Problems before tools: getting AI right in highways and utilities

Start with the problem

3 min read by Triopsis 24th September 2025

AI has the potential to transform the highways and utility sectors — increasing productivity, improving decision-making, and unlocking value from complex data. But like any technology, it only works if it’s applied to the right problem.

Too often, organisations start with the tool rather than the problem. It’s the “hammer and nails” effect — once you’ve got a shiny new AI tool, everything starts to look like a problem it can solve. This is the wrong way round.

Start with the problem

The starting point should always be to define the problem you’re trying to solve — and assess its value. Ask:

  • What exactly is the issue?
  • How much does it cost in time, money, or missed opportunities?
  • Is it significant enough to justify intervention?

If the problem is clearly defined and valuable to solve, then — and only then — should you look at whether AI is the right way to do it.

Gartner’s AI adoption research echoes this, noting that “AI initiatives fail when organisations prioritise technology acquisition over clear business objectives” (Gartner, 2023).

Why “AI for reporting” isn’t enough

Take reporting as an example. Saying “we want AI to help with reporting” is far too vague. Reporting isn’t one task — it’s a chain of activities:

  1. Data collection – How is the information captured? Are field teams and mobile tools gathering accurate, consistent data?
  2. Data cleaning – Is the data free from errors, duplication, or gaps?
  3. Report structure – What metrics and visualisations actually matter to your stakeholders?
  4. Report production – How quickly and accurately can the final report be generated?

AI could be useful at one or more of these stages — but only if there’s a genuine bottleneck. If, for example, your KPI reports are late because field teams aren’t entering data correctly, no AI tool will fix that without first improving the mobile app process.

The right order of thinking

  1. Define the problem – Be specific and measurable.
  2. Assess its value – Is it worth solving?
  3. Break it down – Identify each step in the process.
  4. Target AI where it works – Apply it where automation, prediction, or optimisation can genuinely improve outcomes.

By starting with the problem, you avoid wasting resources on AI that looks impressive but delivers little real-world value. In sectors like highways and utilities — where budgets are tight, regulations are strict, and the cost of getting it wrong can be high — that’s not just good practice, it’s essential.

In our next post, we’ll dive into real-world examples from the highways and utility sectors where clearly defined problems met the right AI tools — and delivered measurable results.