Your Buyers Live in a Complex System. Your Website Thinks It’s a Simple One.
A manufacturing company we audited had a textbook funnel on their site. Homepage, product pages, pricing, demo request. Every page pointed to the next step. Google Analytics showed the path clearly: 70% of demo requests came from the pricing page. The funnel worked.
Except buyers weren’t following it.
Interviews with closed-won accounts revealed something the funnel couldn’t. One buyer had first encountered the company through a ChatGPT answer six months before ever visiting the website. Another had asked three different AI tools to compare this company against two competitors before clicking a single link. A third had been referred by a colleague, checked the company through Perplexity, read two articles, bounced, come back through a LinkedIn mention three weeks later, and then requested a demo.
None of these paths showed up in the funnel. The funnel said “pricing page converts.” The reality was that the decision happened upstream, in a system the website couldn’t see and wasn’t built for.
This is the difference between a simple system and a complex one. And most B2B websites are built for a simple system that no longer exists.
Two Kinds of Systems
Dave Snowden’s Cynefin framework gives us a useful distinction. In a simple system, cause and effect are predictable. You do X, you get Y. Turn the steering wheel, the car turns. Publish more content, get more traffic. Run an ad, measure the clickthrough.
In a complex system, cause and effect are only clear in hindsight. The same action can produce different outcomes depending on context. Agents within the system (in this case, buyers) adapt to each other, create feedback loops, and generate emergent behavior that nobody designed.
B2B buying has always been more complex than most marketers admitted. But AI tools have pushed it fully into complex territory, and most websites haven’t caught up.
The Linear Funnel Is a Simple-System Model
The marketing funnel assumes:
- Buyers enter at the top (awareness)
- They move sequentially through stages (consideration, decision)
- Each stage has predictable inputs and outputs
- More content at each stage means more progress to the next
This model works well enough in simple environments. It works when buyers are discovering you through a search engine, clicking through to your site, and evaluating your offering in the order you designed.
That’s not how buying happens anymore.
What Complexity Looks Like in B2B Buying
W. Brian Arthur made the case decades ago that economic actors don’t behave like the rational agents in classical models. They adapt. They learn from each other. They form expectations based on incomplete information, and those expectations change the system itself.
That’s exactly what’s happening in B2B buying right now.
A buyer doesn’t start at “awareness.” They start with a problem and ask an AI about it. The AI gives them an answer that includes your company, your competitor, or (more often) neither. If the AI mentions your company, the buyer arrives at your site already educated about your positioning, your pricing, and probably your weaknesses. They’re not exploring. They’re validating.
This creates several properties that the funnel model can’t handle.
Nonlinear Paths
Buyers don’t move in a straight line. They jump. They’ll evaluate three vendors through AI, visit one website, leave without converting, ask more questions in AI, come back to a different vendor’s site through a different channel, and then reach out. The path isn’t A to B to C. It’s a network.
Feedback Loops
A buyer’s experience with your content in one channel affects their behavior in another. If ChatGPT summarizes your positioning inaccurately, the buyer arrives at your site already skeptical. Your website is now fighting a correction battle it didn’t know it was in. The AI’s output became an input to the buyer’s judgment of your site.
Emergent Behavior
Individual buyer actions produce patterns nobody designed. When multiple buyers ask AI tools the same category of question (“What’s the best industrial chop saw for heavy fabrication?”), the AI’s answer consolidates and amplifies certain signals. Companies that are well-extracted by AI get recommended more. Companies that aren’t become invisible, not because they’re bad options but because they never entered the system.
Sensitivity to Initial Conditions
Small differences in how your information appears to AI tools (accuracy, specificity, consistency across sources) create outsized differences in outcomes. A company whose product pages are cleanly structured and factually precise gets extracted and surfaced. A company whose pages render content through JavaScript that AI crawlers can’t read effectively doesn’t exist, regardless of product quality.
We saw this directly in an audit of an industrial equipment manufacturer. Their entire product catalog was rendered through JavaScript that left it invisible to AI extraction. A buyer asking about their category would never hear about them. Not because the products weren’t good. Because the system couldn’t see them.
The Cost of Treating a Complex System Like a Simple One
When you build a website for a linear funnel and deploy it into a complex buying environment, three things happen.
First, you misattribute conversion. Your analytics say the pricing page converted the lead. In reality, the decision was shaped by AI interactions, peer conversations, and competitor comparisons that happened weeks before the pricing page was ever loaded. You fix the wrong page.
Second, you design for the wrong visitor. A funnel-optimized site assumes visitors are Explorers: people at the start of their journey, open to being educated and guided. But AI-referred visitors are Validators. They already have a mental model. They’re checking whether your site confirms or contradicts what the AI told them. If your site feels like it’s starting from scratch, they leave. This is the Context Gap: the disconnect between what a pre-educated visitor expects to find and what the page actually delivers.
Third, you miss the compounding effect. In a complex system, advantages compound. The company that shows up accurately in AI answers gets more visits, more mentions, more backlinks, and more AI training data, which makes it more likely to show up next time. The company that’s invisible stays invisible. This is the Proof Loop in action, but it only works if your information infrastructure is strong enough for AI systems to extract, correlate, and synthesize it.
What Designing for Complexity Actually Means
You don’t control a complex system. You participate in it. The goal isn’t to force buyers into a path. It’s to make sure your information is available, accurate, and structured enough that the system (AI tools, peer networks, industry forums) can surface it at the right moment.
Three principles.
Make your information extractable. AI tools need to read your content, understand it, and cite it accurately. That means clean structure, factual precision, and claims backed by evidence. Every page on your site should answer at least one specific question so completely that an AI system would have no reason to look elsewhere.
Close the Context Gap on every high-intent page. When a Validator lands on your pricing page, they shouldn’t see generic marketing copy. They should see the specific information they came to validate: how your pricing compares to what the AI told them, what’s included, what’s not, and why. Match the visitor’s Intent Signature, not your internal org chart.
Build for the network, not the funnel. Internal links between your content. External citations that give AI tools provenance. Consistent terminology across every page. The goal is a knowledge graph that reinforces itself, not a series of isolated pages pointing to a demo button.
The Shift That Matters
Here’s the reframe. Your website isn’t a funnel. It’s a node in a network.
Buyers are moving through a complex system of AI tools, peer recommendations, competitive comparisons, and their own evolving understanding. Your website is one touchpoint in that system. The question isn’t “how do I move them from awareness to decision?” The question is “when they encounter my company through any channel, does the information they find confirm the story I want told?”
If the answer is yes, the system works for you. If the answer is no, or worse, if the answer is “what information?”, the system works against you.
Most B2B companies are in the third category. They don’t know what AI tools say about them. They don’t know what buyers find when they validate. They don’t know because they’re still measuring the funnel and calling it insight.
The companies that figure this out first have a compounding advantage. The ones that don’t will keep optimizing pages that fewer and fewer buyers arrive at through the paths they were designed for.
