What does an AI consultant actually do?

The short answer

An AI consultant helps a business decide where AI should do real work, then makes it happen. They study how the business runs today. They pick the few places where AI will save the most time or money. Then they choose the tools, connect them to the systems the business already uses, and make sure people actually use them. A good one is measured by work that gets done, not by how clever the technology looks.

I answer this as an owner, not a vendor

I don't sell software. I came to AI from running businesses.

I started as a coder: I studied computer science at SFU and computer systems technology at BCIT, then took a software job in Kyoto. Then one bar conversation in Saigon turned me into an ad man. I co-founded Sofresh in 2007 with my friend Justin Cohen. We signed with WPP in 2012, and I ended up CEO of Wunderman Thompson Vietnam, 100+ people. Justin and I exited in 2021, after 14 years. Along the way I was a partner in a rooftop bar, Glow Skybar, and a restaurant, Urban Kitchen + Bar. Both were later sold.

An agency and a restaurant teach you the same thing: margins are thin, people are busy, and any new system has to be simple enough to use in the middle of a shift. So I look at AI the way an owner looks at a new hire. What does it cost? What does it take off someone's plate? Who checks its work?

Today I run my own businesses with a team of AI agents.

What I learned before AI was a job title

Clicks are not sales. In 2019 our agency built a Messenger chatbot for a global baby brand. People clicked far more than the target. Almost nobody used the coupon. The bot was fine. The leads behind it were not. Fix the data and the path to the sale before you tune the bot.
Sometimes the right call is less AI. On that same plan we used buttons first and kept language AI for later, because back then it couldn't handle Vietnamese well. Ship what works today, and plan the upgrade.
Bad data, bad AI. Clients asked us about AI in 2018, long before ChatGPT. Our honest answer was "fix your data first". It's still my first answer.
Most failures are handoff failures. After one project went badly, our own review said we were late and coordination was weak. The talent was fine. The handoffs weren't. AI agents break in the same place: where one step hands work to the next.
Protect people's time. When Sofresh passed 50 people, our makers got interrupted all day, so we added a traffic manager whose job was to say no. Good AI work does the same thing. It takes the interruptions, not the thinking.

What the work looks like, step by step

This is how I do it, from the first call to a system that runs every day:

Advice. I look at how each department works today: finance, HR, sales, marketing and operations. Then I tell you where AI will pay off first, and where it won't.
Architecture. Which tools and AI agents to use, how they connect to the software you already have, and who owns each part.
Design. The workflows, written step by step, so people and agents both know their part. This is where the handoffs get fixed.
Setup and implementation. I build it, test it with your team, and keep it running.
The team, if you need one. When a project needs engineers, I build the team in Vietnam and run it. I'm the orchestrator.

How I think about AI tools

I run AI tools the way I ran teams. I hire them, fire them, give them a place in the team, and check what they cost against what they save. A tool that doesn't earn its place gets replaced, the same way a role would be reviewed. Most of the job is not clever prompts. It's knowing the business well enough to point the tools at the right work.

AI consultant, AI agency or fractional Chief AI Officer?

The three names get mixed up. I ran an agency for 14 years, so I say this with respect:

An AI agency sells you a project. It builds what was agreed and hands it over.
An AI consultant sits on your side of the table. They advise, design and often build, and they care whether your team still uses the work after launch.
A fractional Chief AI Officer is an AI consultant who stays on as a part-time leader. They own the company's AI work over time, not only one project. See Fractional Chief AI Officer, in Vancouver and Ho Chi Minh City.

If you're a Western company thinking about building your AI team in Vietnam, read Building an AI team in Vietnam.

I'm based in Ho Chi Minh City and downtown Vancouver. I meet clients in person in both cities and fly to Vancouver often. I work in English and Vietnamese.

Questions owners ask

Do I need an AI consultant, or can my team figure it out? If someone on your team has the time and owns the result, start there. Bring in a consultant when nobody's job is to make AI work, or when tools were bought and nobody uses them.

Do we need technical staff first? No. Part of the job is deciding what you need. If the project needs engineers, I build the team.

Can you work with the software we already use? Yes. The architecture starts from the systems you already have.

Do you meet in person? Yes. I meet clients in person in downtown Vancouver and in Ho Chi Minh City, and work online in between.

How do we start? A first call. I ask how your business runs today and tell you plainly where AI fits and where it doesn't.

Why do AI projects fail? In my experience, most fail at the handoffs, not the technology. One of our project reviews at the agency found weak coordination between teams, not weak talent. AI agents trip at the same spot: the moment one step hands work to the next. The other cause I see most is bad data. Fix the data and the handoffs first, then tune the tools.

Why doesn't my team use the AI tools we bought? Usually because nobody's job is to make them work. A tool needs one clear task, a place in the workflow and a person who checks it. Without those, it sits unused, however clever it is.

Book a call · letstalk@lyvietvu.com