Three Versions of Every Workflow
Inside every company, three versions of reality exist at the same time. There is how leadership thinks people work. There is how people say they work. And there is how they actually work.
The gap between those three is where most AI projects die.
I learned this the hard way. I asked a team member to write down every step of his daily workflow. The first version he gave me was clean, logical, maybe ten steps. It was the version he thought I wanted to see. I pushed him to include everything, including the steps he considered too small to mention. The second version had over a dozen steps, and the ones he left out initially turned out to be the biggest automation opportunities. Opening a browser, pulling up LinkedIn, copying data between tabs, formatting an internal email from scratch every single time. None of that made his first draft because he didn’t think it mattered.
It mattered more than anything else on the list.
Why Companies Buy AI for the Wrong Reasons
The honest reason most companies adopt AI right now is fear. Fear of falling behind. Fear that competitors are doing something they are not. Fear that the board will ask about their AI strategy and they will not have an answer.
That fear leads to impulsive spending. Teams sign up for tools before understanding what problem they are solving. They pay AI to replace humans rather than making the humans they already have ten times more effective.
I intentionally set out to increase my AI spending at one point. It turned out our costs barely moved because we were using exactly what we needed, paying for exactly what delivered value, and automating the rest with internal capability. The companies overspending on AI are the ones who skipped the documentation step.
What “Using AI” Actually Means
Ask most people if they use AI and they will say yes. What they mean is they ask a chatbot questions. They treat it like a search engine that writes paragraphs. That is not integration.
Here is what integration looks like. One of my team members types a single word into a chat channel and five things happen automatically: an opportunity is created in the CRM, an internal email goes out to the right sales managers with the right people copied, context about the prospect is pulled and attached, a confirmation goes to the client, and a follow-up is drafted. What used to take 35 minutes of admin work per meeting happens in seconds. He does not need to understand how any of it works. He just types one word.
Asking a chatbot for help is assistance. Building it into the workflow is infrastructure. Most companies are still on assistance.
If It Requires Training, You Built It Wrong
One of the people on my team is over 55. He does not use spreadsheets. He was not going to learn a new tech stack, and asking him to would have been a waste of both our time.
When he first saw what we were building under the hood, he was frustrated. It looked complicated. But once the system was finished and all he had to do was type into the same chat app he already used every day, he was on board immediately. Not because we trained him. Because we removed every barrier between him and the output.
AI changes too fast for training to hold. The system I presented last week is already being replaced by a better version. My technical team and I update constantly. Teaching people the mechanics of something that will be different next month is wasted effort. Build the interface simple enough that tech literacy becomes irrelevant.
Start With One Person, Not a Department
The instinct is to roll out AI across a team with a training deck and a timeline. That almost never works.
I started with one person. The one who was already performing well and willing to go deep on documenting his process. We captured his real workflow, automated the repetitive parts, and gave him back hours of his week. He did not feel like something was imposed on him. He felt like his own system got upgraded. He owned it because it was built around how he already worked.
The rest of the team saw the results and asked for the same thing. Nobody had to be convinced. The demand came from them.
Build for one. Let the proof spread. Let people pull the system toward themselves instead of pushing it on them.
Same Department, Different Workflows
Here is what you find when you document an entire department at the individual level: people who supposedly do the same job do it differently. They follow the same steps in different order. They skip steps others consider essential. They have workarounds nobody else knows about.
Comparing those differences reveals the actual best process. Not the one in the handbook. The one that produces results.
Once you have that, you can build automation around the optimum workflow and let each person interact with it in whatever order makes sense to them. The system adapts to humans, not the other way around.
The Work AI Cannot Do For You
There are two kinds of work that operators avoid. The first kind is repetitive and time-consuming. Automate that. The second kind is emotionally uncomfortable. That one stays with you.
Setting up accountability systems. Entering goals for each team lead. Running weekly meetings where people have to explain their numbers. Creating consequences when targets are missed. Most operators skip this work because they would rather build the next thing. I did the same for years.
The connection to workflow automation is direct. When repetitive tasks eat your day, you have a convenient excuse to avoid the managerial work. Once those tasks are automated, the excuse disappears. You have the hours. You have the bandwidth. The only thing left is whether you are willing to do the part of leadership that nobody enjoys.
AI handles admin. It does not hold your team accountable. It does not run a difficult conversation. It does not sit across from someone and explain why their numbers are not where they need to be.
The First Step
If you read this and want to start tomorrow, here is what to do.
Pick one department. Document the workflow of every person in it, at the individual level. Do not hand them a template and ask them to fill it in. Sit with them. Push past the first clean version they give you. Get to the real steps, including the ones they think are too small or too obvious to mention.
Do not assume your published internal workflows are being followed. Do not accept the version people think you want to hear. Get a snapshot of where the company actually is. Not where it thinks it is. Not where it should be.
That snapshot is your blueprint. Every AI tool, every automation, every system you build after that point has a foundation.
Companies that skip this step usually end up with more tools and the same problems. The ones that get it right understood the work before they tried to automate it.
0 Comments