Retailers risk missing AI gains with 'tech-first' approach

People, processes, and organisation make up 70% of BCG's AI transformation principle.

Retailers that treat artificial intelligence (AI) as a technology project are failing to capture its value, as most of the work lies in people, process, and organisation rather than in the algorithm.

Speaking at the Retail Asia Summit 2026 in Hong Kong, Boston Consulting Group (BCG) managing director and partner Charlie Reid said AI is “a business and organisational transformation, and not a tech-led transformation.”

Reid explained the gap through BCG’s 10-20-70 principle, which he used to answer why more retailers have not deployed AI tools.

Under the principle, 10% of the decision rests on the algorithm, 20% on technology and data, and 70% on people, process, and organisation.

He cited BCG survey results showing that AI leaders concentrate over 80% of their AI budgets on two categories of initiatives, compared with about half amongst laggards.

These leaders also pursue about half as many use cases as laggards, going deep in “the areas that really matter” instead of covering “a much wider spectrum”.

Reid named three use-case areas: customer experience, retail operations, and enterprise AI.

Customer experience covers how retailers serve their customers, including agentic commerce.

Retail operations covers merchandising, promotions and pricing, which he said most retailers are starting to deploy and which gives a big uplift in the short and medium term.

Enterprise AI covers back-end efficiencies in finance, HR, procurement, legal, and technology.

Reid said the real drivers of impact are customer experience and retail operations, and BCG’s advice is not to start with enterprise AI because it does not drive much profit and loss impact.

In BCG’s research in Asia, about 40% of respondents each cited cross-functional collaboration and speed of execution as slowing strategy execution, whilst fewer than 20% cited the technology platform.

Reid said the best tool will fail if merchants do not trust its recommendations, decision rights are unclear, or store operations are out of sync with the calendar of changes. “You’ll get no benefit out of it at all,” he said.

He named two must-haves: clean data and an orchestration layer that lets different technology stacks work together within governance guardrails.

On agentic commerce, Reid described a customer asking an app for three dinner options for a family of four on a $200 budget, with the agent drawing on the customer’s shopping habits.

He said the technology is at a very early stage, but Walmart reports about 50% larger baskets and Kroger customers reach the end basket five times quicker.

On assortment, AI can combine customer preferences, product benefits, store demographics, and store constraints to design an assortment down to a single store.

He said retailers can get upwards of 5% like-for-like improvement from AI-driven assortment, with further margin gains from more own-brand sales and less waste.

Reid used supplier negotiations to show AI taking over months of data gathering by category teams, leaving them to focus on category strategies, supplier relationships, and negotiations. He added that this does not make category teams redundant.

On how to start, Reid described a strategy-led approach for organisations with a company-wide AI agenda, and a single proof point for those without the organisation behind them.

“Either take the strategy-led approach or the singular proof-led approach, but go and get value immediately and mobilise the organisation behind that,” he said.

His other takeaways were to double down on fewer, bigger bets, and to secure early wins that give people in the organisation comfort.

Reid said AI is not a strategy in itself but a “powerful enabler” of that goal. The real strategy is to “win the retail customer and gain market share”.