"AI agent" has become a bit of a catch-all term, particularly when it comes to implementing AI within an organisation.
Here's how I break it down.
Start with an assistant
That's your everyday AI, the thing you open to think out loud, draft something, kick around an idea, or knock out a one-off task.
Then you've got agents
They do a specific job, like coding or project management, and they handle the things you do over and over again.
And there are two kinds of agents.
Personal agents work just for you. Things like generating your weekly report, sorting and drafting your emails, or prepping your CRM follow-ups. Quietly handling repetitive tasks in the background.
Shared agents are used across the team. Think of an agent that drafts sales proposals from your company content, one that triages and routes customer requests, or one that answers questions based on your policies.
Assistants and agents are only as good as the information they can easily access and use.
The short version
- Use an assistant to think and explore.
- Hand the repetitive work to agents. Make them personal when the job is yours, shared when it's the team's.
Then focus your organisational strategy on three things:
- Making sure the data access layer is clear, easy to use, and well governed.
- Ensuring everyone in the organisation has the skills to build simple agents for themselves.
- Putting each function, such as Marketing, Sales, Legal and Leadership, in charge of building and managing shared agents, both for use within their own function and for helping other teams self-serve.
