A health clinic reached out to me last year, wanting to implement AI automation across their operations. Scheduling, patient intake, documentation — a lot of moving parts.
My first instinct wasn't to propose. It was to say: before I can tell you what to build, I need to see how you work.
So that's what we did.
The Observe-First Model
For any complex AI project — one with multiple departments, multiple tools, and real workflow dependencies — I use a flat-fee on-site day before I scope anything.
Here's what that day looks like:
I show up with no slides and no pitch. I walk their floor. I watch how things run. I sit with the team and ask questions — not the kind designed to identify upsell opportunities, but the kind designed to understand what's actually happening. I audit the tech stack they're actually using versus the one they think they're using (often different). I find out what breaks, what people do manually even though a tool supposedly handles it, and where the friction actually lives.
At the end, I deliver a report.
Not a sales deck. An actual report — specific recommendations, prioritized by impact, with enough detail that another consultant could implement them without ever talking to me.
The Counter-Intuitive Part
You might think that's a bad business decision. Why would you hand someone a full implementation roadmap?
Here's what I've found: they never use it to go hire someone else.
And it's not because the report is secretly vague or because I'm holding back the good stuff. The report is genuinely complete.
But the report isn't what they're buying. They're buying the context that only exists because I was there. I sat with their front desk coordinator. I watched the doctor flip between three different systems during a patient handoff. I know which team member is resistant and which one is secretly doing workarounds that no one else knows about.
That kind of understanding doesn't transfer in a document. The report proves I have it — but it doesn't hand it to the next consultant. So there is no next consultant.
Why This Works Better Than Proposals
The standard consulting model goes like this: prospect describes their problem, you propose a solution, you negotiate, you start.
The problem is that “prospect describes their problem” and “what's actually happening” are usually different things. Not because clients lie — but because nobody has perfect visibility into their own operations. They tell you what they think is the issue. The real issue often lives one layer underneath.
A proposal based on their self-diagnosis is a proposal based on incomplete information. Which is fine for simple projects. But for complex AI implementations, you're stacking automations on top of existing workflows. If you misread the workflow, you build the wrong thing.
The observe-first model inverts this. You understand first, then scope. The proposal that comes after a day on-site is sharper, more accurate, and usually more trusted — because the client watched you do the work of understanding their business before you told them what it needs.
On Pricing the Observation Day
The flat-fee observation day should feel like a fair trade for both sides. The client pays for your time and expertise. You get the access and context you need to propose something real.
I've found that clients who won't pay for an observation day often aren't ready for a full engagement anyway. They want a free assessment followed by a proposal they can use to compare vendors. That's a different kind of client — and not a bad one — but the observe-first model isn't designed for them.
The clients who engage with this model are usually the ones who already know their problem is complex and want someone who's going to take it seriously. Charging for the day signals that you do.
What I Took From Watching Good Leaders
The best business leaders I've observed don't make decisions from the conference room. They go see for themselves.
I heard about how the Whole Foods CEO would spend hours in stores, talking to employees and customers, before making decisions. Not because store visits were required — because firsthand observation changes what you prioritize.
Same thing applies in consulting. The days I've produced the sharpest recommendations are the days I didn't come in with assumptions to confirm. I came in with questions and let the operation tell me what it needed.
Propose after you've seen it. Every time.
Thanh Pham is the founder of Asian Efficiency and an AI consultant based in Austin, TX. If you're building an AI consulting practice, the 4-Day AI Sprint covers the foundations.
