Technologies sell productivity, but businesses want revenue

The other day, I explained why AI has not delivered on its promises in terms of productivity, that this was not surprising and could be explained very rationally, but that we should not throw the baby out with the bathwater, because the technology itself was not to blame (The great illusion of technological productivity gains (including AI)).

One of the points I mentioned sparked some interest, namely the confusion between productivity and revenue. This led to a few contact requests from people who were really interested because it put into words a problem they were clearly struggling to articulate, and I thought it was worth coming back to the subject here in more detail.

The prevailing discourse on AI promises businesses huge productivity gains. However, a business does not pay its employees or shareholders with productivity. It needs revenue and cash flow. The challenge, therefore, is to transform productivity into income, which is far from automatic.

In short:

• Productivity is not synonymous with income: the productivity gains promised by AI do not automatically translate into revenue or cash flow.
• Measuring productivity in knowledge-based professions is complex and often biased, making it difficult to accurately assess the benefits of AI.
• Individual efficiency gains do not guarantee collective improvement if overall processes or skills do not follow suit.
• Even when productivity increases, it does not always generate more revenue, particularly in the case of a limited market or an unsuitable billing model.
• In the absence of direct economic benefits, productivity often becomes a pretext for cost cutting, with long-term negative effects on innovation and growth.

Productivity, an elusive indicator

Before even attempting to convert it into revenue, it must first be measured. And that’s where the problem lies. In knowledge work, it is impossible to time each action as in Taylor’s day. We don’t know where we’re starting from or where we’re going, yet we have to provide indicators that are, in many cases, logically biased or even fanciful (How can we measure the productivity gains of AI?).

Most studies are based on self-reported data: employees are asked if they feel more productive. But their perception is biased by the novelty effect, the desire to please management, or the fear of admitting that the tool provided is useless. And even when there are real gains, they are often offset by new tasks: checking, correcting, supervising, learning how to use the tool. We gain on one side but lose on the other (Are you familiar with the Jevons paradox at work? When efficiency leads to inefficiency.).

The sum of individual gains is not a collective gain

Let’s assume, nevertheless, that employees become more efficient. This does not mean that the organization becomes more productive. The sum of individual productivity does not equal collective productivity (Local optimum vs. global optimum and the theory of constraints: why your productivity gains sometimes serve no purpose).

For example, in response to a call for tenders, 80% of the team may be able to work 20% faster. But if final validation remains a bottleneck, the customer will not receive anything any sooner. In other cases, some employees do not have the necessary skills to use AI properly, which slows down the flow and cancels out the gains made by others. As a result, more intermediate deliverables are produced, but the final result does not progress.

Added to this is another well-known phenomenon: the gains are captured elsewhere. By software and infrastructure providers, who charge for their services. By employees, who choose to use the time saved to improve quality, train themselves, or go home earlier. In reality, every business does a mix of all of these things. The promised productivity evaporates, and its link to revenue becomes even more diffuse.

Faster does not mean more revenue

Let’s assume that the organization has successfully eliminated bottlenecks and that productivity has indeed improved overall. One final barrier remains: revenue.

I am surprised that many voices have been raised, and rightly so, to criticize the limitations of the MIT study which tells us that 95% of AI pilots failed (MIT report: 95% of generative AI pilots at companies are failing), particularly for reasons of sampling and methodology, but in my opinion its bottom line is based on a major misinterpretation that I have not seen mentioned anywhere. We are told that businesses are not seeing any growth in revenue, but we must remember that the promise of technology, and AI is no exception, is productivity, not revenue.

However, revenue does not depend directly, automatically, and linearly on productivity. Put another way, and more mathematically, revenue is not a function of productivity.

So when you buy productivity, it may be a shame not to see an increase in income, but it is neither abnormal nor surprising.

If you sell a packaged service, such as design or logo production, delivering three times faster makes no difference if sales don’t increase. You’ll just have employees twiddling their thumbs, but at least you won’t lose revenue.

If you bill by the day, like freelancers or IT services companies, producing in eight days what used to take ten days means billing for two days less. Unless you immediately have a new customer, your income will decrease and, at a constant level of activity, your annual turnover will decrease in proportion to your increase in productivity. Even if the sales pipeline is well stocked, your turnover will always depend more on your salespeople than on your operating profits.

Finally, if the market is limited, you can produce faster, but that won’t change anything: there won’t be more customers to buy. In this case, productivity only translates into work comfort or investment in training, not cash flow.

Let’s be clear once and for all: if you cannot increase your business volume by at least the same amount as the increase in productivity, you will lose income. In other words, increased productivity only becomes income under two conditions: first, you must be at maximum production capacity (there is no way to produce more without adding resources), and second, you must be growing.

If the first condition is not met, you could have supported growth without needing to invest to generate productivity gains. If the second condition is not met, the investment made to achieve these gains will not be “covered” by additional revenue. In this little game, an 8% increase in your productivity, if commercial development does not follow or without new offers or new products capable of generating new demand, will allow you to send everyone on vacation from December 1 until the end of the year, but will not earn you a penny, even though you have invested to achieve it.

Productivity does not pay the bills: only the ability to generate revenue does.

When productivity becomes a pretext for cost cutting

When productivity does not translate into revenue, the temptation is to justify the investment in other ways, and many businesses then fall back on costs.

Klarna recently made headlines with its massive shift to AI and the promise of huge savings. Behind the rhetoric, the logic is clear: if you can’t demonstrate an impact on the top line, you justify the investment by reducing headcount. The problem is that these drastic cuts ultimately weaken the ability to grow and innovate, and even lead to the loss of what made the service unique in the market (AI will not create a competitive advantage). The immediate gain in margins is paid for later in revenue, and Klarna ended up backtracking and rehiring massively (AI, layoffs, productivity and The Klarna Effect).

This reflex is reminiscent of the distortions of Lean. Lean was never designed to “cut heads”, but to develop people and streamline flows. Yet how many businesses have reduced it to a cost-cutting mechanism? AI is now going down the same path: instead of using it to enrich skills and the value produced, it is being used as a pretext for cutting jobs (95% of Enterprise AI Pilots “Fail”–Just Like Lean? Not So Fast). We are forgetting the end goal, and the result will ultimately take its toll.

When productivity doesn’t translate into revenue, it ends in layoffs.

Let’s be cynical for a moment. If AI vendors are so keen to highlight cost savings, it’s not because it’s virtuous, but because they have no other immediate argument. They know they can’t yet demonstrate a clear impact on revenue, so they sell the only thing that can be measured in the short term: savings. It’s a short-term logic that reassures some executives and appeals to the markets, but traps businesses in a spiral where AI becomes a tool for downsizing rather than a lever for growth.

Bottom Line

It is easy to see why promises of productivity are misleading. While they may make executives’ eyes light up and attract investors, they offer no guarantee of increased revenue.

In reality, there are three key points to remember: an increase in individual productivity, even a massive one, does not guarantee an increase in collective productivity and therefore does not necessarily create more billable value. Even when there is a collective increase, it is often absorbed internally or captured by other players, which reduces its economic impact. Finally, productivity only becomes a lever for growth if the business knows how to convert it into revenue: through more customers, new offerings, or better valuation of its services. Otherwise, it becomes a simple cost adjustment variable, to the detriment of the long term.

This is where the difference lies between organizations that believe in the promise of technology and those that derive a tangible competitive advantage from it.

To answer your questions…

Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)

Related articles:

Who is handling your artificial intelligence projects? Probably not the right people. Cost reduction : a false good idea for organizations in tough times ? AI in the workplace: going beyond augmentation to actually transform Generative AI in the workplace: revolution or illusion? AI and jobs: why I don’t believe in the “great replacement” of humans by machines How can we measure the productivity gains of AI?