Where AI could make advertising more useful.
What becomes possible when a customer can explain a need, explore alternatives and take a useful next step in one conversation?
These are three experiments I would like to explore with businesses and product teams.
Start with what exists
OpenAI describes advertising that is separate from ChatGPT’s answers and clearly labeled. Its public guidance describes ads below responses and testing visual ads during image generation. Advertisers do not receive people’s private conversations.
Those boundaries matter. A useful commercial experience should help a person evaluate an offer without buying influence over the assistant’s answer.
Three opportunities I would test
From a developing need to a relevant offer
A person may start with a broad goal and discover a specific need through conversation. Could an offer become more useful once the customer has clarified their constraints? I would test relevance and qualified enquiries, and examine when showing no ad produces a better experience.
From an ad to a useful conversation
A click often sends someone into a page that starts the discovery process again. With the visitor’s permission, could a short handoff carry the context they choose to share into a business conversation? I would test whether that reduces repeated questions and improves completed bookings or purchases. A click alone would not establish success.
From visual exploration to practical delivery
Someone exploring a design may need a printer, installer or local specialist to make it real. Could a clearly labeled offer connect the idea to a business that can deliver it? The test would include fit, accurate expectations, delivery capacity and the resulting customer experience.
Design the smallest useful test
These are proposed experiments, not results or platform commitments. I would agree on a baseline, an evaluation window and a minimum worthwhile improvement with each business before testing.
Local services · clarify the need
Test: compare a needs-led offer with the existing service page for one trade and area.
Measure: qualified enquiries, cost per qualified enquiry and the team’s response burden.
Stop: qualification worsens or the business cannot respond. The acquisition lead and service manager own the decision together.
Appointment businesses · preserve context
Test: offer an optional, visitor-edited handoff summary before booking; compare with the current form.
Measure: completed and attended appointments, repeated questions and handling time.
Stop: visitors misunderstand what they are sharing, or handling costs exceed the agreed limit. The customer-service lead owns the handoff.
Printing · turn an idea into a brief
Test: help a customer translate a visual idea into a production brief a printer can quote.
Measure: quote-to-order conversion, clarification rounds and fulfillment margin.
Stop: inaccurate specifications or delivery expectations create rework. Sales and production jointly review the result.
Look beyond a single interface
Conversation, visual creation and agent-assisted research offer different ways to understand an opportunity. I’m interested in how a business makes its offer understandable across them: clear capabilities, current information, honest limits and a next step an agent can navigate. An agent-readable website is useful whether or not it receives paid traffic.
I would begin with one segment, interview customers and owners, define a measurable hypothesis and compare it with the existing path. Useful measures include incremental business outcomes, conversion, cost to serve and repeat business. The question is whether the experience creates value for both the customer and the business.
Read the sources
The platform descriptions below inform these proposed experiments.
Explore an opportunity together