Agentic Leader
AI Fluency Matrix Level 4: Transforming

Build an AI agent that runs one of your real workflows, without code.

A four-week live program for non-technical leaders who already build AI assistants. You redesign one real workflow, build an agent that runs it within limits you set, and prove it creates value before you scale it.

Leaders from these companies have trained with us
4 weeks
About 3 hours a week
Weekly 90-minute live session
No code
Build in the AI tool you already use
OpenAI - ChatGPT Logo SVG
ChatGPT
Next cohort: [Dates]
Agentic Leader

Build an agent that runs one real workflow, in four weeks.

Daan van RossumPawel GorskiHenrik JärleskogPhil KirschnerGary C. Tate
Led by Daan, Pawel, Henrik, Phil and Gary, who run their own work on agents
Expected price
[$1,399] per person
Format
Live online, four weeks
✓
One working agent on a real workflow
✓
Four live sessions, all recorded
✓
Test evidence and a net-value case
✓
A 15-minute intake interview before week 1
✓
Lessons for ChatGPT Work, Copilot Studio and Claude Cowork
✓
Join alone, or as a group of 3 to 4 with 20% off per seat

Join the waitlist to hear the dates first and get first access to seats.

30-day money-back guarantee
Lead with AI, across all our programs
4.9/5
average participant rating across our programs
3,000+
leaders trained by Lead with AI
100+
assistants, agents and automations a typical team in our programs builds within three weeks
Why this matters now

Agents are scaling faster than the people who can design them.

The tools are no longer the barrier. ChatGPT Work, Microsoft Copilot Studio and Claude Cowork let leaders build agents without writing code, and large companies are already scaling them.

The scarce skill is design and supervision. Poorly designed autonomy creates more checking and risk than it removes, so this program teaches that judgment first and the build second.

40%

of respondents at large organizations report scaling AI agents in at least one function, up from 27% a year earlier.

Source: McKinsey, The State of AI in 2026, August 2026
0.2%

of knowledge workers are estimated to be at Level 4, able to design and supervise agentic systems.

Source: Lead with AI Fluency Matrix estimate, September 2026
Who it is for

For leaders who already build their own AI assistants and do not write code.

You have been working with AI for a while and want to go a level deeper. You do not need a technical background to get there.

You are ready if:
✓
You use AI for real work every week.
✓
You have built at least one reusable assistant, such as a custom GPT, a ChatGPT Project or a Copilot agent.
✓
You have a recurring workflow you would like to stop operating step by step.
✓
You can use ChatGPT Work, Microsoft Copilot Studio or Claude Cowork, through your company or your own plan.
✓
You can set aside about three hours a week for four weeks.
Start somewhere else if:
✕
You are new to AI. Start with AI Leader Advanced and come back afterwards.
✕
You want to write code for agent frameworks. This program builds in the tools leaders already use.
Leaders and managers

Hand a recurring operational workflow to an agent and put the time back into judgment, people and decisions.

AI Champions

Build the first governed agents your colleagues can see working, then help them design their own.

Consultants and business owners

Run more of your business on agents you designed and can explain to clients.

What you will learn

Four live weeks, about three hours a week, from one real workflow to a working agent.

Before week 1

A 15-minute interview with Lexi, our AI interviewer, maps your role, suggests workflows worth delegating and checks that your workflow fits the tools your company allows, so the agent you build can actually run. You bring one real workflow to the first live session.

Week 1 · Delegate
Choose work worth handing over.
  • How an agent works: it gathers context, decides, acts and checks the result. (The agent loop)
  • Pick the right workflow: one that is bounded, testable and possible on your platform. (Agentic Fit Check)
  • Decide how much to trust it: five levels of authority, set for each action by what a mistake would cost. (The autonomy dial)
  • Decide who does what: each step goes to the agent, a fixed automation or a person. (O.V.E.R. mapping)
Live: Delegation and workflow design clinic
You leave with a reviewed Agent Build Plan and a baseline of how the work runs today.
Week 2 · Build
Turn the plan into a working agent.
  • Create your agent in ChatGPT Work, Microsoft Copilot Studio or Claude Cowork.
  • Write instructions that hold up, using AI to get the know-how out of your head. (CODO)
  • Connect only the knowledge and tools the agent needs.
  • Set approvals and stop rules, then test one boundary on purpose.
Live: Build and reliability clinic
You leave with one controlled end-to-end run on real inputs.
Week 3 · Prove
Find out if it is reliable and worth it.
  • Make it dependable: fixed automation for repeatable steps, and AI only where judgment is needed.
  • Test it like a manager would: real cases that check the result, the process and the limits.
  • Test, inspect, improve and retest until the results hold.
  • Prove it is worth it: time saved minus setup, review and fixing time. (Net value)
Live: Build clinic and Q&A
You leave with an agent tested on real cases, and the evidence to show for it.
Week 4 · Demo Day
Show what it handles and decide what comes next.
  • Present your agent: what it handles, what still needs approval, your test evidence and one failure you fixed.
  • Decide whether to scale, hold, redesign or retire it.
  • Plan the next 30 days and pick your next workflow.
Live: Demo Day with your cohort
You leave with a decision, a 30-day plan and your [certificate name].

Every week includes about 30 minutes of lessons on learning days, one 90-minute live session offered in two time zones, and recordings of everything.

Joining as a group? You follow the same four weeks on one shared workflow. You map it together in week 1, build your parts in week 2, test the full run in week 3 and present together on Demo Day.

What it looks like in practice

Agents you could build, with a human in charge.

Swipe through twelve agents across industries and roles. Each card shows what the agent does, the result it is built to deliver and the skills you pick up building it. In the course, you build one for your own work.

Semiconductors · Partner management

Partner meeting briefer

Before every partner call, it pulls the account history, open support cases and roadmap commitments, and posts a one-page brief in Teams an hour before the meeting.

Target result
4 hrs
back each week for every account lead
Based on 8 partner meetings a week, with prep cut from 45 to 15 minutes.
Runs in
Copilot Studio
Autonomy
Prepare
It flags anything it could not confirm from a source instead of guessing.
You learn
Calendar trigger
CRM connection
Source citations
E-commerce · Customer service

Returns resolver

It reads each return request, checks the order against your returns policy and settles routine refunds on its own, day and night.

Target result
15 hrs
back each week for the service team
Based on 60 routine requests a day at 3 minutes each.
Runs in
ChatGPT Work
Autonomy
Run and escalate
Refunds above your limit, damaged-goods claims and upset customers go straight to a person.
You learn
Refund limits
Helpdesk connection
Escalation rules
Retail · Store operations

Stock-out watcher

Every morning, it reads yesterday's sales and stock for each store, spots best-sellers about to run out and drafts stock transfers for the regional manager to approve.

Target result
100%
of stores checked before opening
Based on a 120-store region where a manager can only spot-check a few stores by hand.
Runs in
Copilot Studio
Autonomy
Act with approval
No transfer goes out until the regional manager approves it.
You learn
Daily schedule
Reading sales data
Approval step
Higher education · Admissions

Applicant reply agent

It drafts a reply to every prospective student from your approved program pages, fees and deadlines, and books advisor calls for applicants who are ready.

Target result
16 hrs
back each week in peak season
Based on 400 enquiries a week, 3 minutes to write a reply and 30 seconds to approve one.
Runs in
ChatGPT Work
Autonomy
Act with approval
An advisor approves every reply. Visa, scholarship and fee questions go straight to a person.
You learn
Knowledge sources
Inbox connection
Approval step
Any industry · Marketing

Monday campaign analyst

Every Monday, it pulls the week's ad, web and pipeline numbers, explains what moved and why, and proposes where to shift budget.

Target result
5.5 hrs
back each week for the marketing team
Based on 6 hours of manual reporting across four channels, now a 30-minute review.
Runs in
ChatGPT Work
Autonomy
Prepare
It proposes budget moves, and you decide on every one.
You learn
Data connections
Weekly schedule
Judgment rules
Any industry · Finance

Variance commentary drafter

It compares actuals with budget, finds every variance over your threshold, pulls the invoices and emails behind it and drafts the month-end commentary.

Target result
9 hrs
back every month-end close
Based on 40 cost centers, with 18 minutes of writing cut to 4.5 minutes of review each.
Runs in
Copilot Studio
Autonomy
Prepare
Any variance it cannot trace to a document is marked as unexplained.
You learn
Thresholds
Excel and email flow
Audit trail
Any industry · Operations

Board pack builder

It collects the latest numbers and updates from each team's folder, builds the board deck draft in your template and checks every figure against its source.

Target result
1.5 days
back every month
Based on 16 hours of monthly assembly, now 4 hours of review.
Runs in
Claude Cowork
Autonomy
Prepare
Any figure that does not match its source is highlighted for you.
You learn
Working across folders
Template rules
Fact-check pass
Any industry · People and HR

New-joiner coordinator

When an offer is signed, it requests the laptop and system access, books the first-week meetings and chases any missing paperwork.

Target result
40 hrs
of admin back each month
Based on 10 new joiners a month at 4 hours of coordination each.
Runs in
ChatGPT Work
Autonomy
Act within bounds
Anything outside the standard first-week plan, such as a senior hire or a missing contract, comes to you.
You learn
Multi-step flow
Action limits
Reminders
B2B software · Sales

Inbound lead researcher

It researches each demo request, scores the fit against your ideal customer and drafts a personal reply for the right rep to send.

Target result
8.5 hrs
back each week for the sales team
Based on 40 inbound leads a week, with research cut from 15 to 2 minutes each.
Runs in
ChatGPT Work
Autonomy
Act with approval
A rep approves every reply, and strategic accounts are flagged for a personal call.
You learn
Web research step
Scoring rubric
CRM update
Higher education · Academic quality

Student feedback analyst

It reads every end-of-term comment across all courses, groups the themes, compares them with last year and drafts action notes for each program head.

Target result
40 hrs
of reading done in one afternoon
Based on 5,000 comments at about 30 seconds each.
Runs in
Claude Cowork
Autonomy
Prepare
Any theme backed by fewer than five comments is marked as weak evidence.
You learn
Large-file analysis
Testing against a sample
Evidence quotes
Built by Phil Kirschner

Newsletter results check

It pulls the latest newsletter results in Copilot Studio and sends a go or no-go card in Teams.

Real agent
In use
in Phil's own work
Runs in
Copilot Studio
Autonomy
Act with approval
Phil says yes before the full report goes out by email.
Built by Pawel Gorski

Weekly deal follow-up sweep

It reads the week's deal updates, checks each one against the CRM and creates the missing follow-up tasks.

Real agent
In use
every week by Pawel's team
Runs in
Claude
Autonomy
Act within bounds
Pawel confirms any date it had to assume, and it never completes or deletes a task on its own.
Swipe to see more agents

Each card shows a possible workflow. Its target result is worked out from the volumes on the card, not from a client. In the course, you build a version of your own workflow that fits your system access and the four weeks, then work out its net value.

Better as a group

Bring 3 or 4 colleagues, redesign one shared workflow, and save 20% per seat.

Most workflows cross more than one desk, and that is where the waiting happens. Bring the 3 or 4 colleagues who share one workflow, from your department or across functions. You map it together, each build the part you own, and the agent takes over the handoffs.

Example: new customer handover in a B2B software company
1
Sales
The deal closes in the CRM

An agent gathers the contract details and drafts the handover note.

Prepare
2
Legal
The contract check

A person reviews it, and the agent flags non-standard clauses first.

Suggest
3
Finance
Billing set-up

A fixed automation creates the account and the first invoice.

Fixed automation
4
Customer success
The kickoff

An agent drafts the kickoff plan and books the meeting once you approve.

Act with approval
Target result
2 days instead of 9
from signed deal to kickoff
Based on four handoffs that each wait about two days today, now passed on the same day.
One workflow, one shared plan

In week 1, you map the whole workflow and decide which steps go to the agent, an automation or a person.

Each person owns a part

You build your part in week 2 and test the full run together in week 3.

You present together

On Demo Day, the group shows the workflow before and after, and each of you earns the certificate.

Groups get 20% off per seat. Choose Me and a few colleagues in the form to join as a group. Join the waitlist

Outcomes

Leave with a working agent and test results that show what it saves.

A working agent

It carries one real workflow forward on a trigger or schedule, within the limits you set.

Proof it is worth it

Test results that show the agent gets the right result and stays inside its limits, plus a before-and-after count of the time it saves after review and upkeep.

A method your team can reuse

The fit check, the autonomy dial and the build plan work for every agent your team builds next.

Sample Agentic Leader certificate of completion from Lead with AI, for a Level 4 program in the AI Fluency Matrix
Your certificate
A certificate that shows what you built.

Finish the four weeks and present your agent on Demo Day to earn your Agentic Leader certificate, awarded for completing a Level 4 program in the AI Fluency Matrix. Add it to your LinkedIn profile and show the agent behind it.

Your facilitators

Taught by a team with careers at McKinsey, Sodexo and WeWork, who run their own work on agents.

Every facilitator builds and supervises agents in their own work. Each live session is led by one or two of them, so you learn what holds up in practice.

Daan van Rossum
Daan van Rossum
Founder and CEO

Leads the design of this agentic curriculum. He runs about ten scheduled agents in ChatGPT, from one that keeps his priorities current to a research agent for his book, plus a Claude agent that screens his LinkedIn invitations against his own rules.

LinkedIn Top Voice. Research cited by Harvard Business Review and Fast Company.

Pawel Gorski
Pawel Gorski
Co-founder and Chief Growth Officer

Runs a weekly Claude agent that turns the team's deal updates into follow-up tasks, and leads Lead with AI's own internal AI transformation.

Previously CEO of Tribee AI.

Henrik Järleskog
Henrik Järleskog
Co-founder and Managing Director, Europe

Runs a chief of staff agent and 16 automations across his part of the business, including a morning brief that triages a 21-meeting day before he wakes up. His own estimate is about 16 hours a week handed back.

20+ years in strategy and operations, including Sodexo's 65,000-person European operation.

Phil Kirschner
Phil Kirschner
Transformation Partner

Facilitates the agentic block of a champions program at a global pharmaceutical company, where participants build agents in Copilot Studio. He tells people plainly when a workflow is not worth an agent.

20+ years at McKinsey, WeWork, JLL and Credit Suisse.

Gary C. Tate
Gary C. Tate
Chief AI Officer

Leads the team building LeadOS, the platform that interviews every participant and matches them to the workflows worth delegating. He taught our earlier Agentic AI cohorts.

35+ years in technology across AI, cybersecurity and SaaS.

Results

One leader took both courses, then brought them to about 200 colleagues.

From participant to company program

A director of people analytics at a global semiconductor company took AI Leader Advanced and our agentic course himself. Then his company booked a private cohort that combines both. In 2027, about 200 people in its People group go through it.

50+

AI agents built by Catawiki teams and now used in daily workflows.

“Lead with AI delivered high-impact, tailored programs across our Leadership, People/IT, Legal, Finance, and Marketing teams.”

Iolanda Meehan, CHRO, Catawiki

Leaders from these companies have trained with Lead with AI. Source: Lead with AI program data.

Program at a glance

Join the waitlist for first access to seats, backed by a 30-day money-back guarantee.

Format
Four-week live online cohort
Time
About 30 minutes on learning days, plus one 90-minute live session a week in two time zones. Every session is recorded.
Platforms
ChatGPT Work, Microsoft Copilot Studio or Claude Cowork
Prerequisite
AI Leader Advanced, or equivalent experience building assistants
Investment
[$1,399] per person, with a 30-day money-back guarantee
Teams
Private cohorts on your platform, with your policies and your examples
Next cohort
[Dates]. Waitlist members hear first.

Join the waitlist

Be the first to hear the dates. Planning for a team or your organization? Tell us below and we will reply within one business day.

We only email you about this program, and you can unsubscribe at any time.

Frequently asked questions

Do I need to know how to code?

No. You build in ChatGPT Work, Microsoft Copilot Studio or Claude Cowork with instructions, connected knowledge, tools and approval steps. The skill this program teaches is designing and supervising an agent, and that skill carries over to whichever tool you use next.

Do I need to take AI Leader Advanced first?

It is the best preparation. If you already build reusable assistants, such as custom GPTs, ChatGPT Projects or Copilot agents, you are ready. Your intake interview confirms the fit before the program starts.

Which platforms does it cover?

ChatGPT Work, Microsoft Copilot Studio and Claude Cowork, with demos for each. You build on the platform your organization approves. If you work in another platform, mention it in your intake interview and we will advise you.

What if my company has not switched on agent features?

In week 1 you check that your platform can reach the information and tools your workflow needs. If a capability is not available to you, we help you design a version that works within what is approved.

What if I get stuck building my agent?

Bring it to the weekly live session. Weeks 2 and 3 are build clinics, where facilitators help you fix what does not work yet. Between sessions, you can send a facilitator your build questions in writing.

Is this safe for company data?

You build only with the tools, data and connectors your organization approves. Every agent you design has explicit approval steps and stop rules, and you test its boundaries before you rely on it.

How much time does it take?

About three hours a week for four weeks. That is roughly 30 minutes on learning days and one 90-minute live session. Every live session is recorded.

What if I miss a live session?

Every session is recorded, and you can ask questions in the cohort community between sessions.

Can we run this privately for our team or company?

Yes. The same program runs as a private cohort on your platform, with your policies, approved tools and examples. Choose a team or organization option in the form and we will reply within one business day.

Can my company pay?

Yes. After you join the waitlist, you get an approval email to send your manager. If your company needs an invoice, reply to our email and we will send one.

Can we join as a group?

Yes. Bring 3 or 4 colleagues who share one workflow, from one department or across functions. You all need the same platform and access to the workflow's data, and one of you owns the workflow and signs off on what the agent may do. Groups get 20% off per seat. Choose Me and a few colleagues in the form.

What happens after I join the waitlist?

You pay nothing now. You hear the dates first and get first access to seats. If you chose a team or organization option, we contact you within one business day.

What if the program is not right for me?

Every seat comes with a 30-day money-back guarantee.

Which workflow would you hand over first?

Bring it to week 1. Join the waitlist to hear the dates first.

Join the waitlist
[$1,399] · 30-day money-back guarantee
Join the waitlist