AI ROI in 2026: it comes down to CEO accountability and cost visibility

KPMG's Q2 2026 Global AI Pulse survey of 2,145 senior business leaders across 20 markets is the clearest read yet on what the retailers getting paid back on AI actually do. Two findings do most of the work: organisations where the CEO is personally accountable for AI outcomes are nearly three times more likely to realise meaningful business value, and organisations with strong cost visibility are five times more likely to have established AI ROI. For a retail CEO, that is not a survey result. It is a list of two things to do this quarter.

The headline in the KPMG data

Adoption is moving. KPMG's "driving-adoption" cohort jumped from 13 per cent in Q1 to 22 per cent in Q2 - the biggest single shift on the maturity curve they have measured. AI spending is steady at $188M on average. Seventy-nine per cent of leaders now name AI as a key investment area.

And yet only 7 per cent of leaders surveyed have established ROI. Twenty-four per cent are under pressure to prove value to investors. The capital is flowing. The returns are still concentrated.

That gap is what makes the two correlations in the data interesting. They are not soft factors. They predict who is in the 7 per cent.

CEO accountability is not sponsorship

KPMG is careful with the language. Only 24 per cent of leaders say their CEO is genuinely accountable for AI-driven business outcomes. Another 29 per cent point to "broader C-suite responsibility" - which the report frames as sponsorship rather than accountability. The two are not the same.

The numbers reward the distinction. Where the CEO is accountable: 60 per cent confidence in the AI strategy, versus 22 per cent. Fifty-seven per cent realising meaningful business value, versus 21 per cent. Fourteen per cent with established ROI, versus 4 per cent.

For a retail CEO, the practical translation is direct. AI cannot sit with the CIO or the Chief Digital Officer as a delegated technology initiative. The commercial outcomes - basket size, conversion, margin recovery, store productivity, supply-chain working capital - belong on the CEO scorecard, with named targets, named owners, and a quarterly cadence. The CIO runs the platform. The CEO carries the P&L impact.

That looks like three concrete moves. Put AI outcomes on the executive scorecard alongside same-store sales and gross margin. Pick two or three commercial metrics where AI is meant to move the number, and report them to the board every quarter. Insist that AI strategy updates report dollars earned, not pilots launched.

Cost visibility is the second lever, and it is harder than it sounds

The cost-visibility finding is the more uncomfortable one. Forty-two per cent of leaders have only partial visibility into AI spending. Thirty-three per cent cite difficulty understanding cost structures, including tokens. Twenty-three per cent struggle specifically with usage-based pricing. And the kicker: organisations with strong cost visibility are five times more likely to achieve ROI (15 per cent versus 3 per cent).

This matters more in retail than in most sectors. Retail margins are thin enough that an AI workload running unmonitored against a personalisation engine, a customer-service agent, or a search-rerank model can quietly chew through the gross margin it was meant to defend. Token costs scale with traffic. Traffic scales with promotional cycles. A campaign that drives a 30 per cent traffic spike can drive a 30 per cent compute spike, and nobody finds out until the monthly invoice lands.

The retailers getting this right do four things. Cost dashboards by AI workload, refreshed at least weekly. Cost reviews embedded in the approval gates for any new AI use case - not added afterwards. Unit economics calculated per workload (cost per recommendation served, cost per query answered, cost per personalised email sent) and tracked against the revenue or margin that workload is meant to produce. And clear ownership for the bill - usually the commercial leader of the function the workload serves, not IT.

KPMG's Rob Fisher puts it cleanly in the release: AI is now as much a financial-management priority as a technology one. For a retail CFO, that reframes the AI conversation completely. It is no longer "did the pilot work?" It is "what is the gross margin contribution of this workload, and is it improving?"

What the two findings look like together

The reason these two findings matter together, not separately, is that they close the loop. CEO accountability without cost visibility produces ambition without instrumentation - the CEO can be on the hook for AI outcomes but cannot see what the AI is actually costing or earning. Cost visibility without CEO accountability produces good dashboards that nobody acts on.

The retailers in the 7 per cent have both. The CEO owns the commercial outcomes. The CFO owns the unit economics. The CIO runs the platform. Each role has a different report, but they reconcile to the same P&L line.

What to do this quarter

A pragmatic sequence for a retail CEO acting on this data:

First, put your name against two or three AI-driven commercial outcomes for FY27. Not initiatives. Outcomes. Conversion lift, margin recovery, working-capital reduction, cost-to-serve. Report progress at every board meeting.

Second, ask the CFO for the AI cost report. If one does not exist, the gap itself is the answer. Get a workload-level cost dashboard built within the quarter, and embed cost review in the approval gate for any new AI use case.

Third, push back when the AI conversation drifts back to pilots launched and platforms selected. Those are inputs. The KPMG data says the inputs are no longer the constraint. Accountability and visibility are.

The capital is committed. The technology works. The 7 per cent are run differently.

Source: KPMG, Growing adoption signals progress as cost visibility and accountability drive AI value (June 2026).