AI automation with a person in the loop

Work that comes back the same way every week: putting a newsletter together, sorting enquiries, moving data from one system to another, preparing copy for publication. That work is predictable enough to hand to a system, and it costs your team hours that are better spent elsewhere.

We do not start with the technology but with the process. Who does it now, how often, what goes wrong and what it costs. Only once that is clear do we decide whether a simple automation will do or whether an AI agent is needed.

An agentic workflow is a sequence of steps an AI agent carries out on its own: gathering information, drafting, checking against your rules, getting it ready. That goes further than a fixed automation, because the agent decides the next step case by case.

That is exactly why there is always a person in between. Everything that goes out, an email, a post, a message to a customer, is reviewed first. A system that publishes unchecked is a risk, not a saving.

Your brand and your quality rules live in the system itself, not in someone's head. You can follow what it does, you get an alert when something goes wrong and you can pause it at any time.

After delivery we walk your team through it and document how it works. After the first month we evaluate together: what it does, where it needs adjusting and what could be added.

Our own work runs the same way. The articles in our knowledge base and our podcast go through a system of agents that research, write and check, and nothing goes live without a person's approval. What we build for clients, we use every day ourselves.