Kathy Gibson reports – Enterprises that are reshaping and reinventing business domains around artificial intelligence (AI) are on track to experience great return on investment (ROI) and become more competitive.
But there is a currently a big disconnect between AI investments and returns, says Sylvain Duranton, global leader of BCG X. While 75% of executives rank AI or generative AI (GenAI) as a top three strategic priority, just 25% report that they are making money from it.
“The big paradox, and the big gap we have to solve, is why is AI so powerful but not making money.”
It boils down to how companies are embracing AI, he says.
Deploying AI by bringing in AI assistants to perform everyday tasks is useful: it helps people to save time – but it won’t produce much of a return.
Reshaping business by using AI in workflow functions is a way to start rethinking processes and organisations around AI in single domains. The ROI here will be much higher than by simply deploying AI.
“Then there are new things I could do as a business if I am AI-first and working with AI to invent,” Duranton says.
These are the businesses that change they way they work, or invent new businesses that disrupt the market – and they are the ones seeing the big ROIs.
“Companies that want to make money on AI spend 80% of their money on reshaping and inventing; and just 20% on deploying,” he points out.
But it’s not easy to make these changes, and technology is only a small part of the process.
Duranton believes the AI model should account for just 10% of the investment, with 20% going to the technology and data. “The big thing – 70% of the investment – is about changing the way people work.”
This means that an AI programme cannot focus just on the algorithms, he adds. And the business people need to be in charge. “Only the business owners can realise the 70%. You can have the best CIO and the best tech, with an open chequebook, and they cannot do that.”
The role of the CEO
The AI-first organisation won’t be possible if it’s not driven from the CEO, and top leaders are rising to the challenge.
A BCG study shows that 72% of CEOs consider themselves as the man decision-maker on AI. And they have doubled the resources they commit to AI this year over last.
A massive 73% of CEOs have refocused budgets towards their AI transformation, while 94% says they will sustain their investments even if they deliver low ROI over the next 12 months.
Importantly, 50% of CEOs believe their job stability depends on getting AI right. In fact, a couple of prominent US-based CEOs have stepped down because they believe the company needs a different type of leader.
Not a typical IT project
The way an AI programme is conducted needs to be different from a typical IT project, Duranton explains.
The traditional way that IT approaches a project is to look at what a person does, identify the task and automate it with AI solutions.
An AI-first agency mindset, on the other hand, starts with the business outcomes and the team outcomes, then defines the ideal outcomes to be achieved in the age of AI – probably better than those already being achieved. AI can then be deployed to complete as much as possible of those outcomes.
The roles of the whole team also need to be redefined around the desired outcomes.
Duranton cautions against small-minded thinking when reshaping and inventing domains around AI.
“There is a saying in business that we need to start small then scale. My belief is that if you start small, you will stay small forever, with small ambition, small budget and small support. There will be a gravity in the company to ensure you stay small.”
He believes businesses with big ambitions and big budgets are the ones that will succeed – and reap the ROI rewards.
How to succeed with AI
Successful AI reinventions are driven from the top of the company: the C-suite and exco need to have the vision and ambition.
And the economic outcomes need to be defined. These could be one or two points of margin for big companies, Duranton says, or as much as five percentage points for smaller companies.
The deployment portion of the AI programme should be accomplished quickly, so that staff across the organisation start using AI tools.
Following that, two or three reshaping strategies should be launched and seen through to completion.
The C-suite and exco should explore an invest initiative. It’s important that the business does this rather than the technology team, Duranton says.
But tech and HR must be involved and invested. “There will be some restructuring so, before you press the button, have a clear strategy for the tech and people roadmap.”
While exco involvement is necessary, there must be a champion within the team driving the project forward. The champion needs to be credible from the business point of view but also comfortable with AI. This is also a great proving ground for the next generation of leaders, Duranton adds.
He urges businesses to think about reshaping and inventing as a capex project that will yield results in business outcomes – and this is how it should be monitored.
“The return will be real and big,” he says. “I believe there is no better return on your money. But you will have to spend to get a return.”
Managing expectations is also important. “It is hard to get these things right and often the first version doesn’t work. But when it does work, the ROI can be three to five times.”
Duranton points out that the companies that have succeeded have one thing in common: they find a domain and define the function, then use AI to improve that function.
“These are all super-focused programmes.”
The risk of doing nothing
Doing nothing is no longer an option, Duranton warns.
He points to a study that shows 31% of shoppers use AI on occasion for their purchases. More significantly, LLMs are the number one fully trusted source for shoppers – before friends or experts.
Today, 13% of consumers say they will buy what AI recommends, and 63% say AI gave them access to brands they didn’t know about before.
“So who is your customer? When everyone has their own shopping assistant, what is left for us?” Duranton asks. “It means companies have to start working on how they are perceived by the LLMs.”
Ai implementation
There are a number of issues that need to be thought through before the implementation and operation of AI projects.
AI runs on tokens, and token cost can quickly ratchet up as consumption increases. Today, token cost can be as much as 1% of revenue, and it’s set to rise to 3% soon.
“So being strong and effective at managing tokens will be competitive advantage,” Duranton explains.
Businesses need to weigh up the advantages and costs of production tokens versus open weight tokens, and engineer their software to be token-efficient.
“You need to measure your return on tokens as part of a whole new set of capabilities that needs to be measured.”
Tech functions will also be transformed, Duranton adds. There will be more AI models that need to be managed and secured. In fact, IT will need to invest twice as much as in the past on cybersecurity and responsible AI.
Companies also have to take a decision on their agentic platform, whether to use their own or those inside a SaaS platform, Duranton says.
This may require a complete rethink of how software is deployed. In the past, it was more cost-effective and easier to manage complexity by using one unified platform. But Duranton points out that AI is good at managing complexity, so CIOs are starting to move away from the big unified SaaS platforms.
With AI, dependence risk is a big concern so building in modularity and alternatives is important, along with strategic vendor management.
“In the old days, the question was whether to make or buy software,” Duranton points out. “In the end we are going to make, buy and run. At the same time we are going to have to look at insourcing some critical functions into IT.”
Inside an AI-first organisation
An AI-first organisation will look different from businesses today, Duranton says.
For a start, middle management is going to come under massive pressure. “With a lot of the work middle managers do taken over by AI, these roles will be more of wayfinders.”
They will also give way to super-doers, or agent orchestrators, who will supplant today’s process orchestrators.
Duranton thinks these super-doers will be the AI-native youngsters coming out of university now.
The office could start to look more like a factory: with AI agents close to day to day functions, employees will become operators, training and managing the agents.
But as AI drives down the cost of performing certain tasks to near zero, organisations will be able to offer an abundance of services that were inconceivable before.