Disclaimer: This article is based on Atif's brilliant talk at our AI Powered Finance Summit, which you can watch ondemand on your dashboard.
If I could leave finance leaders with a single sentence, it would be this: finance must become the enterprise's judgment architecture for AI.
Not the automation department. Not the pilot factory. Not the function that reports last month a few hours faster, but the function that helps the enterprise decide next quarter better.
That sentence has a philosophy behind it, a set of practical design choices underneath it, a politics to navigate, and a question of institutional power at its core.
I want to work through all four in that order, because the order isn't decorative. It is the argument.
The world isn't short of people willing to tell you that AI will transform finance. It's rather shorter of people willing to show you how, with enough specificity to act on Monday morning.
Sabrinthia Donnelly
Where the research converges
When McKinsey, PwC, MIT, Oxford, Cambridge, and Harvard arrive at the same conclusion from different starting points, it's worth pausing to notice.
McKinsey's 2025 CFO survey found that 44% of CFOs were already using generative AI for more than five use cases, up from 7% the prior year. But McKinsey's counsel isn't to do more.
It's to prioritize ruthlessly, centralize oversight, and rewire workflows rather than chase diffuse efficiency wins. A 2026 study of 1,200 senior executives found that nearly three-quarters of AI's economic value is being captured by just one-fifth of organizations.
Those leaders are twice as likely to have redesigned their workflows rather than simply adding AI tools to existing processes.
MIT's research shows that AI produces its largest gains at the workflow level, not the isolated task level. Until workflow redesign reaches a certain threshold, the costs of adopting AI outweigh the gains.
Cambridge's 2026 survey found that 81% of financial services firms are adopting AI at some level, yet only 14% view it as transformational to their strategy.
The value picture is where it gets genuinely uncomfortable. Deloitte's most recent CFO Signals survey tells us that 80% of CFOs expect AI to be extremely or very important to finance in 2026, yet only 20% of active users report clear, measurable value.
MIT's Project NANDA found that 95% of enterprise generative AI initiatives deliver zero P&L impact. Its methodology has attracted some criticism, and I note that caveat. Meanwhile, KPMG reports that 92% of finance AI initiatives are meeting or exceeding ROI expectations.
If you think those two figures can't coexist, you're thinking about measurement correctly. The tension between them is itself a governance story about what we measure, and against which benchmark.
The CFO who governs the gap between adoption and integration, rather than merely managing it, captures the value everyone else is leaving on the table.
Teresa Garanhel
Philosophy: a problem of institutional design
I begin with philosophers rather than management frameworks, because this isn't, at its root, a technology problem.
It's a problem of institutional design, which rests on philosophical foundations whether we acknowledge them or not.
Hannah Arendt warned that under conditions of tyranny, it is far easier to act than to think. I'd submit that under conditions of vendor cycles, it is far easier to deploy than to think.
We ship the pilot and demonstrate the art of the possible, but almost nobody pauses to ask: possible for whom, and at what cost to judgment?
Peter Drucker put it more economically: there is nothing so useless as doing efficiently that which should not be done at all.
The real AI question for finance isn't how we automate one more task. It's how we preserve decision quality while moving the function from historical reporting to future-shaping stewardship.
None of that is achieved by bolting AI onto the old architecture.
For expert advice like this straight to your inbox, sign up for Pro+ membership.
You'll also get access to hundreds of hours of exclusive video content, a complimentary Summit ticket, and so much more.
So, what are you waiting for?
Daniel Kahneman observed that algorithms are reliable where patterns are consistent and feedback is immediate and clear.
Strategic finance is the opposite, full of ambiguity, competing objectives, and long feedback loops.
As Kahneman said, fast thinking is something that happens to you, while slow thinking is something you do.
An AI-native finance function must protect the institutional capacity for slow thinking in a world being engineered for speed.
Become an FA Insider
Thank you for subscribing
Get exclusive insights, frameworks, and strategies from finance leaders driving real business impact.
An email has been successfully sent to confirm your subscription.
Follow us on LinkedIn
