Every week I watch AI absorb another slice of the work my teams used to fight over. Drafting, coding, summarizing, first-pass analysis. It is fast, it is cheap, and it is getting better while you read this sentence.
The instinct is to feel threatened. I feel the opposite. When execution becomes a commodity, the things machines cannot fake become the whole game.
I lead AI and ERP change at Novartis. My job is not to admire the technology. It is to get thousands of people to trust it, use it, and change how they work because of it. That job just got harder and more valuable at the same time.
Here is what I see rising in value.
Judgment beats output
AI produces ten plausible options in seconds. It cannot tell you which one is right for your context, your risk appetite, your people.
Judgment is knowing which question to ask before you ask the model anything. It is spotting the confident answer that is quietly wrong. It is deciding what not to automate.
In an ERP rollout, the model can propose a process redesign. Only a human who has lived through three go-lives knows that the "optimal" flow will break at month-end close because a plant in Singapore does things differently for reasons nobody documented.
The scarce skill is no longer generating answers. It is deciding which answer deserves to survive contact with reality.
How to build it
- Make decisions with visible reasoning, then check them against outcomes later. Judgment grows from feedback, not from confidence.
- Study your own bad calls harder than your good ones.
- Use AI as a sparring partner: ask it to argue against your decision, then judge the argument.
Trust is the real bottleneck
I can hand a team a perfect AI-generated plan and watch it die on the floor because nobody believes it, owns it, or feels safe raising the flaw they can see.
Technology adoption is not a technical problem. It is a trust problem wearing a technical costume.
People adopt tools recommended by people they trust. They ignore tools imposed by systems they do not. No model changes that.
How to build it
- Do what you said you would do, at the small scale, repeatedly. Trust is a compounding asset.
- Be the person who says the uncomfortable true thing in the room. Credibility is bought with candor.
- When AI gets something wrong on your watch, own it publicly. Nothing builds trust like accountability that costs you something.
Communication is the multiplier
AI can write. It cannot walk into a tense steering committee, read the silence, and reframe a stalled project so that four executives with conflicting incentives all say yes.
The value is not in producing words. It is in changing what people believe and do. That requires reading a room, sequencing a message, and knowing when to stop talking.
The people who will thrive are translators: fluent enough in the technology to be credible, human enough to make it matter to a finance director who just wants the quarter to close.
How to build it
- Explain something complex to someone with zero background, out loud, until they get it. Do this weekly.
- Cut every message in half, then check if it lost meaning. Usually it did not.
- Notice reactions, not just responses. What people do after you speak tells you if you actually communicated.
Sense-making is the new literacy
We are drowning in generated content. Reports, analyses, options, all technically competent, all pointing in different directions.
Sense-making is the ability to take a flood of information and produce a clear, defensible so what. To find the signal, name the pattern, and give people a coherent story they can act on.
When everyone has access to infinite analysis, the person who can say "here is what actually matters and here is what we do Monday" becomes indispensable.
How to build it
- Force yourself to end every analysis with one sentence a busy person could repeat.
- Practice synthesis across domains. Patterns live in the gaps between fields.
- Ask constantly: if this is true, what should change? Insight that changes nothing is just trivia.
The uncomfortable good news
These skills have always mattered. AI did not create them. It just stripped away the busywork that let us pretend execution was the hard part.
That is the trap and the opportunity. If your value was doing the task, you are exposed. If your value is judgment, trust, communication and sense-making, you just got a very powerful assistant and a lot less competition.
The machines are getting more capable. So should we. The difference is that our growth is the part no one can download.
Cédric Bignet is an AI & ERP Change Management expert at Novartis and founder of AInspire. He writes about change management, AI adoption and enterprise transformation.