AI strategy
We find out where AI pays off in your business, in what order the work should be done and what it costs. The result is a plan with numbers on it, not a document that has to be interpreted.
Short answer
An AI strategy is the answer to three questions: where AI pays off in your business, in what order the work should be done, and what it takes to keep it running in production. It is built on a review of your actual workflows and systems, and it lands in a prioritised plan with indicative prices and a timeline.
Shorter than you think, and more concrete.
An AI strategy is not a 60-page document. It is the answer to three questions: where does AI pay off in your business, in what order, and what does it take to keep it running in production.
Those answers cannot be lifted from an industry report. They sit in your own workflows: what actually takes time, which systems the records live in, who does what when a case deviates from the normal. That is why the work starts with you and not in a library of use cases from other companies.
The result has to be something you can decide on. A prioritised list where every candidate has an indicative price, a time estimate and a reason why the one at the top is at the top. If the conclusion is that you should not do anything with AI right now, we say that too, and we write down why.
Leadership creates direction. The team makes the solution useful. A strategy needs both.
01
Leadership needs to name an owner, choose which problem to solve first and set boundaries for data, security and risk. Without that, AI remains something individual employees try alongside the real work.
02
The people doing the work need to take part when the workflow is mapped and the pilot is tested. They know the exceptions, can see whether the solution saves time and notice when it adds a new manual step instead of removing an old one.
03
Both perspectives meet in a real workflow with a named owner, a clear measure and an operations plan. The pilot should inform the next decision, not merely prove that the technology works.
04
Once the pilot holds up, the team gets practical training and a few internal people take responsibility for helping colleagues and collecting improvements. The new way of working can then spread without every question returning to the consultants.
Four decisions made too early, or not at all.
01
A tool gets bought or a model gets picked, and then someone goes looking for a problem that fits. The order should be the other way around: first which hours go to the wrong things, then which technology is reasonable for exactly those hours. It makes for duller projects and considerably more that survive.
02
Which record applies when three systems say different things is a business question, not a technical one. It needs a name attached to the answer before the build starts. Otherwise the solution rests on material nobody stands behind, and the errors surface only when someone has been invoiced wrong.
03
A pilot with no answer to who gets alerted when it stops working is a demo with a longer shelf life. Who is responsible for it in production, what it costs per month and what happens when one of your systems is updated belong in the plan, not in a discussion after delivery.
04
Five parallel initiatives give you five half-finished ones. One workflow in production gives you something that actually gets used and a basis for the next decision. Prioritisation is the hardest part of a strategy and the part most often skipped, because it means saying no to four things that all sounded good.
Four steps, normally done in one to four weeks depending on scope.
01
Half a day together with you. We go through your workflows, your systems and where the information travels between them. We would rather talk to the people who do the work than only to management, because that is where the exceptions live, and the exceptions decide what can be automated.
02
We put numbers on time and cost per workflow: how often it happens, how long each case takes, how many systems are involved and how large a share are exceptions. Without numbers, prioritisation becomes a matter of taste, and then the proposal from whoever spoke loudest always wins.
03
What you should do first, what comes after that and what you should not do at all. The last column is often the most valuable, since it saves money immediately. Every candidate is ranked by effect over effort, with the reasoning written so that someone who was not in the room can follow it.
04
An indicative price and a timeline per initiative, plus what each one requires from you in time, access and decisions. You should be able to take the material into a board meeting and decide without us in the room, and know roughly what the next invoice will say.
Deliverables, not a meeting that ends with us getting back to you. The mapping gives you the first two, the full strategy gives you all four.
01
Your workflows and systems drawn as they actually work today, including the steps where someone moves information by hand between two systems. That map tends to be useful in itself, whether or not you do anything with AI.
02
The candidates ranked by effect over effort. Each with an indicative price, a time estimate, which systems it touches and what it requires from you. The things that are unsuitable to automate are on the list too, marked as exactly that.
03
Where your records live, what quality they hold, what may be sent to a language model and what may not. Plus which risk class each proposed initiative falls into under the EU AI Act.
04
Internal guidelines for what employees may use AI for and how, written so they can be decided on in a management meeting and then actually followed. A couple of pages, not a framework.
Riktpriser
Two scopes. If you move on to a pilot within three months, the price of the mapping is credited against the project, so the first step costs you nothing extra if you build on it.
Indicative prices excl. VAT. Your project is always quoted at a fixed price before we start.
The mapping is enough for most companies that want to get going. You find out which workflows are worth automating, in what order and what they cost, and you can go straight to a pilot. It is the most common first step at a company with up to about fifty employees.
The full strategy is for those of you who need more than a prioritisation: a review of where the data sits and what quality it holds, security and policy material you can show a board or a customer who asks, and a roadmap covering a year rather than a quarter.
If you are unsure, we start with the mapping. It can be extended into a full strategy afterwards, and you then pay only the difference.
Three things included in the strategy work instead of becoming a separate project afterwards.
01
We go through which personal data appears in the workflows we look at, what it may be used for and what that means for every candidate on the list. A proposal that cannot be carried out under GDPR should fall away in the planning, not halfway into the build.
02
We classify every proposed initiative and document the assessment in writing. Most automations of internal administration land in low risk, but the assessment should be made and written down, not assumed. It is also the material you need the day a customer or an auditor asks.
03
You get a draft set of internal guidelines: which tools are approved, what must never be pasted into a public AI tool and who to ask when you are unsure. It is usually the fastest risk reduction a company can make, and it costs nothing to introduce.
Caesar Katende works on agent architecture and orchestration: how a solution is broken up, which tools the agent is allowed to use, how it is evaluated and where the human checkpoint belongs. Ted Wachtmeister works on backend, data and ML as well as operations: integrations, data pipelines, error handling and the parts that have to hold when volume goes up. Both studied at Lund University.
The same people all the way: the person who draws the strategy writes the code and runs the training. Nothing gets lost in a handover.
In practice that means the mapping is done by people who know what every proposal costs to build. A plan that looks good on paper but takes three months to implement is not a plan, and the difference is hard to see for someone who has never built anything themselves.
Common questions
Not in the sense of a thick document. But you need to know which workflow to automate first and why, otherwise the first thing you happen to try is also the thing you get stuck in. For a company with ten employees the work can be done in a week, and then the mapping is plenty.
An automation mapping takes 1-2 weeks from the first meeting, a full strategy 2-4 weeks. What sets the pace is how quickly we get to talk to the right people at your company and how easy it is to get numbers on volumes, not how fast we write.
Partly, yes. You know best where the time goes, and you can write that list without us. What tends to be missing is two things: what each candidate actually costs to build and run, and which candidates are technically unsuitable even though they sound simple. If you want to start on your own, the AI Check is a simplified version, free and done in five minutes.
A flow map of the current state and a prioritised list where every candidate has an indicative price, a time estimate and the systems it touches. The full strategy adds data and security material, a risk classification under the EU AI Act and a draft AI policy. All of it in writing, and walked through in a meeting.
So that the proposals are buildable. We price every initiative on what it costs to actually build and run, since we are the ones who would do it. You get fewer sweeping recommendations and more concrete prices, and you find out which ideas fall down on what your data looks like.