Rebootwe make IT work

The 100x question

David Mariani · Reboot · July 2026

An owner asks whether their firm should be using AI. It comes up most weeks now, and I have stopped reaching for a careful answer. I reach for a marker and write two things on the whiteboard:

10x. 100x.

Which one do you want to be? That is the real conversation, and it runs short on hype and long on arithmetic.

It was recommending rocks a minute ago

Pay the skeptics their due first: they were right. Not long ago the internet was passing around screenshots of an AI recommending rocks for dinner, and the whole category felt like autocorrect with confidence issues, hallucinating politely in complete sentences. If you laughed, you were right to laugh. I laughed, and I am not a man who laughs at computers lightly.

What has happened since is the part I keep failing to say big enough. I will not hand you a benchmark; I do not trust them much, and any number I typed here would be stale before you read it. I can tell you what changed at my own desk. It argues back now. It catches the hole in a plan at midnight and is still awake at six when I want to argue about the hole. Somewhere in the last couple of years it turned into a real thought partner, and the speed of that trip is honestly hard to fathom. I do this for a living and I am still catching up to it.

One hundred times zero

Now the catch, and the catch is the whole game.

The tool is a multiplier. It takes what you walk in with and it scales that. Walk in with twenty years of your field and it hands back sharper drafts, angles you had not weighed, the tedious half of your thinking finished while you were in a meeting. Walk in with nothing and it scales that instead, at the same speed, in the same confident voice. A multiplier holds no opinion about what it multiplies. One hundred times zero is still zero. It is just a much faster zero than you could have produced on your own.

That is why the prerequisite is so unfashionable: domain understanding. You have to know the field you are asking about, because what this thing augments is judgment, and the judgment has to already be in the room. Watch a veteran CPA point it at a client letter. She smells the wrong sentence before she has consciously read it; thirty tax seasons live in that reflex. Drafted in seconds, judged in seconds, letter out the door in four minutes instead of forty. Hand the same tool to someone with zero tax seasons and it produces a letter every bit as fluent, and nobody at that desk can smell anything. Same tool. Different multiplicand.

Point it at your strong suit first

Here is the advice that surprises people: start where you are already the expert. The instinct runs the other way. Owners aim AI at whatever they are worst at, hoping it will cover the gap, and in the weak field you cannot grade its answers, and an answer you cannot grade is a liability with excellent posture. In your strongest work you catch its mistakes in seconds and you notice when it is genuinely brilliant. The seams show up fast when you are standing on solid ground.

The 10x arrives early there. The 100x comes later, once the seams are familiar and you trust it with bigger pieces of the work you know best, checking faster each round. The distance between those two numbers is a skill, and the skill is yours. Nobody sells it by the seat.

We answer it too

Fair is fair: the question points back at us, and it should. The tools are in this shop the way they are probably headed into yours, drafting and doing first passes on the tedious parts. Two things hold still while all of that moves: a person checks every claim before it reaches a client, and the person who checked it is the one who signs it. The curating is real work, and we own every frame of it.

The owner at the whiteboard eventually asks what everyone asks next: which tool, and how much a month. Reasonable questions with perishable answers; whatever I wrote here would be wrong by autumn. The numbers are the part that keeps. If you are weighing what this looks like in your own shop, book a time and bring the work you know best.

When the conversation winds down I erase the arrows and the tool names somebody always calls out. The two numbers stay up.

Book a 15-minute call →