Building data analytic systems is hard, but a Danish automation expert thinks artificial intelligence will smooth the way, just not instantly.

Heine Krog Iversen said: “Artificial intelligence will not be perfected overnight, so we need to prepare for it. There are still issues. If you talk to an OpenAI model today, it will hallucinate [give wrong information].”

The founder and chief executive of the data warehouse automation firm TimeXtender said problems emerge when we are unsure if our AI model is correct.

“You need to trust the AI model you are using,” he said. “With the industries that you have here in Bermuda, you are not going to take all your proprietary data and just put it in an open AI language model so that everyone out there knows what you are doing. That will not fly.”

TimeXtender partnered with local company Bespoke Analytics in 2019.

Mr Iversen talked on Tuesday at Bespoke Analytics’ Discoveries in Dataseminar, “The Preparation for AI: Automation and The Modern Data Estate”.

He said organisations need to ensure control of their data and model.

TimeXtender offers an automated pilot that helps users with their questions.

“That is just the beginning,” Mr Iversen said. “It is a model that we are training.”

He became interested in automation while working in the pharma industry for 20 years.

Installing enterprise resource planning systems, he noticed that 80 per cent of the back-end data that had to be reported on was “roughly” the same.

“I say roughly because it was not the same, but it was close enough that I started to think about how we could automate,” he said. “With AI coming into the picture, it is clear that we need to shift to doing things in a more automated way.”

He founded TimeXtender in 2006 in Aarhus, Denmark, aiming to shorten development times, simplify testing and automate multiple manual processes.

“We want to empower AI and build a data-empowered organisation,” Mr Iversen said.

He said TimeXtender is backed by private equity, so it has a strong growth plan.

“We are aggressive in acquiring more companies to expand within the space from data quality, data cataloguing and master data management,” he said.

He said in the areas where they have not been historically strong they are “buying their way in”.

“We need to get 20 times bigger, and we will actually do it,” he said. “We are focusing on what I call ‘real data’. That is not text or numbers, but stuff that we can put into fields and boxes.”

He said this is the type of data that needs to be trustworthy, but he estimated that 20 per cent to 30 per cent of the data in source systems is actually wrong in the source system.