01
Learn
Learner
The learner studies the provider's method and sees a worked example.
AI agent
The AI agent receives the approved method, rules, and examples.
For course providers
Graduate skilled people with their own trained AI agents.
We reorganize your course so learners build competence while training their own AI agent on your method, their context, and the work they must do after graduation.
The learner leaves with knowledge, demonstrated skills, and a working AI agent they have trained to apply both. You keep the redesigned course and the evidence behind it.
A sample learning task
One task develops the learner and the agent.
Example: a management course on giving difficult feedback. The subject remains management; the AI agent becomes the working partner each learner trains along the way.
02
Practise
Learner
The learner plans a difficult conversation and explains each choice.
AI agent
The AI agent asks questions grounded in the learner's real situation.
03
Review
Learner
The instructor assesses the learner's judgment independently from the agent's output.
AI agent
The learner checks the output, corrects mistakes, and records the revised approach.
04
Apply
Learner
The graduate conducts the conversation and reflects on the result.
AI agent
Their trained AI agent helps prepare the next situation using those corrections.
What the provider keeps
Your course, ready for the next cohort.
We deliver one course your team can run again. Your method stays yours, and the pilot shows how learners and their AI agents perform against an agreed baseline.
A redesigned learning journey
Goals, content, practice, assessment, and post-course use organized around the skill learners must demonstrate.
A template and training path
Approved sources, rules, examples, and a sequence through which each learner trains a personal working version.
An instructor playbook
Guidance for feedback, assessment, exceptions, content updates, and separate review of learner skill and agent output.
A pilot decision record
Observed learner performance, agent quality, teaching effort, and continued use compared with the starting baseline.
One-course transformation pilot
One redesigned course. One pilot cohort.
We redesign one existing course and run it with one cohort. Your team keeps ownership of the method, learning goals, and assessment.
- 01
Map the current process
Trace the learner journey, instructor work, and evidence available today.
Baseline and one outcome - 02
Redesign the course
Set how learners build judgment, train their AI agents, and demonstrate the skill.
Learning architecture - 03
Build the training path
Turn approved sources, examples, rules, and learner context into working tools.
Agent template and playbook - 04
Run one cohort
Learners solve real tasks while instructors inspect judgment, corrections, and support demand.
Delivered pilot and evidence - 05
Evaluate the result
Compare the pilot with the baseline and decide what to revise, stop, or expand.
Next-cohort decision
What the pilot tests
Judge learning and operations together.
The pilot records a baseline first. The provider chooses the thresholds that matter for the course and its business.
Application in real work
Evidence: an independent task beyond the teaching example, plus the learner's explanation of each choice.
Agent quality
Evidence: the agent's work checked against the approved method, sources, and exception rules.
Instructor and support effort
Evidence: time spent on repeated questions, feedback, correction, content maintenance, and cohort operations.
Continuation after the cohort
Evidence: an agreed follow-up showing whether graduates still apply the method with their AI agents at work.
Why 8Hats Lab
Built by people behind learning systems.
The team combines assessment, national learning platforms, university leadership, learning research, and AI systems in production.
Our Academic Council adds education science and psychology: Dale P. Johnson leads digital innovation at Arizona State University's University Design Institute; Irina Zingerman brings 15+ years in human–AI interaction design.
Meet the full teamOur learning architecture
Keep the whole learner in view.
Human–AI Learning Architecture (HALA) is the public framework we use to design each course stage. It defines how people learn while their own AI agents learn to work with them.
Knowledge becomes observable action
Design the mental model, realistic practice, feedback, and success criteria as one learning loop.
Motivation, energy, and peers count
Account for frustration, attention, pace, recovery, peer feedback, and the safety to expose mistakes.
Learning survives the cohort
The graduate takes a trained AI agent, working artifacts, and the provider's method into real work and teams.
Control stays visible
Start with one bounded transformation.
What does each graduate leave with?
Knowledge of the method, a skill assessed independently from AI output, and a personal AI agent trained through their examples, instructions, corrections, and working context.
What does the provider keep?
The redesigned course, agent template, learner training path, instructor playbook, and evidence record. The provider retains ownership of its method, learning goals, and assessment.
What does a trained AI agent mean?
It means an AI agent configured through the course with the provider's approved method and the learner's examples, instructions, corrections, and working context. It does not mean retraining a foundation model.
Where does each learner's AI agent live?
Before the pilot, we agree the approved host, learner access, usage costs, update path, and available export options. The agent can stay in the provider's environment or move with the learner when the chosen tools permit it.
How is human skill assessed?
Learners complete an independent task and explain their choices. Instructors assess their understanding and judgment separately from the agent's output.
How is our method protected?
Before loading material, agree the sources, access, retention, allowed AI model, distribution, and people who can approve a version.
Do we have to rebuild the whole course?
No. Choose one course, one learner outcome, and one cohort. Keep what already works and change only what the pilot requires.
First conversation
Choose a course to redesign together.
In the first conversation, we choose the skill learners must demonstrate and the work their trained AI agents should perform after graduation.
Discuss a one-course pilot