Most AI projects are born answering the wrong question. The question that dominates boardroom meetings is “which tool should we buy” — when the one that actually determines the outcome is different: “is this process ready to receive AI, or are we just about to accelerate a problem that already existed?”
Two simple questions prevent most of the waste seen today in corporate AI projects — and neither one is about technology.
The two questions that come before any tool
- Is this process formalized as data, or does it only exist in someone’s head? If the answer is “in someone’s head,” there is no AI project possible yet — there is a documentation project first.
- Should this process still exist the way it is? Automating a step that only survives out of organizational inertia doesn’t create value — it just makes the waste run faster.
Ignoring the first question is the most common technical mistake. Ignoring the second is the most expensive strategic mistake. Both lead to the same place: budget spent, tool deployed, results nowhere to be found.
Why formal data comes before the tool
Think of an AI agent as a new employee on day one — no prior experience with the company, no accumulated context, no intuition about “how things work around here.” An employee like that learns in one of two ways: someone formalizes the process in writing, or someone shadows them closely until the routine is internalized through repetition.
An AI agent doesn’t have the second option. It doesn’t observe a coworker over months or absorb unwritten rules through daily contact. It operates on what is explicit: structured data, stated rules, documented examples. When the real process only lives in the memory of whoever has always done it that way, there is no material for the agent to learn from — and the AI project effectively becomes a project of trying to guess what nobody ever wrote down.
That’s why so many AI projects that “don’t perform as promised” fail before the technology even enters the picture. The tool works exactly as designed — the problem is that there was no formal process for it to follow.
The most common mistake: automating the habit, not the process
Even when the process is documented, the second question still needs an answer: does it still make sense? Much of what becomes “official process” in a company started as a temporary fix for a problem that no longer exists — and survives because nobody stopped to question it, not because it’s still the best way to do the work.
Putting a new tool on top of that old structure doesn’t transform the business — it just makes the outdated habit run with more apparent efficiency. The short-term result may even look positive (less time spent, fewer people involved), but the process itself remains the same bottleneck, only faster.
That doesn’t mean mapping the process before automating it is the most exciting step of an AI project. It isn’t — and it’s precisely because it isn’t glamorous that most companies skip straight to the tool. It’s also exactly why most companies are failing at this stage right now.
A quick test before approving the next project
Three questions, in this order, before any AI budget gets approved:
- Is the process written down somewhere other than one specific person’s head?
- If a competitor without this process were born today, would they recreate it exactly as it is, or do it differently?
- Can whoever is approving this investment explain, in one sentence, what business outcome it should generate — not which technology it will use?
If any of these three answers is “I don’t know,” the project isn’t ready for AI yet — it’s ready for an earlier, less visible, and more decisive step: organizing the business itself before accelerating it.
Follow Eleva Tecnologia for more on technology applied to business: follow @ElevaTechno on X or @elevatechnologies on Instagram, or learn more about the group at elevatec.net/about.
