Retailers should shape the agent rules, not receive them: a response to BCG | Particular Audience
A response to BCG's 2026 research on AI shopping agents: feeds versus MCP, sponsored LLM visibility, intent data, and why retailers must lead agent protocol standards.
What BCG said
BCG's research, presented in London as Ascendant Network and published as "How retailers can win with AI shopping agents", surveyed commerce media networks, D2C companies and brands. Its headline is upbeat: 77% of brands see agentic commerce primarily as an opportunity.
Retailers are less relaxed. The biggest concerns are losing customer intent and browsing data (74% large concern) and losing visibility in AI discovery, or having to pay for it (72%). BCG expects onsite to remain the largest share of commerce media revenue, with offsite and sponsored visibility inside LLM platforms growing from a small base.
Where we agree
Retailers need two tiers: serve the human shopper, and be the store an agent chooses. Better product, price and policy data is the first lever, and agent-facing formats will become a real source of commerce media revenue.
We also agree that most networks are still doubling down on core media, with agent-specific capability mostly on the roadmap. That gap is the opportunity.
Push back 1: feeds are not enough
Feeds give reach. But a feed publishes a list price and hands the comparison to the assistant. The assistant decides who looks best, with the retailer's own data.
Governed MCP tools change who runs the comparison. When an agent calls the retailer's tool, the retailer can answer with live stock, fulfilment and a contextual offer price through Price Beat, within its own floors and rules. Contextual, not personalised: the same offer for anyone in the same context.
Tools also let the retailer shape the whole answer, not just the price: merchandising rules, sponsored placements, and cross-sell or bundles that add relevant accessories, which usually carry better margin, and put more items in each shipment.
Assistants already allow more than one path to checkout. Retailers should insist on the same for discovery: feeds for reach, tools for the answer.
Push back 2: sponsored LLM visibility is not only offsite
BCG frames sponsored LLM visibility largely as buying placement on someone else's AI platform. Retailers can also sell sponsored answers and products inside their own assistant and agent surfaces, where they own the inventory, the rules and the measurement.
Push back 3: losing intent data is not inevitable
Good: own the agent requests through your own governed tools, and own the landing page with Intent Landing Panels, so the intent arrives with the shopper. Better than owning the landing page: move the function into the agent experience (see below).
Better than owning the landing page: don't wait for the click at all. Move the retailer's function into the agent experience itself. Search, compare, Price Beat, add-to-basket and sponsored answers run inside the conversation as the retailer's own tools and Functional Ad Units, so the agent does the job on the retailer's terms.
The retailer then keeps the intent, the offer logic, the sponsored inventory and the measurement even when no page view happens. Offers stay contextual, not personalised.
That is a shift from predictive context to semantic context. Even if click traffic falls, the share of signal can grow, because a prompt says more than a click ever did. Retailers who capture it keep their targeting, measurement and media value.
Retailers must lead the protocol, not receive it
Agent protocols are being written now. If retailers simply accept whatever standard arrives, they inherit someone else's rules on ranking, pricing visibility, attribution and who owns the shopper relationship.
Retailers should publish their own governed tools, decide what agents may see and do, and push for standards that keep offer pricing, sponsored disclosure and measurement in their hands.
Retailers and D2C platforms: main opportunities from AI agents
Winning share via better product/price data: 38% large opportunity, 80% large + medium.
Reaching new customers: 36% large opportunity, 68% large + medium.
Higher conversion from LLM traffic: 30% large opportunity, 78% large + medium.
New offsite commerce media opportunities: 26% large opportunity, 72% large + medium.
New agent-facing media formats: 26% large opportunity, 58% large + medium.
BCG survey (2026), n=50 commerce media networks. Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Retailers and D2C platforms: main concerns
Losing customer intent / browsing data: 74% large concern, 91% large + moderate.
Losing visibility in AI discovery, or paying for it: 72% large concern, 90% large + moderate.
Greater price/margin pressure: 50% large concern, 70% large + moderate.
Commerce media revenue shifting to AI platforms: 48% large concern, 73% large + moderate.
Loss of control over CX / brand experience: 44% large concern, 74% large + moderate.
BCG survey (2026), n=50 commerce media networks. Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Brands: 77% see the disruption as primarily an opportunity
Influencing purchases at higher-intent moments: 41% large opportunity, 79% large + medium.
Reaching new customers and shopping missions: 35% large opportunity, 76% large + medium.
Winning share with better product data and fulfilment: 23% large opportunity, 75% large + medium.
New offsite and agent-facing media formats: 17% large opportunity, 70% large + medium.
Reduced reliance on paid discovery if agents surface via merit: 30% large opportunity, 68% large + medium.
BCG Survey of Brands (2026), n=100 brands. Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Brands: main concerns
Limited transparency into agent ranking and recommendation: 56% large concern, 83% large + moderate.
Increasing dependence on a small number of AI platforms: 39% large concern, 73% large + moderate.
Exclusion from AI recommendations: 40% large concern, 72% large + moderate.
Losing ability to measure advertising impact: 36% large concern, 72% large + moderate.
Declining effectiveness of current commerce media formats: 25% large concern, 62% large + moderate.
BCG Survey of Brands (2026), n=100 brands. Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Commerce media revenue mix (excl. Amazon): today vs expected in two years
Onsite: 59% today, 52% expected.
Loyalty and personalised offers: 10% today, 12% expected.
In-store / front-of-store: 19% today, 18% expected.
Offsite (social, search etc.): 11% today, 15% expected.
Sponsored visibility within LLM platforms: 1% today, 3% expected.
BCG surveys of D2C companies (n=50) and brands (n=100), 2026. Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Most networks double down on core media; agent-specific features mostly on the roadmap
65% see improved campaign performance.
55% see incremental brand demand and spend.
40% see increased commerce media pricing.
BCG Survey of Retailers (2026), n=50. Among companies that already launched enhancements: Source: BCG, 2026 (Ascendant Network presentation, London) — redrawn by Particular Audience.
Sponsored Prompts: Sponsored answers inside the retailer's own assistant.
AEO and GEO guides: Be found and chosen by AI search and agents.
MCP and agentic commerce readiness: Governed tools agents can call.
Basket margin with MCP: Accessories, multi-item parcels, sponsorship and stock to cash in one answer.
Intent Landing Panels: Pages built around the intent of AI, ad and search arrivals.
Functional Ad Units: Ads that do a job inside AI chat and agents.
Price Beat: Contextual, not personalised, offer pricing for agents.
Trade Media: Suppliers co-fund the offer, retailers keep margin.
Bundles: In-stock bundles and accessories in the agent's answer.
Average Order Value: Cross-sell that fits the item.
Personalisation: Relevance for anonymous and known shoppers.
Buy Again: Previously purchased items, ranked by query relevance.
Sponsored Products: Core onsite retail media.
Sponsored Brand Pages: Brand pages and stores in a few clicks.
Measurement: Attribution and incrementality across surfaces.
External Demand Governance: Custom build, per retailer need.
Read BCG's article: https://www.bcg.com/publications/2026/retailers-win-ai-shopping-agents