For two years now, artificial intelligence projects have been multiplying in businesses, often with a mixture of urgency and fascination. But despite sometimes flawless technical deployments and widespread individual use, at least in the private sphere (Why the widespread adoption of AI by consumers says nothing about its future in the workplace), the effects on organizations often remain disappointing.
Productivity is increasing, but only in certain areas, and tools are multiplying, but the operating accounts show no sign of improvement. The reason is simple: AI projects are rarely led by the right people, or, to be more precise, they are not led jointly and there is a lack of understanding of the importance of each function within the business in the process.
In short:
- AI projects often fail to deliver results at the business level due to a lack of consistent governance and alignment between technical and strategic functions.
- Individual adoption of AI, while essential, remains insufficient if it is not followed by a transformation of organizational processes.
- The productivity generated by AI only becomes truly beneficial when it is integrated into collectively redesigned processes, involving managers and operational leaders.
- The failure of an AI project should be seen as an opportunity for structured learning, involving HR, operations, and knowledge management, rather than an end point.
- Transforming productivity into economic value requires strong involvement from marketing and senior management to reinvest the gains in strategy, otherwise they may result in job cuts.
The false start of AI projects
Most of the time, AI ends up in the hands of those who know how to use it, but not necessarily those who know what it should be used for. IT departments, innovation departments, and sometimes communications or training departments all embrace it with good intentions, but with a view to tooling or experimentation, rarely with a view to transformation. The business itself is busy testing and exploring, but without ever rethinking its production model.
This confusion between use and purpose is not new, but AI amplifies its effects. It is not enough to have the tools to make change a reality: above all, you need the ability to integrate them into a coherent project, and this is where most businesses fail.
Individual appropriation: the first step, but not the finish line
All transformation begins with the individual, and this is undoubtedly the only part that businesses have understood fairly well. Training, acculturation, “trying things out“: as Frédéric Cavazza notes, all of this is necessary, even fundamental ([FR]The digital divide is a problem that no one can ignore). This phase, which could be called AI literacy, consists of making everyone comfortable with the tool, allowing them to experiment in “sandboxes” ask questions, and express their fears. It is a collective effort, and it must be orchestrated by human resources.
In a more specifically French context, where AI deployments have been rejected by the courts for failing to comply with the legal framework, this phase is also part of the obligation to engage in dialogue with employee representative bodies and to assess the impact on working conditions.
Furthermore, without prejudging the maturity of senior management, the lessons learned from this phase can help them refine their ambitions, set an acceptable and achievable course, and, above all, build the trust that is essential for this type of project (Employees Won’t Trust AI If They Don’t Trust Their Leaders).
This is an essential step, but it is not enough. As Lee Bryant points out, learning how to use a technology is not the same as learning how to leverage it in a work context (Doing the Work: Why Learning is Key). Individual appropriation is a foundation, but it produces nothing on its own unless it is extended by productive appropriation.
Productive appropriation: when AI meets processes
This is where most organizations have fallen short so far. They stop at the individual use phase, congratulate themselves on local gains, and then are surprised when they see no results at the business level (The great illusion of technological productivity gains (including AI)). But they forget that individual gains only become collective if processes are redesigned (Local optimum vs. global optimum and the theory of constraints: why your productivity gains sometimes serve no purpose).
At this stage, operations, process owners, and managers must be involved. Workflows must be identified, mapped, and understood in terms of how tasks are linked and where bottlenecks occur. Injecting AI into an already flawed process not only adds nothing, but often accelerates the dysfunction (Why AI Won’t Save Your Broken System).
Productive AI is not about speeding up tasks, but about streamlining flows, and this requires rethinking interactions, dependencies, and sometimes even the raison d’être of certain steps. In other words, AI is not a performance tool but a revealer that exposes organizational inconsistencies. This is not a bad thing; on the contrary, it helps to build a solid foundation on which to build, but we still need to learn how to capitalize on it.
Learning from failure, an often overlooked step
Once this second step has been taken, it is still necessary to know how to learn. Many businesses discontinue an AI project as soon as it fails to deliver the expected results, instead of treating it as a source of learning, which is a major mistake.
HR teams, Ops, and knowledge management have a role to play here: understanding what didn’t work, formalizing the lessons learned, and reintroducing these lessons into the next cycle. This need to learn rather than simply observing results was already implied in Lee Bryant’s article, but we must also link it to a “lean” approach to transformation (95% of Enterprise AI Pilots “Fail”–Just Like Lean? Not So Fast). Failure is not the verdict at the end of a pilot, but a resource for succeeding in the next one, provided you have a structured approach to experimentation (The experimental organization).
What a business learns from its mistakes is a more reliable indicator of its AI maturity than the number of successful pilots it has completed.
Transforming productivity into value
Let’s assume that the previous steps have been completed: employees have familiarized themselves with the tools, processes have been redesigned, and the first efficiency gains are visible. And yet, very often, the business realizes that this does not translate into hard cash (Technologies sell productivity, but businesses want revenue). It may be working better, but it is not making any money.
The real issue is not productivity, but the purpose we give it. Freeing up time, simplifying tasks, or reducing friction only makes sense if we know how to reinvest these gains elsewhere. If this freed-up capacity is not directed toward new initiatives, new services, or a tangible improvement in what the business delivers, it will be useless, even if we have invested in it.
This is where other functions must come into play: marketing, to imagine how this capacity can be converted into sales on current offerings, into new offerings to conquer new markets, or into perceived value, senior management, to articulate this potential in a coherent economic review and even adapt its strategy in light of this new potential.
It is important to be very clear on this point. AI requires investment, and these investments are visible in the accounts, so the business expects to see compensation in these same accounts in return. If the capacity freed up by productivity gains does not allow for sustained growth through increased sales of existing or new offerings, then it will justify drastic cuts in resources, starting with jobs.
This is the least pleasant scenario, but it is very much on people’s minds today. If marketing fails to transform productivity gains into growth, it will be up to HR to transform them into layoffs. It’s abrupt, but it’s the reality.
But when a business embarks on the path of AI, and today there is no question about it, it must know from the outset that it will have to transform its success into value in one way or another, and it is not at the last minute that we start thinking about strategy and offerings. On the contrary, laying off employees is always easier.
You might think that it will be time to answer these questions when they arise, but keep in mind that your employees will be asking them from the outset. “Where do we want to go with AI, what do we want to do with it?”. Between growth and contraction, the level of buy-in will not be the same.
But today, we have the impression that businesses are entering a tunnel that leads them into the unknown, when in fact, if they manage to get out of it, there are only two options at the end of the road, and we know what they are.
Bottom Line
AI has never transformed an organization on its own, and it never will. What transforms organizations are chains of actors, decisions, and the ability to adapt strategy and offerings.
Entrusting AI to technicians is hoping that transformation will come from a tool. Conversely, incorporating it into cross-functional governance means recognizing that it only makes sense if it connects those who learn, those who produce, and those who sell.
AI does not need more resources; it needs continuity and a few people who can understand how each initiative fits in with the next…
To answer your questions…
Failures rarely stem from technology, but rather from management. AI is often entrusted to technical or innovation teams with no connection to the business lines. The result: local experiments with no overall impact. To succeed, you need shared governance where HR, operations, and management work together on the project’s ultimate goal, not just the tool.
Individual appropriation consists of learning how to use AI; productive appropriation consists of integrating it into processes. The former makes the tool familiar, while the latter creates value. Without rethinking workflows and methods, gains remain isolated and have no overall effect.
Every project that fails is a source of learning. Rather than giving up, businesses must analyze what went wrong and capitalize on these lessons. It is this cycle of continuous improvement that builds true AI maturity.
Efficiency gains are only worthwhile if they are reinvested. To create value, they must be used to develop new offerings or improve commercial performance. Otherwise, sooner or later they will result in cost and workforce reductions.
Success depends on cross-functional governance linking HR, operations, and management. It enables scattered initiatives to be transformed into a coherent and sustainable strategy. Without this coordination, AI remains a tool with no real organizational impact.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
HR and IT merger: Moderna redesigns its organization for and with AI
AI in the workplace: going beyond augmentation to actually transform
Enterprise 2.0 and social business : what to expect in 2012 ?
Social Business and Enterprise 2.0 perspectives for 2014. Hint : it’s not about adoption and poorly digital
HR Tools adoption : is it a change or a design related issue ?
Three concrete scenarios for the evolution of the HR function
