Decisions

AI makes explicit choices that could previously remain implicit: even the absence of a choice produces an effect. If nobody decides, the decision gets made anyway — by development case by case, by the vendor’s out-of-the-box settings, by the intermediary that brokers the relationship, or by the client who asks for a feature because it is the trend. Not deciding is a position, not a postponement.

Assess the strategic decisions AI forces by considering what it can do in relation to our brand, and therefore which positions we need to take.

The decisions fall into four dimensions.

Attract

Bring in users: citation, recommendation, search and discovery.

How much do we want AI to bring people towards us, and who decides whether and how to stay in AI’s answers?

To decide whether and how to be discovered through AI, we need to establish whether and how to stay in AI answers and, at the same time, understand what AI traffic is worth.

Interpret

Shape brand, content, positioning and reputation.

How much do we accept that AI says who we are?

Accepting that mediation also changes how we judge success: if AI is effective at telling people who we are, and users therefore interact less with our content, does engagement stay a KPI? This is also connected to the customer journey shortening. The question is whether time spent on content, sessions and pages are still measures of value, or whether they should be replaced by slower indicators: conversion, lead qualification and customer return.

Assist

Support chat, personalisation and progressive refinement.

How much do we want AI to stand between us and the user during use, acting as the interface between us and our users and replacing, at least in part, the touchpoints we control?

This choice concerns the moments when the relationship runs through us: which interactions should preserve direct contact, data and qualification, and which can pass through an intermediary to satisfy the user’s intent without delay.

Act

Handle agentic interaction, quotes, bookings, purchases and service delivery.

How much do we want AI to do things on the user’s behalf, accepting the risk of commoditisation rather than leaving the channel where comparison will happen anyway?

The delegation scale turns that position into operating rules: which actions the system can perform alone, which need confirmation and which it must not perform at all.

Examples

Four examples, to show there is no single correct answer:

The discovery-led retailer: interpret high, precision low. Travel is the clearest case of a sector accepting a loss of precision because discovery is its lever. Skift’s State of Travel 2026 finds only 6% of hotels appear in AI-generated search results and describes the current equilibrium as “AI recommends, the OTA fulfils”: OpenAI’s Apps SDK (6 October 2025) put Expedia and Booking.com in front of ChatGPT’s 800 million weekly users, preserving the boundary between discovery in chat and booking on the OTA’s own site. Tripadvisor’s CEO describes Perplexity-sourced customers as “incremental” and “high intent, ready to book”. The sector accepts being discovered in a context it does not control because that context brings customers whose intent is already formed.

Media: cannot close interpret down, because it is the product. AI Overviews cut CTR by 50-90% where they appear; traffic from Google to publishers fell by about a third in 2025, and some outlets report losses of up to 90%. The only practical route is to limit dependence on referrals and shift resources toward owned audiences.

The high-volume, low-margin service (where price is the lever): let AI act. In retail and CPG the transaction itself, not just discovery, is already subject to agentic mediation: Shopify Agentic Storefronts have been on by default for every merchant since late March 2026, and all three major card networks support agentic commerce. The citation battle in CPG shows who wins when a machine makes the comparison: P&G is the most-cited CPG brand in ChatGPT (AiRR score 78/100), while Unilever’s heritage brands remain nearly invisible in AI citations despite roughly $50 billion in revenue. It is the clearest illustration of The brand reduced to comparable attributes: when the only differentiator a machine can read is product documentation, whoever maintains it best wins, not whoever is best known.

Luxury: act at zero, interpret at maximum. Luxury’s value is scarcity, narrative control and human intimacy: democratised AI recommendations and frictionless purchasing erode exactly what is being sold. The decision observed in each case is the same: AI works behind the scenes, not in front of the customer.

• Burberry uses it to handle support requests.
• LVMH uses it to analyse data and increase retention.
• Louis Vuitton gives store advisors AI agents to assist them, not replace them. When AI interacts with the customer, it does so for advice, not for a transaction.