Operating Notes

How I run my work with AI agents.

Four standing agents. Eight scheduled runs a week. Nothing publishes without me.

Last revised July 2026


The short answer

I run a small set of scheduled AI agents that do defined jobs on a defined cadence and hand me drafts. They do not act on my behalf. They do not send mail, publish posts, or contact anyone. They read, filter, draft, and stop.

The interesting part of this system is not what it does. It is what I took out of it.

The operating principle

An agent earns its slot by removing a decision I would otherwise have to make, or a search I would otherwise have to run. If it only produces output I never act on, it is not automation. It is noise with a schedule attached.


The four agents

01

Opportunity Desk

Weekday mornings, 8:00 ET

Sweeps one inbox across three tracks at once: roles, teaching openings, and consulting or referral signals. Filters postings against a hard test for level, function, and location, caps the list at five so it cannot pad itself, and separately flags threads going cold, where I owe someone a reply or someone owes me one.

Output: one email. Who needs a reply today, what is going cold, up to five things worth a look. On a quiet day it says so in two lines and stops.

02

Content Desk

Mondays, 9:30 ET

One idea per week, drafted once, adapted across channels. It does not work from a theme rotation. It looks at what actually happened in the last seven days and builds from a real specific, alternating between a professional audience and the families and referral partners my estate services business serves.

Output: one long piece for the site, a compressed version for LinkedIn, and a short social adaptation only where it fits. One idea, not one piece per channel.

03

Blue Moon Desk

Thursdays, 8:00 ET · conditional

Checks whether there is an estate sale on the board in the next fourteen days. If there is, it drafts the full weekend promotion sequence. If not, it reports the last sale date and any unanswered inquiry, and stops. On the first run of each month it appends a short audit of what a customer would actually encounter: search visibility, unanswered reviews, anything broken that would cost a lead.

Output: a full campaign when there is something to promote, a two-line note when there is not, plus one monthly check folded into the same email.

04

Dissertation Desk

Sunday evenings, 5:00 ET

The doctorate is the thing most likely to get displaced by work that feels more urgent, so it gets the only agent whose job is protection rather than production. It surfaces committee mail and deadlines, then asks the same three questions every week: what moved, what is blocking the next step, and what the one real work session before next Sunday will be.

Output: under fifteen lines. No research, no drafting, no encouragement. The value is the standing question.


What is deliberately not on a schedule

The distinction that keeps this small is between a scheduled agent and a capability. A scheduled agent is for work that has to happen on a cadence whether or not I remember it. A capability is something I invoke when I need it, and it costs nothing the rest of the time.

Client research before a walkthrough, proposal drafting, settlement reporting, contract review, CRM cleanup, literature work on the dissertation. All of these are real and all of them run on AI. None of them are on a schedule, because none of them need to fire while I am asleep. Confusing capability with cadence is how four good ideas become thirty-seven runs a week.

The test

Before anything goes on a schedule: would this be worse if I only ran it when I thought of it? If the honest answer is no, it is a capability, not an agent.


What I removed, and why that is the point

I built sixteen of these. At peak they ran roughly thirty-seven times a week. Some of it was genuinely useful. A lot of it was me automating my own anxiety.

37

Scheduled runs per week at peak, across 13 recurring agents

8

Scheduled runs per week now, across 4 agents

78%

Reduction in scheduled compute, with no loss of anything I was acting on

The audit was simple. For each agent I asked one question: in the last month, what did I actually do differently because this ran? Where the honest answer was nothing, it went.

Reddit lead scanner · ran 14x/week, could not reach the source
Three separate daily job monitors · all reading the same inbox
Weekly blog engine · drafts outpacing my capacity to publish
Weekly valuation drill · a good idea, wrong season
Monday content planner · built for a growth posture I am not in
Monthly market brief · interesting, never actionable
Friday reputation check · folded into the weekly desk
Direct board sweep · duplicated the daily desk
Monthly Blue Moon scorecard · merged into one Blue Moon agent
The slot they freed · went to the dissertation, which had none

The design rules I hold these to

What this actually demonstrates

Anyone can stand up sixteen agents. The discipline that matters, and the one most organizations skip, is measuring what came back and being willing to turn off the things that did not earn their place. Adoption is not the hard part. Restraint is.


Paul Tonden · paul@houseofpraxis.studio · linkedin.com/in/paultonden