Published on the 12/06/2024 | Written by Heather Wright

Here’s what sets the high performers apart…

When it comes to generative AI, few are seeing a meaningful share of their earnings which can be attributed to the deployment of the technology – but those that are, show some clear trends, and highlight the challenges ahead for other organisations chasing the genAI dream.

That’s according to a McKinsey report, based on a global survey of 1363 organisations, 878 of which regularly use generative AI in at least one function.

“Capturing the full value of this technology often requires customisation.”

The report highlights the growing adoption of AI and genAI in organisations – 72 percent say they’ve adopted AI in at least one business function, with 65 percent of respondents saying their organisations – and they as individuals – are using genAI.

The investments – across most industries most organisations were investing more than five percent of their digital budgets on genAI – are starting to show some payoffs, the report suggests with organisations investing in genAI most commonly reporting cost decreases in human resources. It was in supply chain and inventory management, however, where genAI is providing meaningful revenue increases, with 30 percent of respondents reporting increases of up to five percent, and 18 percent reporting between 6-10 percent increases.

But despite plenty of optimism about genAI’s ability to create value for organisations, the report shows only a small subset of users – 46 out of the 876 genAI organisations surveyed, or five percent – can report that a meaningful share of their organisations EBIT can be attributed to the deployment of the technology.

Those early movers are attributing more than 10 percent of their organisations’ EBIT to their use of genAI. (Notably, 42 percent of the high performers say more than 20 percent of their EBIT is attributable to the use of non-generative, analytical AI).

So what sets the high performers apart?

The stats show they’re using genAI in more business functions – and average of three, where others average two.

Across the board, genAI’s most common new homes are in marketing and sales and in product and service development.

That holds true for the high performers, too, but they’re more likely than others to be using genAI solutions in risk, legal and compliance; strategy and corporate finance and in supply chain and inventory.

The range of activities they’re using genAI for is also wide – ranging from processing of accounting documents and risk assessment to R&D testing and pricing and promotions.

Bryce Hall, McKinsey associate partner, says among those capturing the most value from genAI, most solutions are highly customised or bespoke.

“While many companies are finding value from off-the-shelf genAI solutions, capturing the full value of this technology often requires customisation – for example training models on proprietary company and customer data or tuning models to improve performance within a specific industry or business context,” Hall says.

“Despite the spike in adoption of generative AI, we are still in the experimentation phase with many organisations seeking relatively simple, one-step solutions,” Alexander Sukharevsky, senior partner and global coleader of QuantumBlack, AI by McKinsey, adds.

Roughly half of overall survey respondents say they’re using off-the-shelf models, rather than custom designed solutions.

“This is a very natural tendency in the early days of a new technology, but it’s not a sound approach as genAI become more widely adopted. If you have it, your competitor probably has it as well,” Sukharevsky says.

But the companies forging ahead with early genAI wins are also paying more attention to the risks, McKinsey says.

“Perhaps because they are further along on their journeys, they are more likely than others to say their organisations have experienced every negative consequence from genAI we asked about, from cybersecurity and personal privacy to explainability and IP infringement.”

Given that, they’re more likely than others to report that their organisations consider those risks, along with regulatory compliance, environmental impacts and political stability, to be relevant to their genAI use and take steps to mitigate more risks than others do, McKinsey says.

Risk related best practice was far more developed among the high performers, who were nearly twice as likely as others to involve the legal function and embed risk reviews early on in the development of genAI solutions. They were also more likely to require genAI risk awareness and mitigation skills in their tech talent, and have models designed to allow audits, bias checks and risk assessment.

“High performers are significantly more likely than others to embed testing and validation in the release process for models, as well as to develop clear processes to iteratively improve model outputs,” Hall notes.

“Over time these kinds of practices will become even more important, as highly customised and bespoke solutions are the ones that will truly be differentiating for companies. Off-the-shelf solutions, by contrasts, are likely to become table stakes.”

Seventy percent of high performers admitted data, including defining processes for data governance, had posed challenges for them, with risk and responsible AI cited as a challenge by 48 percent.

The report sounds a cautionary note for organisations, noting that across the board 44 percent of the organisations using genAI reported they’ve experienced at least one negative consequence from its use. Topping the list was inaccuracy, with cybersecurity, explainability, intellectual property infringement and regulatory compliance rounding out the top five.

“Our previous research has found that there are several elements of governance that can help in scaling genAI use responsibility, yet few respondents report having these risk-related practices in place,” McKinsey says.