You build your first agent. A real one, with your real data.
What you build
A personal site agent from your LinkedIn profile. Not copy-paste. Genuinely better, with guardrails that enforce differentiation and storytelling.
The teaching moment
You use best-in-class positioning frameworks as guardrails. The agent steers toward something you could not write yourself. That is the first insight: with the right constraints, the output is higher quality than what you would produce alone.
02
Building Agent Systems
One prompt got you started. Now you design a system. Multiple steps, each with a job.
What you build
A multi-step pipeline for your company initiative. The agent doesn't just generate. It researches, structures, drafts, and refines. Each step has clear inputs, outputs, and quality expectations.
Why this matters
This is the leap. From chatbot user to system builder. The shift from a single prompt to a designed pipeline changes how people think about AI.
One agent, multiple steps. Useful. Multiple agents coordinating. Powerful. And fragile.
What you build
You split your pipeline into specialized agents. One researches, one structures, one writes, one reviews. They hand off work like an assembly line. Each focused on what it does best.
Why this matters
Coordination is where agent projects succeed or die. You experience both: the power of specialization and the pain of agents that don't understand each other.
An agent that works but isn't secure is worse than no agent.
What happens
Your agent breaks your organization's compliance rules. You watch it happen in real time. Then you build the guardrails to fix it.
Your organization goes in here
Your security policies, your data classification, your compliance requirements. Not our defaults. Yours.
05
Output Quality and Hallucination Control
Module 1, you caught fabrication by eye. Now you catch it with a system.
What you build
An evaluation system for your company initiative. Not "does it look good?" but "can I measure good?" Evals as strategic steering. Making output more differentiated, not just correct.
Why this matters
From hoping it works to knowing. They say evals is the new black. I say so too.
Individual evals catch problems. Evaluation systems catch patterns.
What you build
A systematic evaluation pipeline. Automated checks that run every time the agent produces output. Same discipline software teams apply to code, applied to agent output.
Why this matters
An agent without evals is a demo. An agent with evals is a system you can trust and improve over time.
You've built agents. Now: where do they live?
What you learn
The current platform landscape. What's real, what's hype, what actually works in production. How to evaluate platforms against your requirements. No vendor recommendations. A framework for making your own assessment.
Why this matters
After building in Modules 1–6, you know what you actually need from a platform. You can evaluate for yourself.
08
Agents Building Agents. The Flywheel
The flywheel. The insight that changes how you think about scale.
What you build
An agent that builds other agents. Multiplication, self-improvement, and the shift from project to capability. This isn't a one-time thing. It's a new way of working.
The deliverable
An executive recommendation, co-created live by the group. Not a report about what you learned. A concrete proposal for what to do next.
The best strategies emerge from action, not analysis
We adapt everything to your organization. What your team builds stays yours.
Send me a DM on LinkedIn
Or if you know me, just call me.
What Comes After
Agents 102 is Step 1. After the modules, your people have built working agents. The next steps:
Step 2
Make Your Own
A cohort of your change leaders designs the rollout. 4–6 weeks. Change strategy lab with mentoring.
Step 3
Champions
The strongest people from Step 1 learn to facilitate and carry it forward. Train-the-trainer.
Step 4
Scale
Champions run it inside your organization. Your priorities, your pace. No external dependency.
Step 5
Put Agents to Work
Map your processes. Rework them with agents.
The goal is no external dependency. Your people carry it forward.
What We Believe
Building is thinking
Concepts emerge from doing. We never explain something before the exercise that demonstrates it.
Your data, not toy data
Your profiles, your policies, your domain. Not sample_data.csv.
Show the failure mode
For every capability, you see what goes wrong without the discipline. The violation. The undetected fabrication.
Your reality replaces our defaults
Security policies, evaluation criteria, domain knowledge. Clearly marked places where your organization goes in.
One new thing at a time
Each exercise adds one concept. Never two at once.
You keep everything
Full usage, modification and further development rights. If you stop working with us, you keep it all.
If this is how you think, let's talk
Your initiative, your policies, your systems. Your people leave with working agents and a concrete plan for what to do next.
Send me a DM on LinkedIn
Follow the newsletter
Or if you know me, just call me.
I'm Antti Tevanlinna. Twenty-five years at F-Secure, Nitor, Neste, Posti Group, Kone, S-Group, Wärtsilä, Elisa and Finnair. Over two hundred people trained. I built this because I kept seeing the same gap: everyone talks about AI agents, nobody teaches people to build them. A version of this is already running.
More about Agents 102 by Bosser →