Artificial intelligence is reaching a certain level of maturity and is beginning to raise questions. Heavy investments have been made, and there’s more to come, but at the same time, the businesses that have been promised mountains are struggling to see a return on investment in line with the efforts made. Worse still, some are reaching the end of their investment capacity. But should they put an end to their efforts? No, because that would be taking a risk worse than the financial one.
The questionable ROI of artificial intelligence
There’s no denying it, we’re witnessing a return to earth as far as AI is concerned, which in my opinion is an excellent thing. If we take the Gartner curve as a benchmark, we’ve moved away from over-inflated expectations, and we’re beginning to feel a hint of disappointment, which means we’re approaching consolidation and productivity. At least for existing and well-known forms of AI.
And given the speed at which things are moving, the speed at which the hype is ending gives us hope that productivity will arrive very soon.
Let’s make no mistake. I’m not saying that AI doesn’t work, just that it struggles to produce results commensurate with the investment made, and it’s the “commensurate” that counts.
When I was younger, it was explained to me that if I wanted to launch off-the-beaten-track initiatives , the smaller the I, the less they’d bother me with the R.
Here we’ve got a more-than-capital I, so we’re waiting for a gigantic R, almost visible from the moon, and we’re far from there. At least not today.
““The problem is that the current level of investment — in startups and by big companies — seems to be predicated on the idea that AI is going to get so much better, so fast, and be adopted so quickly that its impact on our lives and the economy is hard to comprehend. Mounting evidence suggests that won’t be the case.” .” (Is the AI Revolution Already Losing Steam?)
Diminishing investment capacity
On the face of it, nothing could be more normal. When you’re involved in disruptive, capital-intensive subjects, you often have to wait a long time to reap the rewards of your efforts, and the AI frenzy isn’t as old as all that.
Unfortunately, financial times are not the same as technological times, and businesses are beginning to find it difficult to finance the future of their AI projects, to such an extent that some are considering divesting elsewhere by selling off non-strategic assets to pay for AI investments (Companies look to sell off assets to pay for AI investments).
On the face of it, this is nothing to worry about, as it could be a good reason to clean up their portfolio and refocus on what’s vital for their future.
But this presupposes the conviction that AI will one day be a success, because it’s better to have a non-strategic asset that can pay off than a strategic asset that becomes unprofitable through risky investment.
What to expect from AI in terms of ROI?
The question then arises as to what we can expect from AI, and here the most diverse estimates coexist. Remember that “Forecasts are difficult, especially when they concern the future. (Pierre Dac).
If we instinctively understand potential and its logic, quantifying it is another matter , and speaking of technology, it’s an old story that keeps repeating itself (Figures and ROI : please admit when you have no clue).
There are three schools of thought: the optimists, the pessimists and the neutrals.
The optimists promise ROI in the short term (2-3 years) of 20-40% per annum, and between 100 and 200% in the long term (5-10 years).
This scenario is essentially based on infrastructure amortization and the learning curve.
The pessimists, on the other hand, are more neutral in the short term (-5 to +5%) and catastrophic in the long term, with a negative ROI of -20 to -30%.
This scenario can be explained by ever-increasing infrastructure costs and poorly conceived projects.
The truth often lies in the middle, so let’s take a look at what the “neutrals” and “reasonable” are saying.
They talk of a short-term ROI of 10-15% per year, and 50-100% in the long term.
A scenario justified by high integration costs, slow adoption by teams, or limitations in data quality (Figures and ROI : please admit when you have no clue).
What determines the ROI of AI?
While we don’t yet know enough to know how much AI will bring in, we are beginning to have enough hindsight to understand what influences its ROI.
Logically, there’s the maturity of both products and customers. Emerging technologies, as is often the case, offer new potential, but present high risks due to the many unknowns that accompany them at the outset.
Then there’s the business sector. Today, retail (personalization), finance (risk management) and logistics (optimization) seem to offer high ROIs, whereas healthcare (heavy regulations) and public administration (organizational complexity) are less propitious areas.
Then, as is often the case, there’s the size of the business: the bigger it is, the more it benefits from economies of scale and a fast learning curve.
Finally, there’s data quality. “Shit in, shit out”: poor data management can considerably reduce the benefits of AI.
Over-inflated expectations and hidden costs
The good news is that, unless you want to follow the Cassandras (and why not?), you can expect a positive ROI. But that’s not the point.
When it comes to investment, the question is not just “is it profitable”, but “when” and above all “how profitable”.
In other words, no one is interested in spending billions for a 5% ROI in 10 years’ time. At least not those holding the purse strings.
To meet these expectations, we’ll have to go beyond generative AI, which, while impressive at first, doesn’t produce results commensurate with its costs. A well-placed person in a large business recently told me that “given the cost of the query, AI won’t be able to be ‘open bar’ for everyone in the business”.
The chatGPT plus subscription at 20 euros per month is losing money. Worse still, Open AI can’t make money either with chatgpt Pro, which isn’t profitable (in the end, OpenAI is losing money on its ChatGPT-Pro subscription [FR]). The more it works, the more people use it, and the more people use it, the more it costs. And we’re talking about Open AI here, I’ll let you guess what it’s like for a business that develops its own models without benefiting from the same economies of scale.
“ AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do” (Goldman Sachs says the return on investment for AI might be disappointing)
So maybe with agentic AI? AGI? But if costs increase with capabilities, we’re entering a vicious circle from which it will be difficult to escape.
In short, everyone agrees that, while the promise is tempting, it’s not being kept.
“Speaking to 18,000 people in the audience, Eschenbach (Workday) stated that the current problem is that “for all the dollars that have been invested so far and all the buzz, we have yet to realize the full promise of what AI can deliver as business leaders”, not everyone is seeing return on investment from AI.” (Workday CEO: ‘For all the dollars that’s been invested so far, we have yet to realize the full promise of AI’).
And then there are the hidden costs that we don’t think about, but which businesses perceive in one way or another. They’re not included in the AI “budget”, but that doesn’t mean they don’t exist.
The main one is the environmental cost. Between energy consumption, the carbon footprint of data centers and the impact of hardware and materials, we’re reaching new heights. And one day
Knowing that most countries are unable to provide enough decarbonized energy for this, and that using carbon-based energy to power AI infrastructures exposes businesses to costs linked to emission quotas and environmental penalties, questions will quickly arise.
So, when we know all this, why will businesses continue to invest in AI? And, indeed, why should they?
The financier versus the entrepreneur, ROI vs RONI
Here we have an age-old debate that I often refer to as the financier versus the entrepreneur. The entrepreneur doesn’t always win, but given what’s at stake, I think he will this time, for better or for worse.
The financier will look at the ROI or Return on Investment and ask himself whether the game is worth the candle and whether it might not be better to invest the money elsewhere.
The entrepreneur, on the other hand, will think in terms of RONI. The risk of not investing.
The risk of not investing, here, is that if one day AI delivers on its promises, even if it’s only half or a third of the way through, those who didn’t get on the wagon will be wiped off the map. They won’t be in a bad way, they won’t have any trouble catching up, they’ll be dead in the water.
In the end, the question is simple: should we continue to invest, even at a loss, even for a long time, because one day it will work, or should we accept the prospect of disappearing if it does work?
A gamble. A simple gamble.
A gamble that everyone will make because the probability of it working is real and infinitely higher than the probability of it failing, even if that probability is incalculable.
A gamble that everyone will make because, while waiting for a potential failure to be demonstrated, in 10 or 20 years’ time, the shareholders, the markets and the employer will punish the person who refused to go for it.
Who will be right? We’ll know in 10 years.
And maybe even then, AI will be considered an unprofitable cost, but essential to a business’s survival. We’ve seen more surprising oddities.
Image: IA ROI by CoreDESIGN via Shutterstock.
