As a first meeting that has become clumsy, some companies find it difficult to gain artificial intelligence could try too hard.
They could undertake too many projects or not understand that IA manneurs often come to reclassle the work, not from “Super Cool” AI engines or Grand-Langue models, said Sylvain Duranton, world leader of BCG X, the Tech Build and Design of Boston Consulting Group.
These types of faux pas can make ball in large frustrations for business leaders, he told Business Insider.
Duringon said that if the big question of CEOs around AI in 2024 was which model to use, their request in 2025 is “where is my money?”
Indeed, he said, there are often challenges concerning the implementation of a wide use of AI.
“The scaling of this thing from a technological point of view – it’s difficult,” said Duranton.
To help companies recover their AI efforts, he said, its “golden rule” is that organizations allocate around 10% of their efforts and their money to algorithms-to build IA or LLM of training. An additional 20% should be reserved for data and technology. Essentially, it is for AI to work in the technological environment of a company, said duringon.
Most of the effort – the remaining 70% – should change the way people work, he said.
“Assuming that you have a technology that can evolve, you must put this in the hands of the people. It is an effort of massive change,” said Duranton, which is based at the Paris Paris office and oversees the BCG X global army of nearly 3000 technologists, scientists, programmers, engineers and others.
Some companies have trouble
The frustration of companies is real. In the last months of 2024, the BCG interviewed some 1,800 leaders of the C continuation of large companies in nearly 20 countries and found that, while 75% of respondents ranked among their three priorities among their three main priorities, only 25% declared that they had seen a “significant value” of technology.
To find more value, Duranton recommends that companies do not try to do everything at the same time. He said that the extent of the change in companies needed probably could not be made with tens or hundreds of use cases.
“It’s not the plan. The plan is to focus on very little and things that matter,” he said.
Duranton said that companies sometimes also turn to “additional initiatives”. He said he thought it is often a mistake. Instead, he said, companies should get into a few things “par excellence”.
For a retailer, said duringon, this could use AI to make sure that a brick and mortar store has the perfect product mixture for this location to better resist competitors nearby and online.
CEOs of retail include the issues, said duringon. They will often tell him something like: “I know that if I don’t do that better than the others, I am cooked,” he said.
Duringon said that another imperative for a retailer could be to develop an AI agent who can buy customers – the one who is so good that users will not want to go to a competitor.
“With these two things, you have both a strategic program and an AI program,” he said, referring to what the inventory is composed and the shopping bot.
The trick is therefore to keep the emphasis on these efforts, said Duranton.
“These quintesses, this is where you put all your money, all your energy,” he said. It is necessary, said duringon, if companies wish to adopt the 10%, 20% and 70% approach it recommends.
AI Bermuda triangle
He said that AI scaling is also often difficult because businesses can feel pressure to compromise on expenses, the quality of the results or the speed at which they are produced.
“You have a kind of Bermuda triangle, where it is either expensive and relevant with good latency, or you must compromise on one of the three, and you must optimize,” said Duranton, referring to the AI results.
He said it was often easy to demonstrate technological magic in a demo. What is difficult, said duringon is to manage millions of requests every day and produce timely and relevant results.
“It’s a different ball game,” he said.
In the end, to succeed with AI, said companies will have to bring people, not just robots.
“Invest in change management, not just technology, and ask your most intrepid and strongest leaders,” he said.