There is an idea from Andrew Ng that I haven't been able to get out of my head: artificial intelligence does not
It is a product that is purchased, it is a resource. Like electricity.
It is worth stopping there, because even the verb changes.
No one ever bought electricity. No one invested in electricity either. What one does with the
electricity means making it available: it brings it into the house, brings it into the workshop, and from that day on, it's there.
The appliances come later. The carpenter bought a light bulb so he could work two hours more
when it had already grown dark. Later he bought a saw, to cut in a little while what previously
It took him half an afternoon. What he sold were chairs. Electricity only allowed him to make
More.
Your business is chairs.
And notice that this still happens today in your office, without you even thinking about it. The only thing you need is
Let there be light. From then on, every appliance decides what to do with it: the fridge cools, the
air conditioning cools differently, the printer prints. Neither asked for your permission nor did you
forced to understand what goes through the cable. You enabled the resource once, and ever since you
You need the devices.
And it's not that people don't want to. Everyone wants to be in on this. What they ask me is
«how do I get in?», «What should I buy?», «What should I invest in?».
All three are the same question, and all three take for granted that there is something on a shelf waiting for
It's as if at the time someone had asked what to buy to have electricity.
There was nothing to buy. It had to be prepared, and then the appliances would come.
And the answer to when it is uncomfortable but simple: it had to be enabled last year, it is possible
enable next week, and from then on it stays.
Now, enabling it in a company is not the same as enabling it on your phone, and that's where
the industry is agreeing: form is agents. That's where everything moved
the world at almost the same time. They are still being built. I've been doing it for a good while
building them, so I know why sometimes they promise quite a bit more than they deliver: between one
there is a much greater distance from a showy demonstration to something that can survive any given Tuesday
what it seems. But they work now, and they work today.
And there is something that took me a while to understand: opening that valve is precisely what teaches you to
See. The carpenter didn't design the two extra hours before having light; it was the light that gave them to him.
It showed. The same thing happens with this. You start to see what can be built with a little bit of AI.
only after having it turned on. But really turned on: in something small and at the same time everyday,
within a process that you actually care about. If you put it to work on something you only touch once in a
when you aren't going to use it enough for it to teach you anything. Nobody arrives with the list
done, and waiting to have it is the surest way to never begin.
What is true is that the day you want an agent to take a job off your hands—from
truth, not demonstration— you're going to have to know how to tell which is which. And that's where technology doesn't help.
Because one thing is a task and another thing is a job. «Answer me these five questions» is
a task: you ask for it, they deliver it, it's over. «Take care of serving the customers who
they write on WhatsApp on Saturdays» It is a job: it has an owner, it has a before and after, and
Can someone tell me if it was done well?.
An agent is given jobs.
And naming a job accurately costs quite a bit more than it seems. Who does it today. What
input: a message, an email, the photo of a handwritten list. What comes out: a quote, a
order loaded, a timely response. And the one almost no one has answered — what does it mean,
exactly, that that job was well done.
You don't have it answered because you never needed to. The person who does that job in your
business knows it, it learned it over the years, and with that it was enough. The day you want that
work is done by an agent, that clarity is no longer optional.
That is where the real investment is: in knowing how to say, in words, what you want to happen. If
If you skip it, whatever you bring after that is going to fail, no matter whose it is or how good it is.
So what I'm proposing has two parts, and neither is technical. The first:
enable. Turn on the faucet and put it to work on something small but everyday — one of those
processes that if they stop, you notice. The second: while you use it, look at your operation and choose one
job. Only one — the one
costing you the most, or taking away the most nights. Write it on a napkin: who
what it makes, what goes in, what comes out, how you know it turned out well.
And then go find an agent who can do it, no matter whose it is.
Because you are not investing in a new technology. You are enabling a new way of
work.
And the mistake is not choosing the wrong job — that fixes itself, you choose another one and move on. The mistake of
truth is the other one, the one that is not seen: to conclude that AI is not for you, or that it is not for your
business. That's like keeping candles in the age of electricity. With a calculator in the era of
Excel. With fax in the WhatsApp era.
And you, what job would you give him?
Daniel — your trusted human.
If you were left wanting to see what that looks like in action, here is the door:
I want a walkthrough of what Max can do for my business