I've Been Inside Three AV Companies Through Launch, Acquisition, and Rapid Growth. Here's What I'd Do Differently With AI.

I have spent 15 years inside AV companies. Not advising them. Not building software for them. Running them. Director of Sales and Marketing. Vice President. I have been on the manufacturer rep side, the distribution side, and the integration side. I helped launch a distribution operation from scratch. I led a company through a successful acquisition. I helped scale a self-managed organization through a period of rapid growth. The kind of roles where you own the number, own the process, and own the outcome when things go sideways at 10pm the night before a client's grand opening.

Three companies. Three very different chapters. Each one taught me the same lesson: the companies that adopt the right tools, and create the space for their people to actually use them, compound their advantages. The ones that wait, or that install innovation on top of a broken culture, fall further behind every quarter.

This is not a pitch. This is the honest account of what I saw, what I learned, and why those lessons eventually led me to co-found AVCodex.

Company one: from rep firm to distributor

The first company was a manufacturer rep firm when I came in. We made the decision to transition into full distribution, which meant building warehousing operations in Vancouver and Toronto. That is a fundamentally different business. You go from connecting buyers and sellers to carrying inventory, managing logistics, extending credit, and owning the fulfillment experience.

Capital was the obvious challenge. Warehouses are expensive. Inventory ties up cash. Credit lines need to be established with vendors who do not yet trust your volume projections. That part of the story is the one everyone sees from the outside.

The part that slowed us down more than the capital was the information gap. Our competitors had been distributing for decades. They knew which product lines moved, which ones collected dust, which integrators paid on time, which regions were growing. They had pricing strategies refined through thousands of transactions. We had spreadsheets and instinct.

We were not short on effort. We were short on accessible knowledge.

Every pricing decision was a research project. Vendor evaluations started from scratch. Technical customer conversations pulled senior people off other work because the knowledge to answer them confidently lived in too few heads. Marketing was guesswork because we did not have the data to know which products to feature, which segments to target, or which campaigns were actually driving orders versus just generating noise.

Looking back, that is the problem I would solve first with AI. Not automation. A knowledge layer that captured everything our experienced people knew and made it available to everyone instantly. If anyone on the team could ask, "What is our margin history on this product line, what did the last three quotes look like for this customer, and what are they most likely to buy next?" and get an answer in seconds instead of pulling someone off the phone, we would have moved twice as fast.

Customer support was the other bottleneck. We were a small team trying to serve hundreds of integrators from Vancouver to Quebec, spanning four time zones. Every product question, every order status inquiry, every RMA request landed on a person. An AI agent trained on our product catalog, our order system, and our warranty policies could have handled the majority of those inquiries without a human touching them. That would have freed up our team to do the work that actually built relationships and closed deals.

We did not have any of that. Nobody did. So we did it the hard way: hiring experienced people, learning by trial and error, and spending months building knowledge that should have taken weeks.

Company two: navigating to acquisition

The second company was an established AV integration firm where I served as Vice President for almost five years. I came in to help revamp operations, implement ERP systems, deploy new software, and build the processes that would let a team of 15 to 20 people operate more efficiently. It was a company with deep expertise, loyal clients, and a strong reputation. It was also a company with bottlenecks everywhere.

The Crestron programming bottleneck was the one that kept me up at night. We had one expert programmer. One. Every project that required custom Crestron work flowed through that single person. When they were heads down on a complex build, everything else waited. When they took vacation, projects stopped. The entire project pipeline was gated by one person's availability. That is not a staffing problem. That is a structural risk.

Getting technicians to site was another constant challenge. Rolling trucks is expensive. Every dispatch costs time, fuel, and a technician's day. And half the time, the issue could have been diagnosed remotely if we had better tools for it. But we did not. So we rolled trucks.

We had 15 to 20 people and the knowledge to serve our clients well. What we did not have was a way to make that knowledge move faster than the people who carried it.

Information flow was the deeper issue underneath all of it. In a team that size, you are too big for everyone to know everything and too small for formal departments with documented processes. Knowledge lived in people's heads. When a client called with a question about a system we installed two years ago, someone had to remember the details or dig through files to find them. When a new technician joined, it took months before they could handle a service call without calling back to the office for help.

Then COVID hit. The AV industry went from full speed to almost zero overnight. Projects froze. Clients locked their doors. The companies that survived were the ones that could adapt their operations quickly and find ways to support clients remotely. We navigated through it, but it tested every process we had and exposed the gaps we had been working around.

Through all of that, we also managed the process of positioning the company for acquisition. We eventually sold to a larger integration group, which was the outcome we wanted. It was a good result for everyone involved.

But I think about how much AI could have helped at every stage of that journey. An AI agent trained on our Crestron programming library could have let junior team members troubleshoot common issues without waiting for our lead programmer. A support bot could have handled the routine client calls that pulled technicians off billable work. A knowledge base agent could have made five years of project documentation searchable in seconds instead of hours. Even the acquisition process itself, the due diligence, the documentation gathering, the operational questions from the acquiring company, could have been dramatically streamlined with the right tools.

We did not have those tools. We had smart people working hard. And that got us to a successful outcome. But it took longer, cost more, and was harder on the team than it needed to be.

Company three: leading an integration inside a fast-growing organization

After the acquisition, I stayed on at the acquiring company as Vice President for about two and a half years. My first job was leading the integration: transferring knowledge, files, processes, and client relationships from the company we sold into the larger organization. If you have ever led a post-acquisition integration, you know it is one of the most operationally consuming things you can do. Every system needs to be reconciled. Every process needs to be aligned. Every person on both sides needs to understand how things work now.

But the integration was only part of the story. The bigger challenge was the organization itself.

This was a self-managed company. No traditional hierarchy. No command-and-control structure. People operated with a level of autonomy that most AV professionals have never experienced. For people coming from conventional organizations, and that included me and the team we brought over, it required a fundamental rewiring of how you think about work. You had to unlearn years of habit around waiting for direction, escalating decisions, and deferring to titles.

Onboarding people into a self-managed organization is not a training problem. It is an unlearning problem. And unlearning is ten times harder than learning.

That transition was one of the most challenging things I have been a part of. It took time. It took patience. It took a lot of conversations. There were no managers to guide you through it. There were people within the organization you could seek out who had navigated the transition themselves, and they were generous with their time. But finding those people, scheduling those conversations, and working through the learning curve one meeting at a time was slow.

It created far more meetings and back-and-forth than should have been necessary. And for many people, it never fully clicked because the support systems to guide them through that shift at scale simply did not exist.

The company was growing fast. That kind of growth exposes every gap in your systems. When I arrived, expenses were being tracked on Excel spreadsheets and mailed in. For a company growing that fast, that is not a process. That is a fire waiting to happen. I helped deploy expense tools and other operational software to bring the infrastructure in line with the reality of the business. But every new tool was another change management exercise on top of a team that was already absorbing a lot of change.

The knowledge gaps were everywhere. Sales needed access to design knowledge. Operations needed visibility into project status. Design needed to understand what sales was promising clients. In a traditional hierarchy, this information flows through managers. In a self-managed organization, it needs to be accessible to everyone, all the time. We did not have that. Information lived in pockets, and people spent enormous amounts of time tracking down answers that should have been available instantly.

I think about what AI could have done for that company constantly. A knowledge base agent trained on the organization's self-management principles, processes, and norms could have guided new team members through the transition of unlearning and relearning at their own pace, 24 hours a day. Instead of waiting for a coach to be available, someone could have asked, "How do decisions get made here? Who do I go to when I disagree with a direction? How does role allocation work?" and gotten a thoughtful, accurate answer immediately.

Beyond onboarding, the operational possibilities were endless. A customer support agent deployed on-site at client locations that troubleshoots issues, gathers the right diagnostic information, automatically creates tickets in the service management system, and deploys technicians based on the client's support contract. That alone would have changed the economics of our service delivery. A proposal assistant that bridged the gap between sales, design, and operations so that scopes of work reflected what the company could actually deliver, not just what the salesperson thought sounded good. An internal knowledge agent that made every process, every standard, every piece of institutional wisdom searchable by anyone on the team.

I wish AI was where it is today three and a half years ago. It could have changed the entire dynamic of what we accomplished there. Not because the people were not capable. They were exceptional. But because the speed of knowledge transfer in a rapidly growing organization is the constraint that determines how fast you can actually grow. AI removes that constraint.

The common thread

Three companies. Three different stages. Three different sets of challenges. The same core problem underneath all of them: institutional knowledge was trapped in the heads of a few people, and the business could only move as fast as those people could share what they knew.

In the distribution company, it was product knowledge, pricing history, and customer support capacity. In the integration firm, it was programming expertise bottlenecked in one person, field knowledge trapped in senior techs, and operational context scattered across a small team. In the fast-growing organization, it was onboarding, cross-functional knowledge sharing, and the sheer volume of operational questions that came with doubling headcount in a few years.

Every time, the solution was the same: hire experienced people, wait for knowledge to transfer organically, and accept that the business would be bottlenecked by the speed of human-to-human knowledge sharing. That was the only option available.

It is not the only option anymore.

Why we built AVCodex

When my co-founder Colby Harder and I started talking about what we wanted to build next, we kept returning to this observation. Colby spent nearly 30 years in AV technology. He built CONTI, a Vancouver-based AV company with a 50-year legacy, and led it through a successful acquisition by ET Group in 2022. He had seen the same knowledge bottlenecks from the ownership side that I had experienced from the operations and sales leadership side.

AVCodex exists because we lived the problems it solves. Every template, every agent type, every deployment option was designed by people who have run AV operations, managed AV sales teams, evaluated and deployed AV software, and run AV companies. Not because we read about the industry. Because we have been part of it.

But here is what matters most: those templates are starting points, not finish lines. Your experience is different. Your bottlenecks are different. Your clients are different. AVCodex lets your team build whatever your business actually needs. You do not need to know how to code. You do not need to understand anything about how AI works under the hood. If your team can describe a problem, they can build the solution. The templates give you a running start. Your people take it from there.

What I would do differently

If I could go back to any of those three companies with the tools that exist today, here is what I would build first:

• A knowledge base agent on day one. Every manufacturer manual, every internal process document, every vendor agreement, every pricing sheet, uploaded and searchable by anyone on the team in plain language. This alone would have eliminated hundreds of hours per year of "let me find that" and "who knows about this" interruptions.
• An AI-powered customer support agent. Trained on the product catalog, order status systems, warranty policies, and common technical questions. Available 24/7 across web, WhatsApp, and phone. Not to replace the support team, but to handle the 60% of inquiries that follow predictable patterns, freeing the humans for the conversations that actually require a human.
• An on-site client support agent with automated dispatch. Deployed at client locations via QR code or web widget. Troubleshoots issues in real time, gathers the right diagnostic information, automatically creates tickets in the service management system, and dispatches technicians based on the client's support contract. No phone call. No waiting on hold. No truck roll for something that could have been resolved remotely. And here is the part most people miss: that same agent becomes a revenue channel. When a client realizes their current contract means a four-day wait for a technician but they need someone tomorrow, imagine the option to upgrade their service contract right there in the moment. That is not a hard sell. That is a client solving their own problem with a better option you made available. Net new revenue from an interaction that used to be a cost center.
• An onboarding agent for organizational culture and process. Especially in organizations with non-traditional ways of working, where new team members need to unlearn old habits before they can learn new ones. An AI trained on the company's principles, decision-making frameworks, and operational norms that people can query at their own pace.
• A proposal assistant that makes every salesperson your best salesperson. Not just first-draft scopes of work from room descriptions and product recommendations from past projects. Guidance on wording and framing that meets the client where they are. Language informed by how people actually make buying decisions. Not scripts. Intelligence.
• A cross-functional knowledge bridge between sales, design, and operations. So that what sales promises matches what design creates and what operations can deliver.
• An AI companion for every software deployment. The hardest part of any ERP implementation is not the software. It is the data migration, the knowledge transfer, and getting people to actually use the system. An AI agent available 24/7 to answer questions about how to navigate the ERP, how to run a report, how to enter an order correctly, would have cut the adoption curve in half.

None of this existed when I was in the thick of it. All of it exists now.

The compounding advantage

Across all three companies, I watched the same patterns play out. The details were different, but the bottlenecks were the same. Knowledge trapped in too few heads. Information that took too long to find. Departments operating in silos. Customers waiting longer than they should have for answers. Technicians rolling to sites for issues that could have been resolved remotely. New people taking months to become productive because there was no scalable way to transfer what the experienced team knew.

When I think about what AI could have done across all three chapters, I do not think about replacing anyone. I think about supercharging every person who was already there. The technician still goes to site when needed. The salesperson still builds the relationship. The programmer still writes the code. But every one of them has an AI layer underneath that handles the repetitive, the searchable, and the routine, so they can focus on the work that actually requires a human.

The point of AI in an AV company is not to do more with fewer people. It is to do more with the people you already have.

That is what compounds. Every knowledge base entry makes the next answer faster. Every resolved support interaction trains the system to handle the next one better. Every proposal the AI assists with improves the template for the one after it. The companies that start building this now, and that give their teams the space to actually experiment with it, will not just be more efficient in six months. They will be operating on a fundamentally different level than the companies that waited.

That is the lesson I took from 15 years inside AV companies. The tools keep getting better. The question is whether you use them to supercharge your team while it still gives you an edge, or start after it becomes table stakes.

Clayton Creed is the co-founder of AVCodex, the AI platform built for professional AV. With 15+ years in B2B technology spanning manufacturer rep, distribution, and integration, plus a Lean Six Sigma Black Belt, he builds the tools he wished he had.