Every Business Needs an AI Architect, but Don’t Bet Everything on One Person
Article 2 of an ongoing series on AI transformation for the office technology channel
In my last article, I said every business in this channel needs an AI architect. One person who owns the strategy, understands the tools and is accountable for moving the organization forward. The response told me the message landed and that the next questions were already forming. Fine, but who? Where do I find this person? What do they own? And what happens if I bet my AI strategy on one individual and that individual leaves?
Those are the right questions. Let me answer them from the inside, because I’m living this transition right now.
But first, let me say something that gets lost in all the urgency: this is the most exciting time I’ve experienced in 35 years at the intersection of technology and business. The ability to learn, build and evolve has never been stronger. A marketer with no coding background can now ship working software. A president can interrogate his own company data in plain English. The tools are on the table for anyone willing to pick them up. Yes, the pace is uncomfortable. It should also be energizing.
Here’s the honest backdrop: I don’t believe anyone can tell you what your business looks like five years from now. I can’t tell you what mine looks like. Anyone who claims certainty is selling something. But that uncertainty isn’t an argument for waiting; it’s the argument for structure. When you can’t predict the destination, you build the capability to adapt faster than your competitors. That capability needs an owner. And later this year, the next wave becomes visible to everyone: robotics. AI is about to stop being only a software conversation and start being a physical one. Another collision between intelligence and work is coming, and the businesses that built adaptive structure for the first wave will absorb the second one. The ones that didn’t will be reacting to both at once.
What an AI Architect Actually Is
AI architect is a function, not a title. It owns the AI strategy, translates business problems into buildable systems and is accountable for results, not experiments. That last word matters. My previous article cited MIT’s finding that 95% of enterprise AI pilots produced zero measurable P&L impact. Pilots without owners become science projects, and the architect exists to prevent that.
At AIS, every build passes through the three filters I described last time:
- Does it make us more profitable?
- Can we sell it to our customers?
- Does it give us better insight into our business?
Those filters aren’t parts of a slogan. They’re the architect’s operating charter, the standard against which every hour and every dollar gets judged.
Here’s what this looks like in practice. I’m currently restructuring my own role. We’ve defined what marketing 2.0 looks like at AIS, and marketing is transforming first, by design, because it’s the department I know best and where I can move fastest. I’m setting the vision and strategy so the majority of my time goes to our AI transformation.
The architect’s first project is often their own job; if they’re not willing to redesign their own role around this technology, they have no business redesigning anyone else’s.
Where You Find This Person
Promote domain knowledge over AI credentials. I will say that again, because it runs against every hiring instinct: your business knowledge is harder to acquire than AI skills. A sharp operator who knows your customers, workflows and P&L can learn to build with AI in months. An AI specialist parachuting in from outside will spend a year learning why your service department actually works the way it does, and your strategy stalls while they catch up.
I’m the proof case. I’m a 35-year marketer, not a coder. And today I’m building agents, writing code for commercially available software that solves business problems and generates revenue for our company, and supporting other industry executives through their own transformations. None of that came from a computer science degree. It came from curiosity, accountability and a decision to take this seriously four years ago.
The credentials that matter: a builder’s curiosity, business accountability and the trust of leadership. The reporting line that matters: direct to the president or CEO. Bury this role under IT and you’ve already told your organization it’s a technology project when it’s actually a business redesign.
The Leadership Model Above the Architect
The architect model fails without one thing: a leader who’s personally AI-fluent.
I have that at AIS in our president, Gary Harouff. Gary isn’t delegating AI and waiting for reports. He works with it directly against our own company data, finding new ways to look at the organization, testing decisions, seeing patterns that were invisible before. Our one-on-ones have become mutual learning sessions. He shares what he’s discovering; I share what I’m building. Without his support, I wouldn’t have the freedom to move at this pace.
This answers a question I know many of you are asking, because this channel is full of hands-on owners: how much autonomy do I give this person? But that’s the wrong frame. The leaders who win this era don’t yield power to an AI architect. They skill up alongside one. The architect proposes and builds within the agreed filters. Leadership sets direction and makes the judgment calls machines can’t. Neither works alone. Every business leader reading this should be modeling that behavior now, personally, before the gap becomes unbridgeable.
The Crusader Problem and the Honest Answer
Now the hard part. In ENX’s State of the Industry report on AI, Erik Braden raised a critique that deserves a direct response: a heavily crusader-driven hierarchy makes AI adoption “about that person’s energy rather than the company strategy.” If the point person burns out, leaves or changes roles, the program stalls.
He’s right. And I say that as the person this channel has nicknamed the Mad AI Scientist.
The answer isn’t architect or committee; it’s architect plus operating system. The real measure of an AI architect is what survives them. Documented workflows. A shared knowledge base the whole company reads from and writes back to. Standards and prompts embedded into SOPs and onboarding. Agents that run whether the architect is in the building or not.
I’ve also built this discipline into my personal practice: a knowledge infrastructure that gives every AI system I work with the full context of our business, so the intelligence compounds instead of living in one person’s head.
This is where the cross-functional structure belongs, and it’s a structure that executes, not a committee that deliberates. Each department head owns the AI goals for their function because they know where the friction lives. The architect owns the system that connects them: discovery department by department, prioritization by ROI, deployment one department at a time. Distributed ownership, single accountability. That’s how you get the durability Braden is asking for without the paralysis that’s usually produced by committees.
This Model Already Travels
I can tell you this structure isn’t channel-specific because I’m watching it work outside our industry.
A $100 million healthcare business with 40 locations, led by a genuinely forward-thinking CEO, has asked me to support them on three fronts: modernize their marketing with AI as fast as possible, the way we did at AIS; train their staff and management into an AI-forward operating model so they begin functioning as an AI business; and then start automating tasks and business functions, supplementing roles first, replacing where it makes sense. Different industry, identical problem, same structural answer.
I’m also supporting an executive at one of the largest dealers in our industry through his own AI journey and volunteering with a national governing body for an Olympic sport on the same challenge. And in conversations with CEOs watching the enterprise world, the pattern is unmistakable: large companies are automating redundant and administrative work at scale, and that wave is now moving down-market to businesses our size. The dealer channel isn’t special. It’s just late, and lateness is now a choice.
Your 30-day Assignment
Here’s the test. Within 30 days, name your architect, and give them the three filters. Have them bring one buildable project — not a pilot, but a project with an owner and a measurable outcome — to your leadership team.
If nobody in your organization owns this by Q4, understand that you’ve still made a decision — you’ve decided to be the laggard from my first article.
In the next installment, we get practical with platforms, large language models (LLMs) and what your team actually needs to understand to start building. AI 101, written for this channel.
The water is still rising. But I’ll leave you where I started: there’s never been a better time to learn. Pick up the tools.
