September 19, 2026 · 6 min read

Dennis Yu: How Google and AI Know Who You Are

Dennis Yu joined me to discuss KGMIDs, the client proof founders leave unpublished, and how documented experience can become content, a repeatable service, and eventually software.

• Interviews
• Personal brand
• AI

Before we started recording, Dennis Yu asked me about some of the work I had done for clients. I could name projects, results, and people I had worked with. Then he asked whether any of it was published.

That question stayed with me throughout this conversation. I had experience I could talk about, but much of the evidence was scattered across old conversations, projects, and recordings. Someone looking me up would not necessarily find it.

Dennis joined me on The Josh Collier Show to talk about how that gap affects a founder’s reputation online. We covered knowledge graphs, personal branding, AI, and the path from doing client work to building a software business. His examples kept bringing us back to the same practical question: where is the proof?

Watch the full conversation

Watch on YouTube · Recorded April 7, 2026.

What is a KGMID?

I asked Dennis to explain a term we had been discussing before the interview: KGMID, or Knowledge Graph Machine ID. He described it as an identifier associated with an entity, such as a person, company, product, or place, in a knowledge graph.

The useful part of that explanation was the relationship between those things. My name, my business, my podcast, and the work I have done are connected in real life. The information published about them needs to make those connections understandable, too.

We used my own search results as an example. Looking up my name brought up other people with similar names. Dennis’s question was whether the evidence about my work was clearly associated with the right person.

That is a useful way to think about an online presence. A bio makes a claim. A named project, an interview, a client’s account of the work, and a clear link back to the person who did it give someone more to evaluate. A KGMID is not a promise of a knowledge panel or an AI recommendation; in this conversation, it was our starting point for discussing identity and evidence.

Start with an inventory of what already exists

Dennis suggested asking an AI assistant to assemble an inventory of a person and their company: what they have done, who they have worked with, what other people have said, and where the information is incomplete or contradictory.

He was specific about asking for the uncomfortable parts as well. An assistant that only repeats a flattering bio will miss the point. The questions he proposed included who you actually serve, what your work suggests you stand for, and which blind spots you may be overlooking.

He also described collecting what he calls positive mentions: comments, interviews, and other accounts from people who have worked with someone. Those references may already exist without the founder having gathered them in one place.

My takeaway was straightforward. Before trying to invent another month of content, I need to look through the work and conversations I already have. For each item, I can keep the original source, the people involved, what happened, and the evidence that supports it. AI can help organize that material, but I still need to check that it describes the right people and says what the source actually supports.

Personal branding includes what other people say

Dennis described a personal brand as the accumulated experience other people have had with you. That includes customers, colleagues, partners, and the people you have helped.

He used Brady Sticker and Church Candy as an example. In Dennis’s account, Brady’s work with churches supplied the material for interviews, books, and practical explanations of his process. The content had a clear subject and a specific audience because the work itself did.

We also talked about the intent behind sharing someone else’s story. Dennis described researching people before meeting them, looking for shared connections, and making introductions that could help them. He sees those relationships as part of the work, rather than something to manufacture for a photo.

That landed with me. There are people I have worked with whose stories deserve more attention. A useful case study gives them room to explain the problem, the work, and what changed.

A daily video quota can produce the wrong content

I brought up the pressure to post a reel every day. Dennis challenged the assumption that more effort automatically produces a better video.

His example was the person posting about how far they had progressed through a video challenge. The creator had kept the schedule, but the viewer still needed a reason to care.

During that exchange, I said I could think of plenty of past projects and conversations worth revisiting. Dennis suggested having someone ask me about them. An interview can draw out the details that are easy to skip when you sit in front of a camera and try to talk about yourself.

For me, that creates a more useful starting point: a real question about a real piece of work. What was the client trying to do? What failed? What did we change? What can we show? The answers can become an article, a clip, or a case study without needing to invent a topic just to fill a slot.

Prove the service before turning it into software

Later, we discussed Marco Cipolla and the progression from agency work to a software product. Dennis described Marco narrowing his agency’s focus to concrete coating businesses, learning how to deliver results for that market, and documenting the process.

He then talked about HVAC Quote, a product built around a specific customer question: how much will a new system cost? The lesson I took from that example was about sequence. Doing the work exposes the recurring problem. Repeating it helps establish a process. That process can then inform the product.

Dennis also pointed out that an agency can mistake a retention problem for a lead-generation problem. If clients keep leaving, acquiring more of them does not explain why the service is failing to hold their interest or deliver what they expected.

We went on to discuss exits, acquisition terms, and buying other companies. Those were examples and possibilities Dennis raised, not a claim that every agency should become a SaaS business or pursue the same outcome.

What I’m applying from this conversation

I am starting with the backlog: client projects, interviews, testimonials, and results that have not been properly documented. Each one needs enough context for someone who was not there to understand it.

That means naming the project accurately, explaining my role, including the evidence I can actually support, and connecting it to the people involved. Then I can publish it in a form someone can use.

Dennis closed by talking about making his processes available as documented tasks and skill files. That fit the rest of the conversation: experience becomes more useful when another person can understand it, follow it, and check the result.

See the client work we discussed, or watch more from The Josh Collier Show.

Written by Josh Collier