Skip to content

What is an AI agent?

A digital colleague you hand a whole task to, not just a question. It understands the goal, works in your software and asks for your approval when it should.

Definition

An AI agent is a program that is given a goal, works out the steps itself to reach it and carries them out in your tools, within the limits you have set.

  • It understands

    An instruction in plain English, a freely written email, a scanned invoice.

  • It acts

    It looks things up, compares, writes, files and records, directly in your software.

  • It reports back

    Every step is logged, and anything that commits the company waits for your approval.

What an agent is made of

On its own, an AI model can read, reason and write. To turn it into an agent, you give it four things, the way you would prepare for a new colleague.

  1. A role

    What is expected of it, written like a brief: “Chase customers who are late paying, tactfully.”

  2. Knowledge

    The documents it finds its answers in: terms of sale, procedures, the history of exchanges.

  3. Tools

    The software it reads and writes in: accounting, banking, CRM, email.

  4. Rules

    What it may do on its own, and what must wait for a person to approve.

It observes, decides, acts, checks

Conventional software runs a script written in advance. An agent looks at the situation, chooses the next step and starts again, until the work is done.

For example, when a supplier invoice arrives by email:

  1. ObserveAn email from Atelier Martin arrives, a PDF invoice attached.
  2. DecideIt is a supplier invoice: it has to be matched with its order.
  3. ActIt finds the purchase order in Sage and prepares the entry.
  4. CheckThe amounts match: the entry waits for your approval.

Agents that work together and take the initiative

  • Automate repetitive work

    Data entry, reminders, documents, reporting: agents take on the recurring work, on a schedule or when something happens.

  • Agents that collaborate

    Each agent has its role and its tools. They hand work to one another to carry a process from team to team.

  • Act without waiting to be asked

    They spot situations that call for action, then act within their limits or bring in a colleague.

For example

The sales agent spots a rise in orders and asks the stock agent to check availability. The stock agent finds a stock-out risk and passes the need to the purchasing agent, which prepares a supplier order for a manager to approve.

Chatbot, automation, agent: what changes

All three can use AI. They don’t do the same job.

Chatbot

You give it
A question
It gives you back
Text to rework
When the unexpected happens
It hands the question back
Where it works
In its chat window

Automation

You give it
A fixed rule
It gives you back
The same action every time
When the unexpected happens
It stops, or gets it wrong
Where it works
Between two applications set up in advance

AI agent

You give it
A goal
It gives you back
The work done, or ready to approve
When the unexpected happens
It adapts, or asks you
Where it works
In the tools you open to it

It works for you, under your control

An agent never sees more than the person it works for, and nothing important goes out without your approval.

It sees only what you see

Its access is that of the person it works for, read or write, application by application.

What matters waits for your approval

An email to a customer, an order, an entry: you see what will go out, then approve, edit or reject it.

Everything is logged

Every step, every exchange between agents and every approval can be looked up.

What an agent takes on in a company

Whole tasks that come back every day, every week or whenever something happens.

  • Accounting

    Match invoices and orders

    It reads each invoice received, finds the order and prepares the entry.

    Whenever an invoice arrives

  • Sales

    Prepare meetings

    History, open quotes, latest exchanges: a brief ready before the call.

    The day before the meeting

  • Purchasing

    Keep an eye on stock

    It cross-checks sales and stock, spots shortages and prepares orders.

    Every morning

  • Human resources

    Answer employees

    Leave, health cover, remote work: the answer drawn from your agreements, with its source.

    On demand

  • Customer service

    Sort requests

    It classifies each message, replies following your procedures or hands over.

    Whenever an email arrives

  • Management

    Produce the reporting

    The week’s figures gathered, formatted and shared.

    Mondays, 8 am

What no software did until now

These tasks always fell to someone, for lack of a tool able to take them on.

Link your software together

Accounting, banking, CRM and email don’t talk to each other: someone copies, compares and checks. The agent moves from one to the next, the way a colleague would.

Read what fits no box

A freely written email, a scanned invoice, a photographed delivery note: no “if this, then that” rule can handle them. The agent reads them and pulls out what matters.

Let nothing slip

Checking every order, chasing at the right time, following every due date: what a stretched team ends up putting off, the agent does every time.

Which task would you hand to an agent?

Bring a real process from your company: we’ll show you how an agent takes it on.

Book a demo

To prepare your demo, tell us a bit about yourself and what you’d like Glove to take on. No clear idea yet? We’ll help you find your use cases.

Obside, the company behind Glove, uses your details to answer your request and to tell you about its offers. You can unsubscribe or exercise your rights at any time by writing to [email protected]. Details in our privacy policy (opens in a new tab).