The agents · David Galvis

I came up through business. Then I learned to build.

My background does not start on the technical side. It starts in business, marketing and sales, which is where I learned what an hour of the wrong work costs. The technical part came after, and it came because I needed it: automating something badly is worse than not automating it, so I had to understand what I was building to make it effective and safe. Then I built a skill that builds the rest, so a new one takes an afternoon instead of a quarter.

What comes back is time. Hours for photography, for the people I want to see, for the projects I take on because I want to be in them. And inside the work at Digitech the question turns into the same one with more riding on it: how much more can we get through, how much faster, without the work getting worse.

Six of them, with the lid off.

What I mean by agent

The word moves depending on who is using it, and the definitions are still being argued. Anthropic draws the line at systems where the model, in their words, “dynamically direct their own processes and tool usage” instead of running a path somebody wired in advance.

Mine is narrower and it comes out of use: something that takes a piece of my week without me in the room, leaves its output where the next one goes looking for it, and gets better because every correction I make goes into the file it reads before it starts.

These six are some of the ones I run. Open one and you get what it takes off my hands, how it works, and the time it gives back.

The one that finds what we have in common.

Solves: A message worth reading meant twenty minutes inside a profile first, so I sent fewer.
How: Reads their profile, crosses it against the company’s numbers, and comes back with the actual overlap: that we both live in Massachusetts, that they just moved into a market I know. I write from there. When there is nothing in common it says so.

Hands off to

pipeline / new row
me / I write it from there

~20 minper message

The one that analyses your competitors.

Solves: Everyone can see what a competitor posts. Working out which part of the market they are covering is the slow part.
How: Reads their campaigns, their web and their positioning, and comes back with what they are covering and what they are leaving open. Pointed at a different question it runs a different analysis. The hard part was never the agent: it is finding data worth analysing.

Hands off to

brand / the angle to take
me / what we do differently

~1 dayper competitor

The one that learns a brand by looking at it.

Solves: Every new piece restarted the same argument about type, colour and voice.
How: Takes the company’s existing assets, works out the system underneath, and builds decks and campaign material against it.

Hands off to

pages / ready to ship
me / the call on tone

~2 daysper brand system

One page, then every format.

Solves: A deck goes stale the day it is exported, and nobody reads the second version.
How: Reports and documents got easy once they stopped being files. It writes the thing once in HTML and the PDF, the deck or the document come out of that, in a fraction of the time it used to take. The ceiling is the data: with a knowledge base worth reading behind it, this is the fastest part of my week.

Hands off to

me / one link
archive / filed by itself

~3 hoursper deliverable

The one that does the research.

Solves: Walking into an enterprise conversation on a website and a funding headline.
How: Pulls filings, financials and news from several databases at once, so the due diligence is done before I sit down. It weighs the sources against each other, assumes nothing, and tells me where they disagree. The call is still mine.

Hands off to

prospecting / the angle
me / before the call

~2 hoursper account

A CRM built out of spreadsheets.

Solves: Every platform wanted me to feed it, and a CRM nobody updates is worse than a list.
How: Spreadsheets, CSVs and Claude wired into the channels the deals run through. It reconciles itself against LinkedIn exports.

Hands off to

me / the week’s queue
brand / who needs material

~1 houra week in fields

One chain, one knowledge base.

None of them talk to each other. They talk to the same folder. Each one writes what it found where the next one is going to look for it, and every correction I make goes into the file they all open before they start.

That part took the longest, and it is the whole difference between owning six tools and running one system.

Claude CodeCodexChatGPTApollo.io ApifySales NavigatorTalent Insights NetSuiteLooker StudioBoomiMuleSoft

The same setup, outside the job.

This is the week nobody pays for, and it is the clearest thing I can say about what they actually are.

5:30 am

The brief. Calendar, inbox and everything that changed on disk overnight, waiting on one page when I sit down.

8:30 am

The inbox, already answered. It works out where each thread stands and leaves the reply written. I approve it or change a line.

12:00 pm

Scheduling. I say who and roughly when. It finds the slot and writes the invitation.

5:30 pm

The close. It reconstructs where the hours went from timestamps and commits, then asks only about the gaps it cannot account for.

What I hold to when I build one.

San Francisco makes one thing easy: the models change every few weeks and somebody at the next table has already run the new one. These four survived that.

Pattern recognition is where the model earns its keep.

It finds the shape in two hundred profiles faster than I can. Choosing which shape matters stays with me.

The loop beats the prompt.

Chasing the perfect wording stopped paying. The run fails, the failure is cheap to catch, and the fix goes into the file it opens next time.

A claim with no file behind it does not ship.

They hallucinate hardest on what sounds routine. Every number carries the route to check it, and Python does the arithmetic, not the model.

Autonomy is priced by the cost of being wrong.

Reading, drafting and filing run unattended. Anything that leaves my hands is written by the agent and sent by me.