- Sector
- Technology consultancy
- Company size
- Our own company
- Business lines
- AI and automation
- Solutions
- AI agents, Corporate assistant, Safe AI and training
A small team with the management workload of a large company
SolveIT runs dozens of projects at once. Accounting, tracking each project and the same old questions were eating up hours of the people who were most needed elsewhere.
Before proposing AI agents to a client, we wanted to have them working in-house, with our own data and our own rules.
- Invoices, bank accounts and due dates that depended on someone remembering
- Projects with no weekly snapshot of how each one was going
- Knowledge scattered across manuals, emails and the memory of a few people
Three agents, each with its own role
Whatever an agent writes is signed as its own: it can always be told apart from what a person did.
Accounting
Every night it reviews invoices, bank accounts and due dates inside Odoo and leaves a summary for the morning. It records supplier invoices from the PDF, reconciles transactions and prepares the tax returns. It doesn't pay or file anything: a person signs that off.
Project coordination
Every week it prepares the project portfolio with a written health criterion and a batch of decisions for management, seven at most. It asks the person in charge in the project task and follows each action through until it's closed.
Company knowledge
Answers the team's questions with what SolveIT already has in writing: procedures, manuals and the history of projects and support.
The rules they work by
The AI model can change from one month to the next. What makes us trust the agents is their rules:
- A written role. Each agent has its mission, its duties and a list of what it can do on its own, like anyone joining the company.
- Money and people are off limits. No agent pays, deletes or makes decisions about anyone on the team. It gets things ready and alerts whoever decides.
- In parallel first. Before being handed a task, the agent does it alongside the person for a while and the results are compared.
- Rules for using AI. Which tools are approved and what data can be given to them is set out in a policy the whole team knows.
It's the same thing we set up for clients: AI agents on their systems, an assistant with the company's knowledge and usage rules and training.
Work that no longer depends on someone remembering
The accounts are kept up to date inside Odoo, projects have someone coordinating them and questions are answered with what the company already knows. Each agent does only what it has been approved to do, and anything involving money or people is decided by someone on the team.
What work in your company could an agent take on?
Bring us a process that repeats and that depends on one person today. Together we'll see whether an agent can take it on, and within what limits.