Does Your Company Need AI or a Better Process? How to Tell Before You Invest
There’s a question almost no company asks itself before buying an artificial intelligence tool: is the problem I want to solve actually a technology problem, or is it a process that never really worked?
It’s an uncomfortable question, because the answer, more often than not, is the second one. And for a company that has already decided this is “the year of AI”, pausing to review the process before buying the tool isn’t always the first thing on the agenda.
The most common mistake when starting out with AI
The pattern repeats itself in organisations of every size: a problem is identified, repeated queries, information nobody can find, tasks that eat up hours, and the first reaction is to look for an AI solution to fix it. The process that created that problem in the first place stays just as messy, only now it’s automated.
Deloitte’s 2026 Global Human Capital Trends report puts a number on this mistake: 59% of organisations take a technology-first approach to AI, choosing the tool before thinking about how people will actually use it. Those organisations are 1.6 times more likely to fall short of the ROI they expected, compared with organisations that start from the human problem and only then look for the right technology.
AI doesn’t fix a badly designed process. What it does is run it faster, which makes the underlying problem more visible, not smaller.
How to tell if the problem is the process, not the technology
Before evaluating any tool, it’s worth checking a few signals. If the answer to several of these questions is “yes”, the problem probably won’t be solved with AI:
- Can no one on the team clearly explain the exact steps of the current process?
- Is the same query or request resolved differently depending on who handles it?
- Does the information that’s needed already exist, but it’s scattered across several places or held by just one person?
- Is the bottleneck a slow approval decision, rather than a lack of tooling?
- Has the team already tried to tidy up this process before and left it half done?
When the problem is one of order, not capability, AI doesn’t solve it: it automates it exactly as it is, mistakes included.
The questions to ask before automating with AI
Once you’ve ruled out a process problem, the next step is choosing carefully what to automate first. Not every task is a good candidate, and not every one deserves the same priority. Before moving forward, it’s worth asking:
- Is the task repetitive, with a clear pattern? The more predictable a task is, the faster you see the results of automating it.
- Can the impact be measured? If there’s no way to know whether something improved (time, errors, satisfaction), there’s also no way to justify the investment to the rest of the organisation.
- Is the underlying information in order? Automating on top of incomplete or outdated data only multiplies the mess.
- Does it require human judgement at some point? Some tasks need oversight, at least at first, and that’s fine.
- Is it a task the team has already flagged as a real pain point? The best first use case isn’t the most innovative one, it’s the one people will be glad to see solved.
This exercise is what allows you to prioritise AI use cases with judgement, rather than choosing because it’s trendy or “the competition is already doing it”.
An applied case: repeated queries and access to information in HR
A common example helps to see this clearly. In many companies, the HR team gets the same questions every day: how to request holiday, where to find a particular policy, who to notify about a change of personal details. They’re simple queries, but they take up hours that could go elsewhere.
That’s where the temptation appears to add an AI assistant that answers automatically. It can work, but only if the information that assistant gives out is already organised, up to date, and the same for everyone who asks. If the holiday policy changed three months ago and two versions are still circulating, no AI assistant is going to fix that: first, someone needs to settle which version is correct.
This connects with something we already explored in another article on how AI is changing the way companies reduce staff turnover: technology strengthens what’s already working well. On what isn’t working, it only adds speed to the mess.
Why it pays to start small and measurable
Choosing your first AI use case in a company doesn’t have to be a big or risky decision. Quite the opposite: the more contained and measurable the first attempt is, the faster you learn what works, and the easier it is to show the rest of the organisation concrete results before scaling up.
For those advising companies on their technology strategy, and in particular for partners guiding this conversation in the HR Tech market in Latin America, this is the conversation that really adds value: it isn’t about selling the most advanced tool, it’s about helping the client work out what they need to solve before deciding how to solve it.
That distinction, process first, technology second, is, ultimately, the difference between an AI investment that pays off and one that ends up as just another tool nobody uses.
