Blog

How we use AI at AnAIHelp

Will Burke ·

We’re a small team, and we run our business the way we set things up for clients. Here is what that looks like today.

Shared memory

Everyone on the team works with AI assistants, and not all the same ones. We fit the model to the task. The trouble is that, by default, each new chat only sees the context specific to that session. Using Project features helps, but it’s still isolated by user. Instead, we keep one shared record of the business: what’s decided, what’s in progress, and what’s next. Every person and every AI agent starts from that same record. It doesn’t matter who did the work last or which tool they used.

A person approves what goes in

Agents write up what they did as they work. Those notes don’t become part of the official record until one of us has reviewed and approved them. If an agent gets something wrong, it’s caught before every later session repeats it.

Bots keep the CRM current

Our CRM tracks the companies and people we talk to. Bots handle its structure and its updates: adding records, logging each contact, and keeping the fields consistent. We spend our time on the conversations.

Research, email, and the daily digest

Agents watch our market for us: industry news, trends, and what people in the business are talking about. They flag what’s worth reading. They also watch our email so nothing falls through the cracks, and they help coordinate schedules. Each day the team gets a digest of project status, what changed, and what’s blocked.

Monitoring with a second reader

Standard monitoring software watches our systems and sends an alert when something crosses a threshold. That part isn’t AI. On top of it, an AI agent reads the logs and health data to spot trouble before it trips an alert. Infrastructure updates run on a schedule, and the agent watches for problems after each update.

The pattern

It’s the same each time. Ordinary software does the routine work, AI does the reading and the judgment calls, and a person approves anything that matters.

It’s also what we recommend to clients: scripting, automation, and existing tools first, with AI models where they add value.