Services|01
AI agents
Software that does a defined first pass, then stops. A person still makes the decision that matters.
An agent is useful when the job is repetitive and the record already exists. Sorting a queue, drafting a reply in your voice, checking that last night's backup actually ran, reminding the person who owns the next step. That is the work. A chat box on a homepage is not.
We start with three questions: what is the job, who checks the output, and what happens when it is wrong. If those answers are vague, it is not ready, and we will say so before anyone pays for a prototype.
Agents sit next to the systems you already have, or inside software we build for you. They do not get the keys to email the whole company, and they do not invent an answer when the record is missing. Where a human has to see it, a human sees it.
This is custom work. The shape depends on the desk it is helping: a helpdesk, a faculty office, a sales inbox, a set of reports somebody still copies by hand. If you can describe the grind, we can usually take the first pass off that person.
What that includes
- A written scope: the job, the checker, and the failure case
- Drafting, sorting, reminding and checking — not unsupervised decisions
- Connected to the tools the team already uses
- A person in the loop wherever the output can cause harm or embarrassment
- Care afterwards, so the agent does not quietly rot when a process changes