The Customers You Can't See Are Already Leaving
Every hour, invisible customers visit your website. They browse your products, compare prices, read reviews, and attempt purchases. When they encounter friction, they don't call support. They don't fill out feedback forms. They simply leave—and recommend your competitors instead.
These aren't bots. They're AI agents acting on behalf of real customers with real purchasing intent and real budgets.
The Infrastructure Shift Nobody Discussed
Whilst we optimised click-through rates and A/B tested button colours, something fundamental changed. Microsoft embedded Copilot into the Edge browser. Google integrated the Shopping Agent into Search. Amazon launched Rufus for product discovery. These aren't experimental features—they're shipping to hundreds of millions of users this quarter.
The web is no longer browsed exclusively by humans.
What's Actually Breaking
Your forms fail because field names assume a human context. Your checkout abandons because error messages disappear after validation. Your product pages are confusing because state changes aren't explicitly marked. Your authentication blocks agents that inherit valid sessions.
These patterns work perfectly for humans. They're invisible barriers for agents.
The fascinating part? Fixing these issues makes your site better for everyone. Semantic HTML helps screen readers. Explicit state management improves mobile usability. Structured data accelerates human browsing. The patterns that accommodate agents are the patterns accessibility experts have recommended for years.
Why The Timeline Compressed
Two years ago, agent commerce was speculative. Eighteen months ago, early adopters experimented. Last year, major platforms built infrastructure. This month, competitors launched compatible systems within the same week.
When Target and Walmart—fierce retail competitors—both adopt the same agent protocol, that's not infrastructure standardisation. The question shifted from "whether" to "when" to "how quickly can you adapt."
The Seven-Day Window That Changed Everything
Between January 6th and 13th, 2025:
- Google launched Shopping Agent with real-time inventory integration
- Microsoft announced Copilot Checkout across retail partners
- Amazon expanded Rufus to handle complete purchase workflows
- Target and Walmart both enabled agent-compatible commerce APIs
This wasn't coordinated. It was convergence. Multiple platforms reached the same conclusion simultaneously: agent commerce is production-ready infrastructure, not experimental technology.
What You'll Learn at the Webinar
I'm presenting at the Boye & Company members' call on Wednesday, January 21st (14:00 London / 15:00 Aarhus / 09:00 Toronto). We're covering:
- The patterns that fail silently - What agents encounter that humans never see
- The quick wins - Changes you can implement this week that measurably improve agent success
- The diagnostic approach - How to identify where your site fails for agents without waiting for error logs that never arrive
- The business case - Why this becomes competitive infrastructure, not optional enhancement
This isn't theoretical architecture. We're examining real breakage patterns from real sites, documented over two years of research for "The Invisible Users."
Why I Wrote The Book
For 25 years, I've built websites optimised for human users. Forms that guide. Navigation that orients. Error messages that explain. Every pattern assumed a human reading the screen.
That assumption broke.
I spent two years investigating how AI agents interact with modern websites. Not scraping. Not automated testing. Actual customer-facing agents attempting legitimate purchases with valid payment methods and delivery addresses. The failure patterns were consistent, predictable, and completely invisible to traditional analytics.
"The Invisible Users: Designing the Web for AI Agents and Everyone Else" documents what breaks, why it breaks, and the specific patterns that fix it. Semantic HTML. Structured data. Explicit state management. Progressive enhancement. None of these are new technologies—they're established best practices that suddenly became infrastructure requirements.
The book launches in Q1 2026, but the patterns are relevant today. The webinar covers the critical issues you can address immediately.
The Convergence Timeline
- 2023: Early agent experiments, limited capability
- 2024: Platform infrastructure development, closed testing
- January 2025: Major platforms launch within seven days
- Q2 2025: Agent commerce becomes standard competitive infrastructure
- Q3 2025: Sites without agent compatibility face a measurable disadvantage
You're reading this in the middle of that timeline. The infrastructure shipped. The question is how quickly you adapt.
What Happens Next
The webinar is free, recorded, and focused on actionable patterns you can implement immediately. Whether you attend live or watch the recording, you'll understand:
- Where your site likely fails for agents right now
- The specific patterns that fix the most common breakage
- How to prioritise fixes based on business impact
- Why do these changes improve the experience for all users
Watching the January convergence unfold in seven days was remarkable. The timeline compressed. The competitive landscape shifted. The sites that adapt quickly gain a measurable advantage.
Join the webinar. Read the analysis. Fix the patterns that fail silently.
The invisible customers are already visiting. Make sure they can complete their journey.
Register for the webinar (January 21): https://www.boye-co.com/blog/2026/1/websites-work-until-dont
Connect: Email: [email protected] Website: https://allabout.network
Tom Cranstoun has built web experiences for 25 years, from early e-commerce platforms to modern progressive web apps. "The Invisible Users" synthesises two years of research into how AI agents interact with websites and the patterns that work for both agents and humans.
Related Articles