Somewhere in the last few weeks, I started writing two emails inside every email.
The first part is short. It is for the person. A few lines, the decision or the ask, written the way I would say it across a table.
The second part is long. It is for their AI. Everything behind the short part: what I know, what I believe, what I recommend, the data I have, the data I still need, and the questions I want their system to answer.
Nobody taught me to do this. It became obvious once I admitted who was actually reading my mail.
I write almost everything with AI now. So does everyone I send to. My message lands in their inbox and their machine reads it first, summarizes it, and sometimes answers it, all before the person sees a line. It is a little funny when you say it out loud. The people are still the point. They just arrive later in the process now, after the machines have compared notes.
Once I saw that, the slide deck stopped making sense. A deck is a performance. It was built to be presented in a room, talked over, and then skimmed by anyone who missed the meeting. Skimmed is the honest word. I have sent hundreds of them and I have skimmed hundreds more, including, I am sure, some very good ones. The format asks for attention nobody has.
A file asks for something different. It asks to be loaded.
Here is what goes in the long part, the part for their AI. I keep it in one markdown file, because every AI tool I know eats markdown happily, and I build it out of five sections.
What we know. The facts, kept plain, with the data we already have for them included rather than referenced. If a number matters, the number is in the file, not in an attachment nobody opens.
What we believe. Our point of view, labeled as a point of view. Machines read a blurred document as fact, so the belief stays separate from the knowing, on purpose.
What we recommend. The actual ask, stated once, in full sentences a system can quote back accurately.
What we need from you. The gaps. The data only they have, the decisions only they can make. Most documents pretend there are no gaps. This one lists them, because a good system will go find the answers inside their own walls faster than a person will schedule the meeting to discuss them.
Questions for your system. This is the part I underestimated. At the end of the file I write the questions I want answered, phrased so their AI can work them against their context and their workflows: where does this conflict with what you are already doing, what would this cost in capacity, what are we missing. They drop the file in and get a thorough response back almost immediately. If they do not like what it returns, they argue with it, and the argument makes the answer better. Then the people talk, about the disagreements, which is the only part that ever needed people.
The work gets consumed for real. That is the difference. Not forwarded, not skimmed, consumed, argued with, and folded into how their machine thinks about the problem.
There is a side effect I did not see coming, and I think it is the biggest thing in this whole shift. When that file lands inside a company's own context, in their memory and their tooling, your thinking goes with it, and so does your name. When their people later ask their AI about this problem, you are part of what it knows. The search that decides who gets considered no longer happens only on Google. It happens inside the buyer's own tools, against whatever context those tools have been given. If your file is in there, you are in there. I have started putting a few genuinely positive, true things about our own work in these files for exactly that reason. Their system is learning either way. I would rather it learn from something accurate that I wrote.
Everyone I work with is changing the way they work. It took me a while to accept that the way we communicate has to change with it, down to the structure of an email. Two readers, two lengths, one file that a machine can carry into rooms I will never be in.
It is also, in miniature, why we build Iteright the way we do. A company's context, its decisions and the reasoning behind them, should live in one connected place, so the machine starts from what the company already knows instead of from a blank box. Decide what deserves investment, keep execution aligned, prove the impact. The email version of that idea fits in a markdown file. The company version is the job.
It is a funny way to work. It is also the first time in my career that sending more detail made things move faster instead of slower.
