A few weeks ago, I shared some thoughts on digitisation projects and the gap between intention and execution. The response suggested many leaders are wrestling with a similar question behind closed doors: are we investing in outcomes, or are we investing in optics?
With my CFO hat on, I remain slightly skeptical about the return most organisations are currently achieving from their AI spend.
That’s not to say the potential isn’t real. Some studies suggest well-executed AI initiatives can deliver around $3.50 for every $1 invested, with top performers reporting significantly higher returns. But the broader picture is more nuanced. Adoption is high, yet the vast majority of organisations still struggle to move beyond pilots and proofs of concept. Many initiatives stall. Others quietly lose momentum once the initial excitement fades.
Deloitte’s 2025 executive survey reinforces this. Most respondents reported that a typical AI use case takes between two and four years to generate a satisfactory ROI. Only 6% achieved payback in under 12 months. That is a very different profile to traditional technology investments, which are usually expected to show value within 7–12 months.
That disconnect matters because AI is still often being sold internally as if it will deliver near-immediate impact.
We also need to be honest about why organisations rush into new technologies in the first place.
Sometimes it’s because leaders genuinely enjoy innovation and want to stay ahead of the curve. Often, it’s because they’re worried about being left behind.
Right now, we are firmly in the “talk it up” phase of the AI cycle. Vendors, consultants, boards and leadership teams are all reinforcing the narrative that you must be doing something with AI or you’re falling behind. That pressure leads to budgets being approved, tools being deployed and pilots being launched before anyone has clearly defined what success actually looks like.
A better question upfront is still the simplest one: what tangible benefit does this deliver to the organisation?
There are also broader market signals worth paying attention to. Two major US indices tracking leading AI and robotics companies, including mature businesses such as Microsoft, Alphabet and Apple, are currently trading at around 28 times earnings. That doesn’t mean AI is a bubble. But it does suggest expectations are extremely high. Markets are pricing in future performance that most organisations have not yet experienced operationally.
Then there are the practical risks that deserve more airtime.
We have already seen high-profile examples of professional services firms being exposed for using AI tools to generate client work without transparency. That’s not a technology failure. That’s a governance and ethics failure, and reputational damage tends to arrive far faster than productivity benefits.
From a finance perspective, data governance is another area that should make leaders pause. Confidential financial data, forecasts, commercial models and client information are increasingly being fed into tools and platforms that many organisations don’t fully control. This week, WhatsApp (owned by Meta) was sued by an international group of plaintiffs, including South Africans, over allegations that private messages may not be as inaccessible as users have been led to believe. Meta strongly denies the claims. But whether the lawsuit succeeds or not is almost secondary. The real issue is that trust, transparency and control over data remain unresolved in much of the digital ecosystem.
None of this is an argument against AI.
It’s an argument for discipline.
The organisations that are genuinely seeing value are not chasing every new tool. They are selective about use cases. They invest in data foundations. They treat AI as an organisational change programme, not a technology experiment. They apply proper governance. And they measure outcomes honestly, even when the answers are uncomfortable.
AI should be treated like any other capital allocation decision: clear problem definition, clear expected value, realistic timelines, strong governance and accountability for outcomes, not activity.
If you’re a CFO looking at technology budgets, are you asking the tough questions about ROI? If you’re a CEO approving major technology investments and you can’t confidently tell whether it’s a true investment or a vanity project, maybe it’s time to bring in an external set of eyes.
Not too slow progress. But to make sure it actually delivers.
This article is adapted from Rowan De Klerk’s monthly LinkedIn newsletter, where he shares strategic perspectives for business leaders.
Subscribe here to stay ahead: Subscribe on LinkedIn