AI development is growing exponentially, and the hype makes it a powerful selling point to the extent that suppliers are peppering CIOs with products they call AI, but aren’t always so.
The current interest in AI is massive, and companies, as well as the public sector, are exploring the new technology in all its capacities as much as possible. But it’s important to be vigilant and painstakingly sort through all products that have an AI label.
“Many handle the word a bit carelessly,” says Charlotte Svensson, CIO at SAS, the Scandinavian airline. “They talk about AI without really seeming to know what it is.”
It can be about anything from classic data analysis and advanced data analysis, to robotics or machine learning. It’s all called AI, she says.
“The vast majority of companies already have a structure for analytics and machine learning, so we’re already there; it doesn’t add much,” she adds. “For those of us who have been around for a while, it reminds us of when the cloud came and it became popular to make everything cloud-based, and to put old legacy systems in the cloud and see it as modernization.”
In addition to many things being casually called AI, the sales pressure has also considerably increased.
“It’s lots of phone calls,” says Svensson. “I didn’t think it could be more. There are sales calls and workshops, and some book meetings right into the calendar. There are events and everyone has something to do with AI. Then they play on the “fear of missing out” if you don’t catch on.”
Even control questions that are asked to gauge interest in AI means time required to spend on it, and Svensson is not interested in that. Instead, she filters out most of what comes in, but if something catches her eye and seems interesting, she hands it over to her AI team for assessment.
“Even if it were to be called AI, even though it’s rather a robotization application, it doesn’t matter if it seems interesting to us. But it’s my team that makes that assessment,” she says.
SAS works a lot with AI already, though, with more traditional machine learning and evolving generative AI tools.
“It’s important to try your hand and see what works,” she says. “Inventing new innovative things in a lab environment isn’t that difficult. What is difficult is seeing what fits the specific logic of your particular industry, and what fits the architecture and what can scale.”
Aiming at the wrong level
Svensson thinks it’s strange that suppliers so often aim for the top of the hierarchy, toward the CIO rather than address the level that directly works with AI.
“If I were a supplier, I’d turn to the AI team instead of thinking a CIO should react to a random email,” she says. “They really understand if you have something unique to offer.
Emil Dahlin, group CIO at multi-technical service provider Bravida, also experiences an increased flow of sales in the inbox, but nearly all first meeting requests are rejected. In addition to the fact that AI is used generously for almost all products, he already has a clear picture of the type of AI tools Bravida needs.
“Five years ago, suppliers tried to sell their AI platforms, or entire solutions, but we didn’t find great use cases for automating with AI,” he says. “We’ve used other methods such as robotization and low code.”
Shoots wide of the target
With generative AI, where functionality can be built into other parts, the focus is now on things like predictive analysis and energy optimization by finding deviations in the property data that Bravida collects. The company also lets AI make 3D models to follow a construction or streamline internal training.
“We know quite concretely what we need, and most people who contact us about their products miss the mark because it’s about other things,” says Dahlin.
Just like Svensson, he’s bombarded with emails and calls and questions.
“They want to get me interested so I’ll then lobby for it,” he says. “But I usually say it’s better to get those who know something in a field to become interested. If in turn that team tells me they think a product could be a good fit, then it’s difficult to say no.”
Of course, he says, it’s interesting to try something experimental, but investing requires greater commitment to the business case.
“Then you have to be able to demonstrate real, quantitative value that creates cash flow and direct savings,” says Dahlin.
Lacks framework
Marcus Matteby, CIO of Sundsvall municipality on Sweden’s east coast, also feels the pressure from suppliers, although it’s a little different in the public sector where products are not only bought in but procured.
He also points out that today there’s a multitude of standards when it comes to AI, and that it’s something to keep an eye on before buying pre-packaged products that are offered.
“This also creates the risk of building a digital legacy with solutions that work on their own, but can’t be connected well together,” he says.
Listen to the words
Sundsvall recently procured development of AI models from a small company run by newly graduated KTH Royal Institute of Technology students, and their way of talking about AI shows they fundamentally understand it, says Matteby.
“That’s how you can tell the difference between those who learned buzzwords and those who really know something,” he says. “So a follow-up question if a salesperson gets in touch might be to ask them to elaborate on what they mean by AI.”
One way to distinguish the more serious players, he adds, could be that instead of speaking broadly about AI, to be more specific and talk about image analysis, natural language, or deep learning.
“It’s the simplest trick,” he says. “But maybe the next step for salespeople will be to learn it too.”
