CFO AI Optimism Hits 73% - But Graduate Hiring Is the Quiet Casualty

Sentiment Has Moved. The Question Is What It Means.

Deloitte surveyed 58 UK CFOs between 1 and 13 July 2026, the 76th edition of a quarterly tracker running since before generative AI was a board-level agenda item. The headline number: 73% of finance leaders now say their optimism about AI improving business performance has improved over the last 12 months. That compares with 59% in Q4 2025 and just 39% in Q3 2024. As CoinReporter put it bluntly, going from 39% to 73% in two years is "a genuine sentiment regime change, not a gradual warming."

Digital spending intent follows. A near-unanimous 96% of CFOs expect UK business investment in digital technology to rise over the next five years; 93% expect it within the next 12 months. Half of respondents said they expect productivity gains from AI deployment within the coming year.

For consulting and business services firms pitching AI transformation programmes, this looks like a green light. And to an extent, it is.

But.

The Graduate Hiring Numbers Are Uncomfortable

Buried in the same survey is something anyone running a graduate recruitment programme, or a firm that staffs them, should sit with. Cost control ranked as the single biggest dampener of graduate hiring for the year ahead, cited by a net 64% of CFOs. AI came second, cited by a net 47%. Outsourcing and offshoring came third, at net 33%.

That ordering matters. Cost control is the polite answer; AI is the structural one. These aren't separate forces. When a CFO says AI and cost pressure are both suppressing graduate intake, what they're describing is a workforce model quietly being rebuilt around different assumptions about how much junior human time is needed.

For recruitment agencies specialising in professional services, management consulting, or business support placement, this isn't an abstract future risk. It's a present-tense compression of the addressable market.

Personneltoday.com, which covered the graduate hiring finding specifically, noted that Deloitte's own CEO Darren Graves framed the AI optimism as encouraging for "productivity and business performance," without addressing the distributional question of who absorbs the displacement.

New Company Formation Tells a Different Story

The sentiment spike among large corporates sits in striking contrast to what's happening at the formation end of the market. According to AI Business Dispatch analysis of Companies House data, only 3,247 new SIC 70.22 (management consultancy) companies were registered in 2026-Q3, a fall of 74.6% against the prior period. UK Intellectual Property Office trademark data tells a similar story: Class 35 filings, the trademark class covering business services and consultancy, dropped to 3,526 in the same period, down 67.4%. A structural fact that doesn't shift quarter to quarter: 98.2% of active SIC 70.22 companies hold no Class 35 trademark at all. (AI Business Dispatch analysis of Companies House and IPO data, as of July 2026.)

That combination points to an accelerating bifurcation. Large, established consultancies with deep AI deployment programmes are pulling away, capturing CFO budget with credible productivity narratives. The long tail of micro-consultancies, many formed in the post-pandemic formation surge, are consolidating, closing, or being absorbed. The formation collapse isn't necessarily a bad-news story in isolation; it may reflect market rationalisation. But the trademark data adds nuance. Firms that aren't protecting their brand are often firms that aren't investing in their future.

The ICO Regulatory Clock Is Running

Consultancies and recruitment agencies running AI hiring tools face a separate, simultaneous pressure. The ICO's consultation on automated decision-making guidance closed on 29 May 2026, and final guidance is expected this summer. The underlying report, published 31 March, drew on engagement with more than 30 UK employers and found a consistent pattern: employers told the ICO their AI tools were decision-support only, with humans making final calls. The evidence suggested otherwise, with tools making substantive screening decisions and human review amounting to rubber-stamping.

The Data (Use and Access) Act 2025, in force from 5 February 2026, updated the legal framework. The near-total prohibition on significant automated decisions has been replaced by a permissive, safeguard-led regime. More flexible in principle; more exposed if firms haven't built the safeguards in. The ICO has already written to 16 named organisations, and enforcement focus on employment and recruitment use cases has been explicitly signalled.

For HR tech vendors selling AI screening tools into the UK, and for the consulting firms implementing them, the summer guidance drop is the one date that matters right now. Final guidance will require that any AI which filters, ranks, or rejects candidates be accompanied by a Data Protection Impact Assessment, meaningful human review with genuine override authority, and a clear candidate challenge mechanism.

Over 70% of organisations anticipate increasing AI and automation in recruitment over the next five years, according to the ICO's own public perceptions research. That ambition and the incoming compliance floor are on a collision course.

The Utilisation Problem Isn't Going Away

The Deloitte CFO data reads as a demand signal. CFOs are spending on AI, they expect returns, and they're redefining what headcount is needed to generate those returns. Yet 50% expecting productivity gains within 12 months, against only 78% expecting gains over five years, reveals the tension: short-term expectations are almost as high as long-term ones. That optimism curve will meet a reality check.

Management consultancies face a structural challenge AI doesn't resolve: utilisation. AI saves time on individual tasks without necessarily creating redeployable capacity blocks. As Dayshape's 2026 analysis noted, shaving hours off a fixed engagement doesn't free someone up for another client without deliberate resource planning. The firms that will capture CFO AI budgets are those that have already worked out how to convert fragmented AI time-savings into billable hours, not those still running unchanged service delivery models with a chatbot bolted on.

Sentiment is up. The work is harder than it looks.