How IMI uses AI to help customers stay ahead of demand

In supply chain, the difference between a customer who runs out of stock and one who doesn’t often comes down to a single question: did you know what was coming? At IMI, answering that question better and faster is exactly where we’ve been putting AI to work.

We sat down with Johan Persson, AI & Data Scientist at IMI, to talk about how the company is applying AI in practice, what it’s learned, and where it’s heading next.

Johan Persson, AI & Data Scientist at IMI

Getting the right product to the right place at the right time

One important area of IMI’s AI work today sits in demand forecasting. Using machine learning and to a growing degree, deep learning, IMI builds forecasts of customer sales that help them purchase the right products, in the right volumes, at the right moment.

It sounds straightforward. In practice, it’s anything but. Supply chain demands an understanding of large, complex flows of data and products, and building models that account for seasonality, supplier lead times, and shifting customer behaviour requires genuine technical depth.

For IMI’s customers, the output is practical: fewer stockouts, less overstock, and purchasing decisions grounded in data rather than instinct. AI is giving the people running these businesses sharper information to act on.

Building for what customers actually need

IMI’s AI ambitions extend beyond forecasting. Johan describes a prototype he built to help customers automatically compare incoming orders against order confirmations, a time-consuming manual task that creates friction and risk in the supply chain.

The prototype used a locally-run large language model, specifically designed so that sensitive customer data would never leave their environment or be shared with third parties. When it was built, the technology wasn’t quite ready to deliver the precision the task demanded.

“In and with the enormous advances in technology, it’s perhaps time to look at that project again,” Johan says.

That’s a deliberate approach: prototype early, evaluate honestly against real customer needs, and return when the technology has caught up. It reflects a broader discipline at IMI. Don’t deploy AI for its own sake, deploy it when it genuinely serves the customer better.

Faster, more tailored solutions, at lower cost

One of the less obvious ways AI benefits IMI’s customers is through the software development process itself. By late 2025, Johan had shifted to letting AI models write the majority of his code, as the next step in a long evolution.

“The programmer no longer needs to write the correct syntax for the code. You can write instructions directly in English or your native language and let the LLM translate that into the specific programming language’s syntax.”

Johan traces this as the latest step in an 80-year arc, from hardware and software being inseparable, through machine code, through programming languages, to natural language today. Each step made software faster and cheaper to build.

For customers, this matters practically. Algorithms that were previously too complex to implement can now reach production. Tailored solutions, built around a specific customer’s workflow rather than a generic template, become more feasible. And the pace of development increases, meaning customer needs get addressed sooner.

A smarter way to invest in AI

Not every company is getting this right. An MIT report from August 2025 found that 95% of businesses investing in AI had not yet reached production stage or seen a return on their investment. Johan sees this as a warning about approach, not ambition.

“Don’t throw yourself into a vaguely defined ‘get AI’ project. Carefully consider whether the problem you want to solve should be approached with or without AI. AI is not by definition better than other solutions.”

At IMI, the question that comes first is always: what does the customer actually need? AI is one of the tools available to answer that question, and sometimes the right tool, sometimes not. That discipline is what separates useful AI from expensive noise.

Where this is heading

Looking ahead, Johan is optimistic about what AI means for the people closest to supply chain problems, not just the developers and data scientists, but the operators and managers on the floor.

“It has now become easy for the average person to develop software. I believe this will lead to an enormous number of smaller problems getting tailor-made solutions, built directly by the people working with those problems, regardless of their technical background.”

For IMI, that translates into a future where the gap between a customer identifying a problem and having a solution for it gets shorter and shorter. Supply chain will always require human judgement at its core, but AI is making it faster, sharper, and more responsive to what customers actually need.