AI intelligence. Human governance.
Sweden’s MASAI trial screened more than 100,000 women. AI-supported screening found 29% more cancers and cut radiologists’ reading workload by 44%. That’s great news for women, and a lesson for anyone using AI at work.[1]
Almost every retail media vendor now says the same thing about AI: it recommends, a person approves. That promise has become table stakes.
The questions that separate one platform from another are sharper. What can AI find that your team can’t? Who owns the controls? And can you show your working when finance asks?
This piece sets out why brands should use AI in retail media, what it uncovers, and how to keep it under control.
A lesson from breast screening
MASAI, published in The Lancet Digital Health in 2025, is one of the clearest tests of AI working alongside experts, comparing AI-supported screening with standard double reading.
The AI didn’t replace anyone; it cleared their plates. It scored each scan by risk and flagged areas of concern, freeing radiologists to focus on the complex cases. Clinicians made every final decision, and each flag could be checked against the image itself.
Three lessons carry straight across to retail media:
- AI adds most value where volume overwhelms people
- Trust comes from evidence experts can inspect, not a confidence score
- Accountability stays with the professional, so the process must show who decided what
Why use AI in retail media
Retail media has the same problem as a busy screening unit: more signal than people can read. A brand on three or four networks faces separate consoles, metric definitions and reporting cycles. Teams spend the week rebuilding campaigns and reconciling spreadsheets.
That is exactly the work AI handles well:
- Spotting pacing issues and anomalies across retailers before budgets run dry
- Comparing performance by hour, day and promotion to inform day-parting
- Surfacing options for where the next pound should go, with evidence attached
- Taking repetitive set-up and reporting off the team’s plate
The goal isn’t fewer people. It’s experienced people spending their time on judgement, not admin.
What the data uncovers, and what it exposes
Put cross-retailer data in front of AI and patterns emerge that no single console shows. Think of the hour a category converts best on one retailer but not another, or SKUs still spending while availability is weak.
AI also exposes gaps. Recent industry research points to two:
Data readiness: a BCG study of 50 large retailers and 100 global brands found only 18% of surveyed retailers considered their marketplaces agent-ready. Product feeds, price, availability and fulfilment data must be right before agents can act on them[2]
Measurement confidence: a sponsor-backed GrowthLoop and Ascend2 survey of 291 US and Canadian marketers found only 40% were very confident in their incrementality measurement
Both point to one conclusion: AI is only as good as the evidence underneath it. Its first job is to make that evidence visible, including where it is thin.
Control it: governance you can show
If human approval is standard, brands should ask for governance they can inspect. Four questions to put to any platform:
- Can I see the evidence behind each recommendation?
- Is there an audit trail of what changed, when, why and who approved it?
- Do we own the controls, such as day-parting rules, or does the vendor?
- Does the platform show where data is missing, rather than filling the gap with a guess?
And one commercial question: if AI helps us scale, what happens to our fees?
Where Triffecta fits
Triffecta is an operating layer for retail media: one login to create, replicate and manage campaigns across the networks we support, with cross-retailer reporting in one picture.
Our AI follows a simple path: evidence, recommendation, controlled action. Each recommendation arrives with its evidence, every action is logged in an audit trail, and your team owns the day-parting controls.
Pricing is a fixed monthly fee. When automation lifts output, there is no growth tax on the result.
Use AI. Control it. Learn from what it finds.
Retail media, made easy.
What would you need to see before trusting AI with your retail media budget?
No obligation chat? Let’s talk
Also referenced: GrowthLoop and Ascend2 survey of 291 US and Canadian marketers, released 30 September 2026. MASAI – Mammography Screening with Artificial Intelligence trial.
#AI #RetailMedia #GovernedAI #HumanPlusAI #CommerceMedia #Triffecta