There is no single right number, but there is a right shape. A realistic AI transformation budget is weighted towards people, not software: training, internal time and change management usually cost more than the licences themselves. If your draft budget is 80% tools and 20% people, it is probably upside down.
The five lines every AI budget needs
1. Licences and tools
General AI assistants (ChatGPT, Claude, Copilot, Gemini) are typically priced per user per month, with business plans costing more than personal ones because of data protection terms. AI features in software you already use may be included or sold as add-ons. Check current pricing directly with vendors, as it changes often.
Budget principle: pay for seats people will use weekly. Start with the pilot group and expand as usage is proven.
2. Training
This is the line most often under-funded. A licence without training is mostly wasted. Budget for hands-on training for the people who will use AI daily, and deeper training for the managers who will lead adoption. External courses, internal workshops and protected practice time all belong here.
3. Integration and configuration
Connecting AI to your documents, CRM or helpdesk, setting up approved workspaces, and configuring permissions. For many organisations this is light in year one, because off-the-shelf assistants need little integration. It grows once you automate workflows across systems.
4. Governance and risk
Writing an AI use policy, reviewing data protection, checking vendor terms, and handling UK GDPR questions. Often this is internal time from legal, compliance or IT, but it is real cost and should be visible.
5. Internal time
The hidden giant. Pilot users spending time learning, champions supporting colleagues, a manager owning the programme. It rarely appears on an invoice, which is exactly why it gets ignored, and why projects run out of steam.
How the budget shifts by stage
| Stage | Where most money goes | What to hold back |
|---|---|---|
| Pilot (first 3 months) | Training the pilot group, a few licences, the owner's time | Enterprise contracts, custom builds |
| Expansion (months 3 to 12) | More licences, wider training, champions' time, light integration | Tools without a proven use case |
| Scale (year 2 onwards) | Integration, process redesign, governance, ongoing training | Nothing by default; fund what the data supports |
Three budgeting mistakes to avoid
Budgeting by headcount instead of use case. "Copilot for all 400 staff" sounds bold, but if 300 open it twice, you have paid for 300 unused seats. Budget for the teams with proven use cases and expand from there.
Treating training as a one-off. Tools change monthly. A small recurring budget for refreshers and new use cases keeps value growing.
Forgetting to budget for measurement. Someone needs time to track time saved, quality and adoption. Without it, you cannot justify next year's budget.
How to present the budget to decision-makers
Frame the budget in two parts: a small, fixed pilot investment with a clear end date and success criteria, and a larger expansion budget that is released only if the pilot hits its targets. Decision-makers approve far more readily when the downside is capped. Our guide on how to build a business case for AI walks through the numbers in detail, and hidden costs of AI implementation lists the items people forget.
Investing in the person who leads it
The highest-return line in many AI budgets is training the manager who will lead the rollout, because their decisions shape every other line. Our Leading AI Innovation & Organisational Transformation course costs £720 live online and finishes with a 90-day rollout plan you can cost and present.
Frequently asked questions
What percentage of an AI budget should go on training?
There is no fixed rule, but if training is a small fraction of licence spend, adoption is likely to suffer. Many successful rollouts spend as much on people as on tools in the first year.
Should small companies budget differently?
The shape is similar, but the absolute numbers are much smaller. A small firm can often run a meaningful pilot with a few licences and focused training.
When should we budget for custom AI development?
Only after off-the-shelf tools have been properly tried and a specific need remains that they cannot meet. Most organisations get significant value long before that point.