When AI Chooses What We Buy
AI can do more than return a list of options: it can recommend a brand, advise against it or try to complete a task for a customer. At BRXND NYC 2026, Alephic's Tom Critchlow explores what that means for discovery and commerce. He introduces task gain as a way to think about whether a brand actually helps an agent accomplish the customer's goal.
Key takeaways
- Look at the quality and context of recommendations, including brand preference and negative mentions.
- Test whether assistants can complete real tasks, such as booking or buying, instead of measuring visibility alone.
- Experiment with connectors and the information agents need, while treating the eventual interfaces and standards as unsettled.
- Consider how a brand remains recognizable when a customer delegates the journey to an agent.
In this session
00:00 Why AI search changes the customer journey
Critchlow opens with a travel-planning example that combines many personal constraints and feels different from a conventional search.
02:09 Recommendations, brand sentiment and measurement
He argues that citation counts miss the sentiment and context of AI recommendations and describes examining what the web says about a brand.
06:38 Brand positioning and testing content with AI
The discussion turns to consistent positioning, factual brand information and testing how AI responds to different content.
08:06 From information search to task completion
Critchlow introduces task gain and the shift from finding information toward asking an assistant to get something done.
10:33 Measuring task gain in real customer journeys
He describes experiments with real customer journeys and the gaps that appear when agents try to complete them.
11:50 Experimenting with MCP connectors
He proposes MCP connectors as a way to learn by building, while acknowledging that the eventual form of agent commerce is uncertain.
12:48 Merchandising and marketing to agents
Critchlow considers how recommendations, merchandising and basket-building could work through interfaces used by agents.
14:42 When agents leave reviews
He explores the possibility of agents sharing reviews of software and services with one another.
15:29 Staying visible as a brand in an agent workflow
The talk closes with the unresolved question of brand visibility inside an agent workflow and a call to experiment.
Recorded at BRXND NYC on November 5, 2026, at The TimesCenter in New York.
Written by
.png&w=3840&q=75)
Alephic Writing
Ideas for building the AI-native enterprise.
New essays on AI, marketing, software, and the organizations being rebuilt around them.
