Turn your people into agent builders.
Across eight modules, your people build an agent system for their own work. They find mistakes, improve the instructions, and build checks so the same problems are easier to catch next time. Then they work out how someone else could use it.
See what your people buildBusiness builders and leaders in product, operations, sales, and other functions. No coding background needed.
A live challenge, your knowledge of the business, and company material you can use in the exercises.
About 50 minutes of setup and prework: Claude Code or Cowork, the training folder, two small builds, and a short read.
The memory, agents, rules, and checks accumulate in one working folder. Between sessions, participants put the system to use and bring back what happened. The next module builds on that evidence.
Start with output you can judge yourself. Move into a live challenge and a system with memory.
Build a personal HTML one-pager and steer it with positioning, strengths, and guardrails. Check the claims against what you know about yourself. Push back on generic answers and save the rules that made the result yours.
Take back to work Your own page, the rules that shaped it, and experience spotting what the agent made up.
Bring a live business challenge. Curate useful sources from company knowledge, recent work, and practitioner methods. Build a memory the agent can use, review its plan, and put a custom agent to work. Observe a scheduled agent between sessions.
Take back to work A working agent, a curated memory, and operating rules you can improve from actual use.
Build source-finding agents for different bodies of knowledge, then coordinate distinct viewpoints to answer a strategic question. Inspect disagreements and handoffs. Decide when splitting the work helps and when one agent is enough.
Take back to work A research and synthesis workflow tied to your challenge, with sources you can trace and an answer you can challenge.
Inspect risk, measure whether claims hold up, and run the checks repeatedly.
Check your system against policy and agent risks. Look at access, untrusted instructions, sensitive context, and tool use. Apply a mitigation, run the check again, and name what is still uncertain.
Take back to work A reusable security check, a mitigation you have tested, and a record of the risks that remain.
Compare four methods for checking claims against sources. Score them on a 30-claim benchmark from your work, then save the strongest method as a reusable judge. Record what it catches and what it cannot assess.
Take back to work A benchmark, a comparison of the checks, and a judge with a measured scope and explicit limits.
Turn the one-off check into a repeatable loop. Separate generation from judging. Keep the judge fixed while the generator changes its approach in response to feedback. Inspect the results and the changes between rounds.
Take back to work A runnable evaluation loop and a record showing how the work changed under the same quality standard.
Carry the system into team use and build the next agents from what you have learned.
Start with a teammate’s job and the outcome they need. Decide whether to share context, a skill, the output, or an interface. Plan for ownership and everyday use, then test the assumptions behind adoption.
Take back to work A sharing plan with technical and people decisions, plus a concrete next move with a teammate.
Use an agent to extend the system you have built. Then bring the group’s systems together around a real company question: find the crux, test assumptions, and challenge proposals before choosing a direction.
Take back to work A proposed direction for the business problem you worked on, agents to support it, and a plan for the next two weeks.
Participants keep their working system and get an HTML workbook, a theory handbook, and practical references to return to. The agents, workflows, and skills they build stay theirs. Training material usage is covered by your agreement.
I built the engineering and business training tracks at F-Secure, trained leaders personally, and developed internal trainers to carry the work forward. Today I’m AI Transformation Lead at WithSecure.
Around 500 people have been through my training programmes. I’ve personally trained about 100 of them.
We’ll agree what they need to learn, which modules to run, and how to fit the sessions around their work.
After the first group, we train people inside your company to teach the next groups.
Talk to Antti