8 seats left before tuition increases. Applications close in:
2.5 weeks · Live and AI-guided · All recorded
Agentic AI stalls on the design, not the build. In 2.5 weeks, apply the MIT CSAIL design method to your own biggest problem, test your decisions in a working simulation, and leave with a method you can use again.
You leave with
Co-led by
Prof. Daniel Jackson
MIT CSAIL
Eagon Meng
MIT CSAIL
A technology collaboration: the Software Design Group at MIT CSAIL and NGL, a new technology start-up.
Full curriculum, weekly schedule, project examples, outputs, faculty, and time commitment.
Opens immediately after submission.
Starts
Tuesday September 15
Runs
2.5 weeks
Live
Tue & Thu, 12 PM ET
all recorded
Guided work
6–9 hrs a week
with Paski, your AI tutor
Fee
$1,695 $2,095
Save $400 · ends Thursday, September 10
9–12 hrs a week in total · no engineering background required
You bring
One real business problem
You apply
MIT CSAIL Concept Design
You produce
Blueprint + simulation
You become
An Agentic AI Fellow
Agentic AI Fellow membership is the Fellows Forum: weekly builder office hours with Dirk and Ivana, a monthly session with Daniel and Eagon, and the growing concept catalog.
Why this capability is rare
Agentic AI has made execution faster. It has not made the underlying decisions easier. The differentiator is no longer who can build. It is who can design.
The situation
Five things leaders in your seat say out loud. They sound like five different problems.
01
“What I asked for is never what I get.”
02
“I approved a demo I couldn’t interrogate.”
03
“I delegated the work and accidentally delegated the decisions.”
04
“We have pilots everywhere and nothing in the portfolio.”
05
“The AI bill keeps climbing and nobody can tell me why.”
They are one problem. Nobody wrote down what the thing was supposed to do, precisely enough for anyone, or anything, to build it right. This intensive is where you write it down.
View the full outlineMake it concrete
“Use an agent to handle customer onboarding.”
Before anyone builds it, the business has to decide how the service behaves. Five questions, and none of them are the engineer’s to answer.
01
What does it own?
Collect information, verify it, create an account, or only recommend the next action?
02
What can it decide?
Can it waive a requirement, reject a customer or approve an exception?
03
Which rules and boundaries apply?
04
What happens off the happy path?
05
When does a person take over?
The outline works all five through on a real example, and shows you what a complete answer looks like.
How the work goes
You already bring
You understand the work, the people and the consequences. But you may not yet have a method for making every important decision visible.
During the intensive
Use the MIT CSAIL concept design method to model how the agent should behave, then test that behavior in a working simulation.
You leave able to
You can interrogate how an agent should behave, explain your decisions, and define what you are prepared to build.
Organizations have always been able to run on implicit knowledge, because people fill the gaps without being asked. Agents cannot do that. You have to describe the business explicitly, in one language people, machines and agents can share.
Prof. Daniel Jackson
Associate director, MIT CSAIL · created concept design
The outline gives you the five units, what you build in each, and the one thing you’ll need to bring.
View the full outlineProof, not notes
The artifact is the evidence. You do not leave claiming to understand agentic AI. You leave having designed and tested how one real system should work.
01
Agent Design Blueprint
One document a person can read and an agent can work from, covering your system’s purpose, authority, rules, handoffs and exceptions.
02
Working simulation
Run the design on your own normal, ambiguous and failure cases before committing to a build.
03
Readiness review
A shared reference executives, domain experts and the people who build can interrogate, approve and build from.
Your first design is guided. The method is yours to run again. You leave with the prompts, sources and method to approach the next project yourself, plus the readiness review and the Fellows Forum.
This is not an attendance certificate. There is no certificate. It is the design itself, plus the tested behavior behind it, in a form your organization can act on.
Written once, read by both. The same document your executives approve or the document your agents are built from, so nothing is lost in translation between them.
Two and a half weeks, led by Prof. Daniel Jackson, who invented the method, with Eagon Meng and the applied faculty.
View the full outlineFaculty
Daniel invented concept design; Eagon is evolving it with him. Dirk and Ivana teach the sessions where the design meets a real organization.
Prof. Daniel Jackson
associate director, MIT CSAIL · author of The Essence of Software
One of the most cited researchers in software design, Daniel created concept design as a rigorous way to describe what systems do, in language every stakeholder can read. His methods are used inside companies from Autodesk to global enterprises aligning teams, products, and now agents.
Eagon Meng
PhD candidate, MIT CSAIL Software Design Group
Eagon’s research turns concept designs into running systems: the translation layer between human intent, agent behavior, and executable code. He builds the tooling that makes a concept specification something you can hand to an agent and trust what comes back.
Teaching in the intensive · weekly office hours in the Forum
.png)
Dirk Hofmann
co-founder and CEO, DAIN Studios Germany · Harvard Data Science Review board member
Co-founded DAIN Studios; before that led global data and AI work at Siemens, Nokia and Deutsche Telekom.
.png)
Ivana Ovcaric
principal data and AI strategist, DAIN Studios
A decade leading AI strategy and implementation across telco, media, banking and manufacturing.
Daniel, Eagon, Dirk and Ivana all teach in the intensive. Afterwards, Dirk and Ivana run weekly builder office hours and Daniel and Eagon host a monthly session in the Forum · included in the fee. You keep access to the growing catalog of vetted concepts, and can submit your own.
View the full outlineWho this is for
You do not write code. You decide what the system may do on your behalf.
Business judgment is the prerequisite. No code, no technical vocabulary, no engineering background.
What this is not
Not prompting. Not requirements writing. System design. No survey of models, no tool tour, nothing that expires with the next release cycle.
Not a seminar or a strategy exercise. Those end at understanding. This ends with artifacts in hand, and the method keeps running inside your organization afterwards.
$1,695 at full tuition, and the Agentic AI Fellows Forum is included.
View the full outlineQuestions
No. It’s a technology collaboration between the Software Design Group at MIT CSAIL and NGL, a new technology start-up: a transfer of the concept design method into your organization, with the reviewed portfolio as the evidence it worked. Executive education certifies attendance. This produces work.
No, deliberately. A certificate is a claim about a person; this is a review of the work. You leave with your concept catalog, the simulation, the kit, and a letter from MIT CSAIL stating that the work meets the standards of their method and that it is yours.
One pain point or one opportunity you care about. It can sit inside a product, a team or a whole function, and it does not need to be decided yet · bring two or three possibilities and the first live session helps you choose. Everything is applied to the one you pick, so there are no toy examples.
Strategy work ends at understanding: you leave knowing why agentic AI design is hard, with a list of mitigations. This ends with artifacts in hand: a concept catalog your organization can align on, materials whoever builds it, or your agents, can work from as written, and a simulation that runs. If you already know the “why,” this is the “what, exactly.”
Something gets built, and you are not the one typing it. You define the concepts; Paski, your AI tutor, generates the translations and a working simulation from them, and your job is to check what comes back against what you meant. If you want to write the code yourself, this is not it.
The outline gives you the five units, what you build in each, who leads them, and the one thing you’ll need to bring.
View the full outlineSeptember Cohort starts September 15 · $1,695 · no engineering background required
Starts September 15
.png)
