When an AI assistant finds the product, compares the options, and completes the purchase, what happens to the brand’s relationship with its customer?
Tom Critchlow joins Alephic to explore that question through Airbnb, Intel, Southwest, and Visa. The conversation covers the decision to connect a business to AI assistants through the Model Context Protocol (MCP), how brands can remain visible, and why better knowledge of a customer’s preferences could make distinctive products more valuable.
Four takeaways
Treat integration as a business decision. Tom argues that offering an MCP changes how customers reach a business and who controls the experience. The timing depends on brand strength, distinctive inventory, and where customers already buy.
Better matching may change what people will pay. Agents can compare prices, but they can also find products that fit a person’s preferences more closely. Tom expects that second effect to create opportunities for differentiated businesses.
Design the information an agent receives. Product details, related products, instructions, and offers all shape an assistant’s response. Tom sees this as an emerging design and merchandising discipline, with unresolved questions about whose interests the agent serves.
Build on foundations that can survive changing workflows. Tom points to consistent brand knowledge and a clear understanding of customer needs. He also expects people’s responsibilities to expand as AI takes on parts of their work.
In this conversation
00:00 Brands in the AI age
The opening question: how should marketers and businesses respond as AI changes the way customers discover, choose, and buy?
00:25 Meet Tom Critchlow
Tom describes his background in SEO, independent consulting, media, and ad tech, and the questions he is now working on at Alephic.
01:21 AI assistants that take action
Tom expects personalized assistants to become more widely used because they can complete tasks, use computers, and act on a customer’s context.
03:23 Should brands build an MCP?
The discussion moves from appearing in AI answers to letting an assistant interact with the business itself. Tom frames this as a strategic choice for brands.
05:11 MCP usage: company growth figures
Tom cites reported adoption figures as evidence of growing usage, then asks what value a brand retains when more activity happens through an assistant.
06:48 Airbnb’s integration dilemma
Airbnb illustrates the tension between opening up to agents and protecting a distinctive app, experience, and customer relationship. Tom expects a period of mixed consumer behavior.
10:08 Who owns the customer?
A look back at the early web suggests that new technology can create new intermediaries while forcing existing ones to adapt.
13:09 Intel Inside and brand visibility
Tom uses Intel Inside to explore how a brand can stay recognizable while operating beneath another company’s interface. The conversation also considers agent access to everyday services.
16:38 Southwest: when to join a platform
Southwest’s approach to third-party booking prompts a question about timing: when does joining a platform become more valuable than remaining separate?
18:17 The 2×2: visibility and customer ownership
Tom introduces two dimensions for assessing a brand’s position: how visible it is and how much of the customer interaction it owns. Visa provides another example.
20:19 Will AI make buying cheaper?
Better price comparison could lower some costs. Tom argues that better preference matching could also increase willingness to pay for the right product or experience.
24:57 Designing brands inside AI assistants
The group compares early AI integrations with the beginnings of mobile apps and considers how much room brands may gain to shape their own experiences.
27:27 How brands shape AI recommendations
Tom discusses how endpoints, returned information, related products, and offers influence what an agent presents. He also raises questions about transparency and consumer protection.
30:48 The next platform wars
The conversation considers how developer ecosystems, exclusive integrations, and commercial deals could shape competition among consumer AI assistants. These are possible futures, not announced arrangements.
34:29 Businesses built for AI assistants
Tom describes services designed to be adopted and used through an agent. How agents discover and evaluate those services remains an open question.
38:18 AI reviews and service feedback
MCP connections can carry information in both directions. Tom explores how agent-written reviews and usage feedback might create new services and improve existing ones.
41:04 Marketing around customer tasks
Tom suggests looking at the larger task a customer wants to complete, such as planning a vacation, and deciding how much of that journey the business can serve.
46:36 What should you build now?
The discussion weighs improving today’s work against preparing for changing customer behavior. Tom sees richer understanding of customer preferences as a promising place to focus.
50:46 Brand consistency and expanding roles
Tom offers two principles: build enduring brand knowledge, and expect people to cover a broader range of tasks as AI handles more of the individual steps.
53:51 Should Airbnb build an MCP?
Tom’s conditional answer considers Airbnb’s own AI experience, how distinctive its inventory really is, and whether customers can book the same stays through competing services.
