Automate / agentic workflows

What is an agentic workflow?

An automation is a row of dominoes. One thing happens, the next one falls, the same way every time. An agentic workflow puts one step in the middle that can read what came in, look things up, and decide what happens next. Pick a job below and watch it run.

The 30-second version / sound on, captions built in

The interactive version needs JavaScript. Here is one run, written out.

  1. A customer writes in through the website. That is the trigger.
  2. An AI agent reads the message and checks the instructions it was given.
  3. It looks up what it needs: the price list, the calendar, the customer's history.
  4. It decides: reply on its own, ask the owner first, or hand it to a person.
  5. Whatever it did is saved to the customer's record.

Want one like this? Ask me →

Four parts.
One of them thinks.

A trigger

Something happens. A form is submitted, an email lands, the clock hits five. That starts the run.

An agent

The step that thinks. It reads what came in and works toward the goal you gave it, in whatever order this case calls for.

Rules and tools

Plain-English instructions, the kind you would give a new hire, and a short list of things it is allowed to touch. Nothing else.

An outcome

What it is allowed to do at the end: reply, update, print, or stop and ask you.

Most jobs don't
need an agent.

If the steps never change, use plain automation. It is cheaper, it is faster, and it never surprises you. An agent earns its keep where the input is messy: email written by people, PDFs, web pages, a customer who is upset.

I build both, and I'll tell you which one the job needs. I run agents on my own businesses before I sell one. The morning briefing above is real: mine prints on a thermal label printer at 5 a.m.

What would you
hand off?

Tell me the job you keep doing by hand: the inbox, the quotes, the listings, the Monday report. I'll tell you straight whether an agent is the right tool, and build it if it is.

Ask me about a workflow →