Artificial intelligence
Automating a bad process only makes it bad faster
The useful question about artificial intelligence is not which tool to use. It is which process is worth automating, and whether that process deserves to exist as it is.
Almost every conversation about artificial intelligence inside an organization starts with the tool. Which platform to buy, which model is better, what the competition did. It’s an understandable conversation and almost always a premature one.
The question that decides whether AI will be useful is a different one: which process are you going to touch, and does that process deserve to exist as it is?
Automation amplifies
Automating doesn’t fix a process. It multiplies it. If the process is good, you get the same result faster and with less effort. If the process is bad, you get the same bad result, faster, at higher volume and with fewer people watching.
A common case: a sales team answers prospects with a generic message because it has no time to personalize. The obvious fix is to generate replies with AI. But if the real problem is that prospects arrive without enough information to qualify them, automating the reply only speeds up the moment the team discovers it’s talking to the wrong person.
AI doesn't fix a confused process. It makes it confused at scale.
Three questions before automating
1. What result does this process produce today?
Not what activity it generates. What result. If nobody on the team can say why a step exists, that step is a candidate to disappear, not to be automated.
2. Where is the real friction?
The most annoying task is rarely the most expensive one. Mapping the whole process, even on a single sheet, usually reveals that the bottleneck is a wait, an approval or a handoff of information between people, not the visible task.
3. Which human judgment can’t be lost?
Every process has moments where someone decides with context that isn’t written down anywhere. Those moments need to be identified before automating, to protect them or document them. Automate them without noticing and quality drops silently, and nobody can explain why.
What is worth automating
The best candidates tend to share three traits: they repeat often, they follow clear rules and their errors are easy to spot. Sorting incoming requests, turning a long document into different formats, preparing a first draft from structured information, moving data between tools that don’t talk to each other.
In those cases AI doesn’t replace anyone’s judgment. It frees up time so that judgment can be used where it matters.
The starting point
Artificial intelligence doesn’t need its own strategy. It needs to become part of the strategy the organization already has, applied to processes the organization already understands.
That’s why the first useful deliverable in an AI project is almost never a tool. It’s a map: what gets done, why, where it gets stuck and what would happen if it were done differently. With that map, choosing the tool becomes the easy part.