Agents & Automation · Intermediate
AI Agents and Workflow Automation
Build AI agents and automations on your own processes, tested and running by the end of day two.
In two days your team hands its repetitive work to AI agents and automations that read the request, look up the data, update the record and send the reply. You leave with workflows built on your own processes, tested on real cases and running, with a person approving the steps that matter.
What the research shows
Figures from published studies and surveys. Each one links to its source.
43%
less time on Microsoft Entra access-policy tasks for identity admins given an AI agent, at equal accuracy (randomized trial, 162 admins)
Bono, Cheng & Lozano (Microsoft), arXiv, 202550%
more ads per worker when people teamed with an AI agent instead of another person (randomized experiment, 2,234 participants)
Ju & Aral, Collaborating with AI Agents, arXiv, 202626%
more completed tasks (pull requests) per week for developers who used an AI coding assistant, in randomized trials with 4,867 developers at three companies
Cui et al., Management Science, 2026What you leave with
- Working automations on your own processes, built and tested during the course and ready to switch on
- A step-by-step map of each workflow showing what the AI does, what a person approves and where the data comes from
- An AI assistant that answers from your company documents and acts in your tools, such as updating a CRM record or drafting the reply to a ticket
- A test set of real and awkward cases for every automation, so you see it hold up before it touches live work
- Guardrails your IT and data protection people can sign off: human approval on risky steps, logging and a fallback when the AI is unsure
- A handover pack with an owner, documentation and monitoring, plus a ranked backlog of the next tasks to automate
- On the technical track: a coded agent prototype that calls your systems through tool use or MCP, retrieves from your own data and comes with evals and tracing
What the course covers
From chat to agents
What separates a chat assistant, a workflow and an agent: the model, the tools, the memory and the loop that ties them together. You see working examples and learn which kind of solution fits which job.
Choosing and mapping your processes
Each team maps its own repetitive tasks step by step and scores them on volume, risk and data access. By the end of the module you have the processes you will build on for the rest of the course.
No-code workflows
Build triggers, steps and conditions in Make, Zapier, n8n or Power Automate, connecting email, forms, spreadsheets, your CRM and Teams. The AI step reads, classifies, extracts and drafts, and the workflow routes the result to the right place.
Prompts built for automation
Write prompts that return the same structured output every time, using examples, output formats such as JSON and clear rules for when to stop and ask a person.
Agents with tools and company knowledge
Build an assistant in Copilot Studio, a custom GPT or a Claude Project that answers from your own documents and takes actions through connectors. You decide what it may do alone and what needs a person's approval.
Testing and guardrails
Put every automation through real cases, edge cases and deliberate attempts to break it. Then add human approval, logging, data protection rules and a fallback path, with GDPR and the EU AI Act in mind.
Handover and the next backlog
Document each automation, name an owner, set up monitoring and agree how changes get made. To finish, you rank the next tasks to automate so the work keeps going after the course.
Technical track: APIs, tool calling and retrieval
Developers work in code with the Claude and OpenAI APIs: tool calling, MCP servers that expose your systems safely, retrieval over your own data, and evals and tracing to debug behavior and keep cost and latency in check. Coding agents such as Claude Code and GitHub Copilot speed up the build and write the tests.
Who it is for
- Operations and process owners whose teams move data between systems by hand every day
- Analysts who rebuild the same report or data pull every week
- Power users who use ChatGPT, Claude or Copilot daily and want AI to finish whole tasks on its own
- IT and digital workplace specialists who support, secure and scale what business teams build
- Developers and technical specialists who want to build agents with APIs, tool calling, MCP and retrieval
- Finance, customer service, HR and sales teams with high-volume, rule-based work such as invoices, tickets or applications
Tools
We reply within one working day.
From the blog.
27 October 2026
Agent Skills moves to the Linux Foundation. Instructions you write for one AI tool now have a neutral home
26 October 2026
Estonia's AI agent register is heading to parliament. Cybernetica, ITL and two regulators want it rewritten
6 October 2026
An AI agent that goes rogue was given too much freedom. The EU’s cyber chief on where to look.
Or we build one around your team.
Closed group sessions for your own team, or a custom curriculum from scratch. Tell us the problem and we shape the program.
Ask for a quote