Three questions your organisation should answer before approving an AI budget

Three questions before the budget moves

Most enterprise organisations heading into an AI investment can speak fluently about the technology. Fewer can confidently answer three questions about the decision itself. These are not trick questions. They are not complex. But the inability to answer them clearly is one of the most reliable indicators that the investment is not ready to be made.

There is a secondary cost to this that receives insufficient attention. Organisations that commit to AI without answering these questions do not just risk a failed implementation. They risk landing on the least ambitious version of the opportunity. When the problem is unbounded, accountability is diffuse, and governance is an afterthought, the organisation defaults to what is easiest to justify and simplest to govern. Cost reduction. Headcount efficiency. Incremental automation. These are not wrong outcomes. But they are not what AI can deliver to a business that is genuinely ready to use it. The organisations that use AI to reshape how they compete, serve clients, and make decisions at speed are the ones that did the governance work before the investment was made, not after.

The first: what specific, bounded problem are we solving, and how will we know when it is solved?

Not a capability. Not a use case category. A specific problem, with a definition of done that is concrete enough to evaluate and narrow enough to govern. AI investment decisions that begin with broad capability statements — we want to improve customer experience, we want to accelerate decision-making — are not decisions. They are hypotheses dressed as commitments. The technology budget should not move until the problem statement is narrow enough to be governed and specific enough to fail against.

The second: who is the decision owner, and what is their actual authority?

Not the steering committee. Not the executive sponsor by title. The individual or body that will make the binding call to proceed, to pause, or to stop, and will be accountable for that call in twelve months. If the honest answer is that no single person holds that authority, the governance work is not complete. Diffuse accountability at the decision point becomes diffuse accountability at every subsequent point.

The third: what are the governance requirements for production, and is the organisation ready to meet them?

Regulated sector organisations in financial services, insurance, government, and healthcare carry obligations that shape what AI can do and how it must be governed in production. Not all of those requirements are fully settled. The ones that are known, however, should be understood before the investment is made, not after the pilot concludes. Microsoft's Responsible AI Standard, the Australian Government's AI Ethics Framework, and the emerging ISO/IEC 42001 AI Management System standard converge on the same principle: governance must be designed before deployment, not retrofitted after it.

Three questions. If all three can be answered clearly, the organisation is not just ready to commit. It is positioned to use AI at full power, not just minimum viable justification. If they cannot be answered, the investment that needs to be made first is in the decision, not the technology.

The AI Decision Acceleration Framework helps organisations work through these questions before committing to a budget, in a fixed-scope, artifact-driven engagement designed for the decision layer. April bookings are open at https://calendly.com/jradvisory/initial-discussion

Is your organisation approaching AI as a cost reduction exercise or as a genuine capability shift? Share your experience in the comments.

If someone in your network is about to approve an AI budget without a clear answer to all three, pass this on before the decision is made.

Sources: Microsoft. Responsible AI Standard. https://blogs.microsoft.com/on-the-issues/2022/06/21/microsofts-framework-for-building-ai-systems-responsibly/ Australian Government. Australia's AI Ethics Framework. https://www.industry.gov.au/publications/australias-artificial-intelligence-ethics-framework ISO/IEC. ISO/IEC 42001 AI Management System Standard. https://www.iso.org/standard/81230.html