The learning path: Professional
Connections, loops, statuses, agents with tools, widgets and APIs with keys.
- 1
From a CSV to personal e-mails
You have a list of contacts in Excel or as a CSV, and everyone should get their own personal e-mail. In this path you put the list in a table, show it as a chat widget, let a workflow go through the rows one by one, have an AI Colleague write each mail and let Outlook send it. Along the way you connect Microsoft 365. You need: a CSV file and an app registration in Microsoft Entra ID (tenant ID, client ID and client secret).
- 2
Connect a partner's API
Your supplier or partner has a REST API with a base address and a key. You record it once as a connection; the key stays encrypted on the server. From then on any workflow calls the API with just a path, without ever seeing the key. You need: the API's base address and a bearer token.
- 3
Process row by row with a loop
A workflow reads the orders from a table, calculates the amount including VAT per order and writes it to a second table. That is how you learn the Loop (for each): steps that repeat per item, with variables you name yourself.
- 4
A funnel with statuses and rules
Quote requests go from request to quote and then to accepted or rejected. With a status column you record which step may follow which, and with rules per status who may see a row, change it or move it to a status. Flow enforces that for people, APIs and workflows, and writes every status change to the audit log.
- 5
An AI Colleague that uses workflows
An AI Colleague can start a workflow itself when it needs one, for example to look up the stock. You choose which workflows it may use, whether they only read or change something, and how often per run. You try it out and see exactly which tool it called, before you publish it.
- 6
A chat widget from a template
A chat widget shows data in the chat as a card, table, key figures or chart, instead of long text. You start from an example that works straight away, test and publish it, and then fill it from a workflow. An AI Colleague that may use the widget shows it in the chat itself.
- 7
Publish an API with keys
You create an API with one endpoint that runs a workflow, and give a consumer a key that opens only this API and has only the permission the endpoint asks for. Then you call the endpoint the way the consumer does. Every call can be traced to a consumer, with a key that expires.
- 8
Compare and restore versions
Version control keeps track of what changes in your environment. You commit a version, change something, see line by line what is different, and restore one item to an earlier version without touching the rest. Restoring is itself a change you commit: history is never lost.
- 9
Your team and their roles
In the portal you invite people to your organisation, each with a role. A builder builds in Development and publishes up to Acceptance; a colleague uses what has been built and tests and approves; the owner manages everything. You see per environment what a role may do, and how your colleague gets in.
- 10
An AI Colleague with knowledge and tools
An AI Colleague is only really a colleague when it knows what you know and can do things for you. Knowing comes from a knowledge base: Flow cuts your documents into pieces and, for every question, looks up the pieces that are about it. It hands those to the AI model, so the answer comes from your own text, with the source alongside. That is called RAG: look it up first, then answer. Doing comes from workflows the AI Colleague may start as tools. You build a course guide for a Flow course: it answers questions from the course material, looks up a student and enrols a new one. You try it, publish it and talk to it in the chat.