Let’s begin with a striking statistic from The GenAI Divide 2025, courtesy of MIT NANDA.
| 95% of enterprise AI pilots fail. |
Let that sink in. 95%. But, on the bright side this means that 5% of enterprise AI pilots are succeeding. They are generating some form of business revenue in a successful manner.
So, what do the 5% of the enterprise AI pilots do differently from the other 95%? Here is a hint: it has to do with the role of L&D in the AI strategy.
Why Most Enterprise AI Pilots Fail
At the start of 2025, every organisation had one goal – to implement AI and get the high returns it was projecting. But very few companies managed to turn that “potential” into “returns.”
In our last The Learning Buzz episode, our guest, an executive adviser and a former CLO, Christopher Lind, framed the issue well.
“We were already really bad at understanding how work happened and what people did and what problems we were trying to solve. And so then you throw AI, the world’s most powerful accelerant and amplifier, just on top of it. … AI turned the volume up.”
The problem wasn’t that AI budgets weren’t enough. It wasn’t even that AI was the wrong tool. Put simply, the problem was that we threw an amplifier at the problem we were trying to solve instead of at the solution.
What Successful AI Adoption Looks Like
Let me give you an example.
When BMW transitioned its process of painting, sanding, and polishing the car body from human assisted to fully automated, it wasn’t throwing AI at a problem. It was implementing AI around a solution that its best employees had already perfected.
BMW saw what other companies failed to see – the best use of AI is in the hands of those who already know the work like the back of their hand. Consequently, they programmed the AI and the machines to follow the processes that the employees had already refined.
Here’s the detail that matters: one of the issues while painting the body is the small imperfections that show up in random places. The employees had come up with the solution to this – manual scoping and identification. And when AI was implemented, BMW didn’t find a new fix. It scaled the one that already existed. Now every time the body is painted, the computer scans and highlights imperfections that are resolved in the next step. The entire process happens faster and smoother thanks to human effort and AI integration.
Much like this, your AI integration becomes a success story when you implement AI on a solution – not a problem. For this, you have to start at the root cause. Talk to the people doing the work every day. How do they do the work? What are the everyday problems that arise and how do they solve them? What are repetitive, consistent tasks that can be amplified with minimal effort? These are the questions that make AI integration actually work.
The L&D Role in Enterprise AI Integration – A Four-step Framework
The job of L&D just doesn’t start at the tail end to simply “train employees on the new tools” once it’s already live. Our work starts at the heart of the job – with understanding what every day work looks like.
We follow a four-step process:
1. Diagnose the current state
We map how people and workflows actually work today. At this stage, we collect data on the work being done, the steps and processes involved, and the outputs produced.
2. Identify the skills gaps
Once we know how work actually happens, we look at where skills gaps actually live, and why, using data collected. Where does the friction occur, and what’s causing it – a process breakdown, a knowledge gap, or a confidence gap? How well does the employee understand the why behind each step, versus relying on instinct? Can they judge the quality of their own output?
3. Design learning into the flow of work
The advantage here is two-fold. First, the employee is not taking extra time to learn the “tool” separately – learning happens at the exact point of friction, inside the work itself. Second, we are able to see the effect of learning instantaneously. As soon as it is learned, it is applied and the effects of the learning reflect with immediacy on the dashboards – faster work and better output.
4. Measure Real Business Impact
Of course, all the learning is for nothing if the effects do not show up in the long term. So, we measure whether this sticks over time. Does the employee retain the new method months later, not just during the first week? Has the productivity gone up and stayed up?
L&D resides at the heart of AI integration as its enabler. Put simply, we help identify the gaps and close them with the right solution and the right learning, at the right point in the workflow.
When building enterprise scale AI solutions, L&D is no longer an afterthought – it’s an undeniable part of the strategy.
If you’re ready to move your AI pilots from the 95% to the 5%, let’s talk. Reach out to Apposite. Because the organizations succeeding with AI treat L&D as a strategic architect of workflow change, not a tool-training function.