowing a colleague doesn't count.
04Is it done in a few steps?
No long chain of judgements building on each other.
05Does it go roughly the same way every time?
A fixed pattern, not an exception every time.
06Is a mistake cheap to fix?
Yes: a mistake shows and is quickly corrected.
07Can the outcome do without a detailed justification?
No if it needs a legal or in-depth explanation.
Does the human stay in charge?
- Is original creativity the core?
- Does someone need to be physically present?
- Is a person personally accountable for the outcome?
Result
11 / 14
AI fit · Strong
This task suits AI well.
Role of AI
Human approves
AI does the work, a person checks before it goes out.
Confidence
79%
This is the first test of the APO method. In the AI Quick Scan we run it per process step, with the people who do the work.
07 / Four roles
AI isn't all or nothing.
The AI fit leads to one of four roles between human and AI. By far most opportunities land on "human approves" or "AI assists". Fully autonomous AI is the rarest role: only with a very high score and with no creativity, physical presence or liability involved.
Pick a task
Human
A person does the work. AI stays out.
AI assists
A person does it, AI drafts or looks things up.
Human approves
AI does the work, a person checks before it goes out.
Fully AI
AI does it by itself. A person spot-checks.
Where it lands: Human approves
AI prepares the answer with the right order details. An employee reads along and sends: tone and exceptions stay human.
08 / Four build layers
Yourselves with AI, or NinA builds.
Every opportunity gets a build layer. That decides who makes it and how much work it is. Your team does the bottom two itself, with training and the right prompts. We build the top two. We link every opportunity to one of 24 fixed building blocks with an hour estimate in three scenarios, so the estimate doesn't depend on who runs the session.
Layer 1 · yourselves with AI
Prompts and knowledge
The team does it themselves, with good prompts and training. Think of a prompt library for recurring writing work.
Layer 2 · yourselves with AI
Assistants and projects
Your own GPTs and Claude Projects with your knowledge inside, run by the team. For example an assistant that sorts email or writes copy in your house style.
No data or no examples? Then 40 percent confidence at most. However good the opportunity sounds.
That's the most important brake in the method. Low confidence isn't a rejection but an assignment: foundation first, then the build. And never 100 percent saving: per step we calculate with 90 percent at most, because exceptions, checks and management remain.
09 / Prioritising
What do we tackle first?
Every opportunity gets a place on two axes: expected saving per year against expected build hours. The default thresholds are € 15,000 a year for high impact and 60 hours for low effort, adjustable per client. We rank the list by ICE: impact times confidence times ease, each on a scale from 1 to 10. Because confidence is a multiplier, a missing precondition pulls the whole score down. That's deliberate.
| Impact | Effort | What we do | |
|---|---|---|---|
| Quick win | High | Low | Do it straight away, in the first phase |
| Strategic opportunity | High | High | Plan it, with sign-off on the business case |
| Filler | Low | Low | On the shelf, only if there's time left |
| Thankless | Low | High | Don't do it, or dig deeper first |
Swipe sideways to see all columns →
10 / The AI Future Plan
A plan in three phases.
The last step of the method. The plan is the starting point of working together, and in a partnership we update it every quarter.
- 01
Quick wins
0 to 3 months · layers 1 and 2
What the team picks up itself with prompts and assistants. Often results within a few weeks, without a build project.
- 02
Building
3 to 9 months · layer 3
Workflows that take over processes, connected to your systems. Live one by one, with a person keeping an eye on them.
- 03
Strategic
9 to 12 months and beyond · layer 4
Custom work: agents, search across your own documents, platforms. Only once the foundation is in place.
11 / Where you'll meet the method
In the Quick Scan, and every month after.
The method isn't a separate product. It's in everything we do.
- In the AI Quick Scan we run through it in one day at your site. Within a week you get the baseline, a business case per opportunity, the impact-effort matrix, a twelve-month roadmap and package advice. € 6,000 one-off, also available on its own.
- In the AI Partnership the Quick Scan model is the yardstick. Every month we show what's running and what it delivers, against the business case we made together.
- In a workshop we use parts of it: mapping processes, the AI fit check, the four roles.
Want to start on your own? Take the AI fit check above for one task from your own work.
“Don't let it drive you crazy. Start small.”
12 / Questions
Questions about the APO method
What does APO stand for?
APO stands for AI Process Optimisation, the method of NinA AI. You map processes in measurable steps and decide per step whether AI can take it over, what that saves and what it costs to build. Until August 2026 the method was called APR, AI Process Redesign.
How long does an APO session take?
A full session in ten blocks takes four to four and a half hours, including a break. In the AI Quick Scan we do this in one day at your site, and within a week the results are ready: baseline, business case per opportunity, roadmap and package advice.
Who needs to be in the session?
The people who do the work, and someone who can make decisions. The opportunities have to come from the participants. A session with only the leadership team produces a nice list, but little support on the work floor.
We already know where our problems are. What does the method add?
The numbers. What does it cost now, what does AI realistically deliver and what does it cost to build? Without those three, there's no decision to make. If you already know the problems, the session simply goes faster.
Our data is a mess. Is this still worth it?
Especially then. It comes up explicitly in the session. Opportunities without data or examples get 40 percent confidence at most, plus a concrete assignment: build a source or collect examples first. Better to know now than halfway through a project.
How do I know the savings are right?
They come from your own numbers: your times, frequencies and rate. We only fill in the build hours, from a fixed library of 24 building blocks. And we advise on the expected scenario, not the most favourable one. We also show the cautious scenario, as a floor.
We've already bought an AI tool. Does that fit in?
Yes. The method isn't about tools but about processes. An existing tool simply counts. The question is which steps that tool really takes over and what that delivers.
Can we apply the method ourselves?
Partly, yes. Pick one process, give each step three numbers, decide the role of AI per step with the AI fit check and work out what it delivers with your own hourly rate. What you miss without us: the fixed hour estimates for the build, someone to guide the conversation and the experience of what does and doesn't work.
Want to see what the method delivers for you?
In a fifteen-minute intro call we look at which process qualifies first, and whether the AI Quick Scan is the logical start.
01
Intro call
Fifteen minutes, no strings. Where does work get stuck and what do you want to achieve?
02
AI Quick Scan
One day on site. Within a week, a business case and a roadmap.
03
AI Partnership
A dedicated team that builds, trains and shows what it delivers every month.
160+ organizations · Sessions rated 9.3 · Based in Amsterdam