The AI budget conversation your firm needs before 1 July

That conversation feels productive. It generates decisions. But it's the wrong conversation, and by the time most firms realise it, they've committed a budget to a pilot that was always going to drift.

BCG's research shows 74 per cent of organisations struggle to achieve and scale AI value despite widespread adoption. Firms average more than four pilots, but fewer than one in four reach production with measurable returns. The constraint isn't technology. It's the question the budget conversation started with.

The myth that's costing firms real money

Myth: Choosing the right tool is the hard part. Once the firm selects a solid platform, adoption and value will follow.

Reality: The tool is rarely the constraint. What kills AI investment in accounting firms is the absence of a defined outcome before the licence is signed. A mid-tier firm I worked with approved an AI-assisted document review platform mid-year. By the following March, the same partners who approved it had stopped using it - not because it didn't work, but because no one had defined what "working" looked like for their practice. The tool did exactly what it was supposed to. The firm had just never decided what they needed it to do. Twelve months of licence fees, three weeks of onboarding, zero measurable return. Problems first, platforms second. Every time.

Three questions before the budget is approved

The most useful thing I can offer any firm heading into a new financial year budget cycle is a pre-approval discipline. Before any AI initiative receives funding, the proposing partner should be able to answer three questions clearly and specifically:

What is the sentence?

The initiative must be describable in a single, complete sentence: "This investment will produce [specific measurable outcome] by [specific date], and we will know because [specific indicator]." If the sentence can't be completed, the initiative isn't ready. Vague goals – "improve efficiency," "reduce admin burden," "explore AI capabilities” – are not outcomes. They're intentions. Intentions don't produce returns.

What is the real cost?

The licence fee is the smallest number on a properly constructed AI budget. The real investment includes data preparation, integration with existing practice management systems, change management – the structured work of helping people actually adopt the tool – AI literacy training that goes beyond product onboarding, and governance setup that defines who is accountable for what the system produces. In my experience, the total cost of a mid-sized firm AI implementation runs two to three times the platform fee when these elements are included honestly. Budget for what it actually costs, or budget for a pilot that will quietly fail.

What is on the stop list?

Before any new initiative is approved, every current AI initiative deserves a simple audit: Is it being used? By whom? Has it produced the outcome it was funded to produce? Any pilot running longer than six months without a defined path to production is not a technology problem. It is a leadership decision to avoid accountability. The stop list is not a failure report. It is the foundation of a serious budget – and the single most valuable conversation a firm can have before committing to new investment.

If an initiative can't pass all three, it doesn't belong in next year's budget. That's not a harsh standard. It's the standard that separates firms building genuine AI capability from firms accumulating expensive experiments.

What the new financial year actually demands

The firms I see building real capability aren't necessarily spending more. They're spending with more discipline - funding fewer initiatives, defining outcomes before platforms, and treating the people's side of adoption as the primary investment, not an afterthought.

AI is 20 per cent technology and 80 per cent people. The firms that understand that heading into the new financial year will be well ahead of those still treating it as a shopping problem.

The budget that earns trust from your partners and your clients isn't necessarily the biggest one. It's the one that knows what it's there to prove.

Andrew Cooke is founder and managing director of Growth & Profit Solutions AI.