What an AI employee is, and how it differs from a workflow

An AI employee is a workflow that has been given a job. Underneath it is the same chain of steps: a trigger, data, an AI step that reads and decides, and an action that sends the result somewhere. What makes it an employee is how you work with it: it has a role, a patch it covers, hours it works, a person it reports to, a budget, and a record of what it did each week. You manage it the way you would manage a colleague, not the way you would debug a script.

This article is for managers and team leads who keep hearing both words and want to know whether the difference matters. It explains what the employee view adds, what stays exactly the same, how a role is decided, and how to hire your first one.

The short version: same engine, different job description

Both columns describe the same flow. The right-hand column is how xMap Flows presents and runs it.

Nothing is stored twice. Open an employee and you are in its flow, block by block. Change a block and the employee changes with it.

Where the role comes from

An employee's role is read from what its flow does, mainly from the block that brings in its data. A flow that searches for places and keeps only the ones it has never seen before becomes an Opening watcher. One that follows a brand's branches becomes a Rival watcher. One that scores businesses as prospects with AI becomes a Sales scout; with contact details added, a Lead finder. Job postings make a Hiring scout, road incidents a Road watcher, ships a Port watcher, satellite pictures a Site inspector.

The subject comes from where it looks and what for: "Nairobi pharmacies", "cafes in Lima and 2 more", "Starbucks in Seoul". So a card on the Flows page reads like a line from an org chart: Opening watcher, Nairobi pharmacies, works every morning, reports by email.

If you run seven employees with the same role, give each a name of its own. Rename sets the employee's name and the flow's name separately, so "News watcher" can become "Port news" while the role stays visible underneath.

What the employee view actually changes

You hear from it, instead of checking on it

A workflow waits to be looked at. An employee writes to you: one note a day in its thread in the Team inbox, saying what it did, what it found, its best few and what it needs from you, and a separate note at once when a run fails, its budget runs out or an approval is waiting. See running a team of AI employees from one inbox.

You decide how much it does without you

Each employee that sends something has one setting for that: Ask me first, Act and tell me or Fully on its own. A workflow would need you to add and remove approval steps by hand; the employee does it for you. See how much an AI employee should do on its own.

You judge it by its week

Every working employee has a report card: what it found this week against last, the strong fits among them, an estimate of the hours it saved and what it used. On Mondays the cards arrive in one email, with a short review where you keep it, raise its budget, change it or let it go.

What does not change

• It is still exact. The data blocks, filters and schedules run the same way every time. The AI step decides within the rules you set; it does not rewrite its own job.
• It does not learn from your clicks. An AI employee follows its instructions and the data. To make it better, change its instructions or its filters, in the editor or by replying in its thread.
• A test run never sends. Trying an employee is safe; only Run once and deployed runs act for real.
• Permissions are the same. Who may see, change or deploy a flow decides who may see, change or deploy the employee.

An example: one job, two descriptions

Take a pharmaceutical distributor in Cairo that wants to know when a new pharmacy opens in its delivery area. As a workflow, that is: a schedule at 07:00, an area covering three districts, a places search for pharmacies, a step that drops ones already seen, a scoring step, and an email. As an AI employee it is an Opening watcher, Cairo pharmacies that reports to the sales manager every morning, asks before emailing anyone outside the company, has its own budget for the month, and says on Monday that it found, say, nine new pharmacies this week against five the week before (made-up figures). Same six blocks. The manager now has something to manage.

Hiring your first AI employee

1. On Flows, press Hire and pick a ready-made employee from the templates, or describe the job in one sentence in the box at the top.
2. Answer its questions: usually the city and the kind of place. It remembers your company's city and the kind of place you work with.
3. Press Hire. It takes a first look, shows what it found and starts work on its timetable.
4. Choose how much it does alone and, if you like, a monthly budget, in its panel (click its picture on Flows).

Good first hires are the plain ones: new places every morning, watch new competitors or companies hiring near you. Once two or three are working, a weekly brief employee reads what the others found and writes one summary for the team.