What does an AI feasibility study cost?
An AI feasibility study typically costs SEK 50,000–150,000 and takes 3–6 weeks. You get a scored list of possible use cases, a review of your data maturity, a risk analysis against GDPR and the AI Act, and a recommended pilot. The feasibility study is the cheapest insurance against betting millions on the wrong use case.
An AI feasibility study is the smallest investment you can make in AI – and the one with the biggest leverage. For SEK 50,000–150,000 and a few weeks of work, you get the answer to the question that decides everything that follows: which problem should we solve first, and is our data ready for it? Skipping that step is how companies end up in million-kronor projects that never should have started.
What the feasibility study costs, and what drives the price
| Scope | Typical cost | Timeframe |
|---|---|---|
| Focused study, one department | SEK 50,000–80,000 | 3–4 weeks |
| Broad study, multiple functions | SEK 80,000–120,000 | 4–5 weeks |
| Study with an in-depth technical data review | SEK 120,000–150,000 | 5–6 weeks |
The price is driven by how many parts of the business need to be mapped, how many interviews and workshops are required, and how deep the data review goes – from a high-level assessment to a technical review of actual systems and datasets.
The deliverables to demand
A feasibility study is only worth its price if it results in concrete material you can act on. Demand these four:
- A use case inventory with scoring. A list of possible use cases in your business, scored by business value, feasibility, and risk – not an inspiration catalog, but a ranking with reasoning you can question.
- A data review. Where is the data for the top use cases, in what condition, with what permissions? Data maturity is the biggest cost factor in AI projects, and this is where it becomes visible before it becomes expensive.
- A risk analysis against GDPR and the AI Act. Which use cases involve processing personal data? Does anything fall into a higher risk class under the AI Act? What does that require, if so? This sometimes drives the ranking more than the business value does.
- A recommended pilot. A justified proposal for a first step, with a sketch of scope, price tag, and measurable goals, so the next decision can be made directly from the feasibility study.
If you get a report full of generic AI musings and a sales pitch for the vendor’s platform, you haven’t bought a feasibility study, you’ve bought a brochure.
How the work actually happens
A typical feasibility study follows three phases. First, a mapping phase with interviews and workshops where processes, pain points, and ideas are gathered from the business – this is where your employees’ knowledge becomes the study’s raw material. Then an analysis phase where the use cases are scored, the data sources reviewed, and the risks classified. Finally, an alignment phase where the conclusions are presented to decision-makers and the recommendation is discussed, so the report becomes a decision made rather than a shelf-warmer. Your own time investment is a few hours per key person – small relative to what it steers.
The worked example: what does the wrong use case cost?
Here’s what the leverage looks like in numbers. A company has two candidates: an AI solution to automate contract interpretation and an assistant for internal documentation. Without a feasibility study, they pick contract interpretation – it sounds most valuable. After eight months and SEK 1.5 million, it turns out the contracts exist as scanned images of varying quality, half the information is in email attachments, and the margin of error in the interpretation requires manual review of every contract anyway. The project is shut down.
A feasibility study for SEK 100,000 would have caught this in two weeks: the data review would have shown the state of the contracts, and the scoring would have ranked the internal documentation higher – less glamorous, but buildable on existing data with measurable value. The feasibility study costs 5–10 percent of a misdirected project. That’s the whole calculation.
The right place in the AI staircase
The feasibility study is the first step in an investment staircase where each step buys knowledge for the next: feasibility study, PoC, pilot, and production rollout. Companies that follow the staircase rarely bet millions on the wrong thing, because the mistakes get caught while they still cost tens of thousands of kronor.
At Weapp we run feasibility studies with exactly the four deliverables above, and sometimes recommend the conclusion “wait on AI, start with the data instead.” Want to know what a feasibility study would cover at your company? Get in touch and we’ll have an initial conversation.
Frequently asked questions
Do we need a feasibility study if we already know what we want to build?
If the use case is well defined, the data is known, and the business value has been calculated, you can often go straight to a PoC. The feasibility study is for the stage before that: when AI feels urgent but nobody can pin down which problem to solve first, or when several ideas are competing for the same budget.
What's the difference between an AI feasibility study and an AI strategy?
The feasibility study is concrete and short: it scores specific use cases, reviews your data, and recommends a first step. A strategy is broader and covers organization, skills, and long-term direction. For most companies, the feasibility study is the right place to start – the strategy gets better once it's built on real experience.
Who should take part from our side in a feasibility study?
People who know the business's processes and pain points, someone who knows the systems landscape and the data, and a decision-maker who owns the budget for the next step. Expect a few workshops and interviews – the quality of the feasibility study depends more on your involvement than on the number of consulting hours.
What does the AI Act say about our planned solution?
The EU's AI Act divides AI systems into risk levels with different requirements. Most internal efficiency solutions fall into the low-risk category, but use cases involving employees, credit decisions, or biometrics can be classified higher and require more. A good feasibility study classifies your use cases and flags the requirements before you build.
Can the feasibility study be done in-house instead?
Parts of it, if you have people with both AI knowledge and the time. External support mainly adds two things: experience with what tends to work and what doesn't, and an independent assessment that isn't colored by internal preferences. A combination, internal process ownership with external methodology, tends to give the best result.