Everyone is talking about agents. What the word actually means, what an agent can safely do for a small business today, and what it can’t.
“Agent” has become the word of the year in AI. It gets used for everything from a chatbot to a fully automated department, which makes it hard to know what you are being offered.
A simple definition
A chatbot answers. An agent acts. It is an AI model given a goal, a set of tools and permission to use them in a loop: look at the situation, decide on a step, take it, check the result and continue until the job is done or it needs help.
The tools are what make it an agent. They might include searching your files, reading a calendar, filling in a form, sending a draft email or clicking through a website.
What agents do well today
- Multi-step lookups: “find every client whose documents expire this quarter and list what each one needs.”
- Preparing work for a person: drafting replies, filling in forms, assembling a report from several sources.
- Routine checks on a schedule: comparing orders to stock, bookings to staff, payments to invoices.
- Working through old software that has no other way in, carefully, one screen at a time.
What still needs a person
Anything irreversible: sending money, deleting records, messaging a client on your behalf for the first time.
Judgement calls: a refund for a regular, an exception to a policy, a sensitive conversation.
Anything where a mistake would go unnoticed. If nobody would catch an error, the agent shouldn’t be doing that task alone yet.
Give an agent a clear job, the fewest tools that will do it, and a person to report to.
How we approach agent projects
We start with a task that has a clear finish line and an easy way to check the result. The agent drafts and reports at first; it doesn’t send or change anything. Every run leaves a log. Once it has been right consistently, we let it take the next small step on its own.
That pace can feel slow next to the demos. It is also how agents end up trusted and used rather than switched off after the first surprise.
A quick test for any agent pitch
Ask three questions: what exactly can it touch, what does it do when it is unsure, and how will we see what it did? If the answers are vague, the agent is not ready for your business yet.
From the DYOL blog. Photo by Franck V. on Unsplash.
Keep reading
All posts
What’s new in AI this autumn, in plain language
Cheaper fast models, agents that run on a schedule and new rules on labelling AI. A short tour of what changed and what it means for a small team.

AI agents on a schedule: what it means for a small team
Agents can now wake up at 7 a.m., check something and report back. What that looks like in a café, a clinic or a club, and where to be careful.

How to ask AI for what you actually want
Better results from AI rarely need clever tricks. Five habits that help any business owner get useful drafts, summaries and answers the first time.
A free 20-minute conversation, then an honest assessment. You’ll know what it costs and how long it takes, in writing, at a fixed price.
Or email lloyd.delacruz@outlook.com. We reply within 24 hours.
