Home / AI debt
The hidden AI debt
What does it cost you that AI is not yet used systematically?
Your AI debt is what it costs you not to use AI fully. For many organisations it is at least SEK 75,000 per knowledge worker a year in lost capacity and missed delivery. [1]
Understand your AI debt and how to start reducing it. The video is in Swedish.
How much money are you losing?
Work out what insufficient AI use is costing you
Enter the number of knowledge workers and choose how cautiously you want to calculate. The result updates instantly and can be saved as a briefing for your manager and leadership.
/ year
This is an estimated loss of productivity and capacity, not an IT cost. Even small losses of time per person add up to large sums over a year. That is why this is mainly a question for leadership.
Results, briefing and next steps
Your next steps
Judge how quickly you should act
The higher the AI debt, the more reason to have the first conversation now.
What the figure means for the business
The calculation makes visible a loss of productivity, capacity and delivery that otherwise easily stays hidden in everyday work.
Why the briefing is worth sharing
The result is designed to be passed on internally and give your manager or leadership a concrete basis for prioritising the next step.
Why a meeting is the natural next step
In a short review we show how to read the result, choose the right starting point and get going without making the change bigger than it needs to be.
The way forward
From AI debt to your own AI capability in five steps
The figure shows why you should act. Your AI journey shows how, without making it too big, difficult or technical. See the whole AI journey →
- A first conversationFree of charge. You get a first picture of where you stand.1
- Where you stand, in depthThe obstacles and opportunities in your organisation.2
- Your planWhat to do first, and why.3
- Support and trainingTraining for employees, advisory for leadership, or both.4
- You manage on your ownThe knowledge stays with you.5
The first conversation is free of charge. We show you where it is easiest for you to start.
Book a free reviewMethod and sources
How we calculated
What the research actually shows, which levels the model uses and what it takes to realise the value.
1. How should the three levels be read?
The levels are the same three as in the calculator and the briefing. Read them as three decision levels within the same model, from a cautious estimate to higher potential.
| Level | Per knowledge worker a year | How to read it |
|---|
Potential beyond the chosen level
The highlighted level shows which assumption is used in the calculator and in the briefing. It is meant to give a clear and comparable baseline for decisions.
At the same time, research suggests that greater value can be released when AI is used more systematically in ways of working, governance, coordination and implementation. Organisations that go further on their AI journey can often release additional potential.
2. What drives the different cost levels?
- Knowledge intensity: the more the work relies on analysis, synthesis and handling information, the greater the potential.
- The content of the work: analysis, writing, documentation, coordination and decision support are often affected more than purely transactional tasks.
- Degree of under-use: the broader and more uneven AI use is across roles, the greater the lost value.
- Implementation maturity: access to tools is not enough; actual adoption in workflows decides the outcome.
- Governance and ability to scale: governance, follow-up and clear ways of working decide whether the value is realised safely and at scale.
- The jagged technological frontier: the value is higher when the work lies within AI’s useful range and lower when the task falls outside that frontier.
- The economic weight of quality: where better quality and less rework matter a great deal, the higher potential level becomes more relevant.
3. What does the evidence show overall?
Taken together, the evidence shows that the value of AI does not mainly come from access to tools, but from how well AI is turned into real ways of working, priorities and leadership. Research supports that the effect increases as use becomes more integrated, more structured and more firmly anchored in the organisation.
That is also why the levels in the model should be read as decision levels, not as an absolute ceiling for potential. The chosen level gives a reasonable and useful basis for investment. Organisations that move from a first activation to more systematic implementation can often release further value.
For decision-makers, this means the question is not only whether AI should be used, but how quickly the business can move from limited use to a way of working that strengthens profitability, reduces dependence on consultants and builds internal capacity over time.
4. Exact sources
- Dell'Acqua, F., McFowland III, E., Mollick, E., Lifshitz-Assaf, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F. and Lakhani, K. R. (2023). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality. Working Paper 24-013. Harvard Business School.
- Forrester Consulting (2025). The Total Economic Impact™ Of Microsoft 365 Copilot: Cost Savings And Business Benefits Enabled By Microsoft 365 Copilot. March 2025. Forrester Research, Inc.
- Gartner (2024). Gartner Identifies Four Emerging Challenges to Delivering Value from AI Safely and at Scale. Press release, 21 October.
- Gartner (2025). Gartner Survey Shows Supply Chain GenAI Productivity Gains at Individual Level, While Creating New Complications for Organizations. Press release, 5 February.
Next step
Start paying off your AI debt
Book a review and see which workflows to start with to save SEK 75,000 per knowledge worker and build your own AI capability.