Where 70% of AI value comes from (and what it means for retail)

BCG's Julie Bedard and Vinciane Beauchene published a piece in February 2026 that puts a number on something most retail boards already half-know. Only 10 per cent of AI value comes from the algorithms. Another 20 per cent comes from the implementation technology. The remaining 70 per cent comes from the workforce - leadership, adoption, skills, operating model. For retail leaders deciding where the next round of AI spend goes, that ratio matters more than the use-case shortlist itself.

The 10-20-70 breakdown, in plain terms

The headline figure is the 5 per cent. In BCG's Build for the Future 2025 study, only about one in twenty companies have produced real financial gains from AI - revenue or cash flow increases, plus material process improvements. That same 5 per cent posts three-year total shareholder returns roughly 3.6 times the laggards.

What separates them is the share of the program they put into people. The 10-20-70 split is BCG's distillation of what actually moves the needle: a small fraction in the algorithms, a moderate fraction in the implementation stack, and the majority in workforce change.

The implication for budget conversations is direct. The spend ratio needs to mirror the value ratio.

What the 70% looks like in a retail organisation

The 70 per cent is abstract until you put it into a retail operating context. In a retail business, the workforce change shows up in specific places.

Buyers and merchants using AI inside range planning, OTB cycles, and supplier negotiations. Planners running forecasting and allocation with AI sitting alongside the system. Store managers using it in rosters, labour deployment, and shrink response. Ecommerce merchandisers using it in onsite curation and search tuning. Marketers using it in creative production, audience build, and campaign analysis. Customer service teams using it in triage and response drafting. Supply chain using it in demand sensing and inbound flow management.

Each of those is a behaviour change inside a team that already has a full day job and a quarterly target. None of them happens because IT deployed a tool. They happen because a manager in that function is using the tool themselves, the team has had real time to learn it, the operating rhythm has been rebuilt around it, and the KPIs reflect the new way of working.

That is the 70 per cent. And it is where the workforce spend ratio needs to land.

Why leadership engagement is the first lever

BCG is direct about where the workforce work starts. Executive engagement is one of the strongest predictors of AI maturity. Companies that treat AI as a CEO-level priority, not a tech initiative, scale faster and capture more value.

The data point that lands hardest is the manager-modelling gap. At future-built companies, 88 per cent of managers actively use AI in their own decisions and daily work. At laggards, the figure is 25 per cent. Employees in the first group see AI as something their direct manager uses every day.

For retail CEOs and exec teams, the practical move is uncomfortable. Before asking a buying team or a store leadership team to embed AI, the executives leading those functions need to be using it themselves, in their own decisions, visibly. That single shift does more for adoption than any internal launch event.

The discipline of fewer priorities

The other piece BCG flags is focus. The future-built companies align on three or four central AI priorities, not thirty or three hundred. The discipline is to pick the three or four that tie directly to enterprise outcomes - margin, conversion, customer lifetime value, supply chain cost - and put the workforce investment behind those.

Governance, KPIs, protected learning time, manager modelling, operating model redesign. All of it focused. The pilots that do not make the cut either get parked or get sponsored at a lower investment level with no expectation of enterprise impact.

This is the move that frees the workforce budget the 70 per cent actually requires.

Upskilling that actually sticks

Three conditions BCG keeps coming back to on upskilling are worth taking at face value. The learning needs to sit in the flow of real work, not in a separate training calendar. The design needs to draw on behavioural science - leaders modelling, early visible wins, a clear why. And progress needs to be tracked against business metrics, not course completion.

For a retail organisation, that translates into AI learning anchored to actual cycles. Buyers learn AI by running their next range review with it. Store managers learn it through their next roster cycle. Marketers learn it on a live campaign brief. The work itself is the curriculum.

The companion move is workforce planning. Future-built companies are five times more likely to do strategic workforce planning than laggards - mapping which roles change, which emerge, which contract, and what the skills profile of the function needs to look like in two to three years. In retail, that workforce planning conversation is the one to start now.

The actual point

AI in retail is not a technology investment with a workforce side-effect. It is a workforce investment with a technology component. The 10-20-70 split tells you where the spend, the attention, and the executive bandwidth need to sit if the program is going to produce financial returns rather than a portfolio of pilots. The retailers that get this right will be the ones that treat AI maturity and operating model maturity as the same project.

Source: BCG, AI Transformation Is a Workforce Transformation (February 2026).