Pattern 01Stuck at chat
You gave everyone ChatGPT.
People write emails faster and look things up. No cost, cycle time, or revenue number moved.
Market signalMIT NANDA, 2025
5% of AI pilots show measurable P&L impact
Anonymous · one response per card
Eight weeks inside your company, built by your people
You bought the AI.The work didn't change.
Work That Matters teaches your people to hand real work to AI, measure what it's worth, and keep rebuilding it as AI improves, so more of their time goes to problems no playbook covers.
The pattern
Every company is doing something about AI. Almost none of it changes who does the work.
Pattern 02Gone when they leave
You hired someone to build automations for you.
A few shipped. Then the models changed, and nobody inside the company knows how to rebuild them.
Anonymous · one response per card
Pattern 03Faster, not different
You bought an AI tool for sales. And one for support.
Each function got a little quicker. How work moves through the company didn’t change.
Market signalSMB Group
Most mid-sized businesses stall at one or two use cases
Anonymous · one response per card
Pattern 04Never reached production
You gave people time to experiment.
Good demos. Nothing the business actually depends on.
Market signalMIT NANDA, 2025
5% of custom AI systems reach production
Anonymous · one response per card
Each of these can help. None of them builds what you actually need: people inside your company who know how to redesign their own work around AI.
The numbers behind the pattern
The evidence
Most pilots move nothing. Very few companies make the leap from isolated use to a new way of operating.
No measurable P&L impact.That's the result for nearly every enterprise AI pilot.
Reach advanced adoption. Most mid-sized businesses stall at one or two use cases with no strategy behind them.
Sources: MIT Media Lab / Project NANDA, “The GenAI Divide: State of AI in Business 2025,” preliminary findings · SMB Group.
The uncomfortable truth
The workflows you automate this year won’t be your advantage next year. The people who can rebuild them will be.
AI will keep getting better for years. Anything built to today’s models is a snapshot. No consultant will be there to rebuild every workflow each time the models change, and you wouldn’t want to pay one to be.
Where the value is
Most of your payroll goes to work that already has a playbook.
Reconciling invoices. Building the weekly report. Qualifying leads, scheduling, onboarding, answering the same forty support questions. Hard work, done by skilled people, but known work. For decades software couldn’t do it, so you hired people to. AI can now do a growing share of it.
Work with a playbook
Hard, but knowable. Someone could write it down and train a new hire on it.
Reporting, reconciliation, scheduling, research, support, sales operations.
AI can increasingly do this.Work without one
Nobody knows the answer yet. You have to probe, test, judge, and adapt.
Why aren’t customers buying? What should we build next? How do we make this line of business profitable?
This is where your people should be.The prize isn’t making the eighty 20% faster. It’s moving that line.
That’s what we mean by agency. It isn’t a feeling; it’s capacity. Your people can start work without waiting for a new hire, and get to the problems that sat untouched because execution ate the week.
The method
Treat AI like a capable new hire, not a tool.
Software waits for someone to operate it. An agent takes an objective, works through the steps, uses your systems, and comes back with finished work. Getting value from that is a management skill, and almost nobody has been taught it.
- 01
Give it a job
A clear responsibility, not a prompt.
- 02
Give it access
The information and systems the work actually needs.
- 03
Set its authority
What it can decide on its own, and what comes back to a person.
- 04
Define good
What finished, correct work looks like.
- 05
Check the work
Every output is reviewed by the person who owns the outcome.
- 06
Widen the leash
Autonomy grows as it earns trust.
None of this depends on our software. It’s a way of working your people keep, whatever the models look like next year.
What could one of your people hand off first?
Three inputs. We’ll show you work someone already on your payroll could hand to AI in eight weeks, and the metric each one moves.
Turn one of these into a real eight-week build.Request a free 30 min working session
How it works
Your people build it. On real work. In production in eight weeks.
Each participant takes work they already own, builds an agentic workflow that does it, and runs it on real company data. We agree the baseline before anything is built, so at the end you have a before-and-after number, not a demo.
Week 01
Phase 01
Find it
Work they already own, with a playbook, that eats real hours. We measure how long it takes today.
Week 02
Phase 02
Frame it
The responsibility, the systems, the limits of its authority, and what good output looks like.
Weeks 03–06
Phase 03
Build it
An agentic workflow, built by them, never for them.
Weeks 07–08
Phase 04
Run it
Live on real company data, reviewed by the person who owns the work, measured against the week-one baseline.
8 weeks6–12 people8 protected hours a weekExecutive sponsorAccess to real systems and dataBuilt by your people, never by us
The return
You’ll know what you spent on AI, and what it bought you.
Every workflow is measured three ways: the human hours the work took before, the hours it takes now, including the time spent checking it, and what that difference is worth at what you pay the person who used to do it.
- 01
Workflows
Agentic workflows doing real work, in production.
- 02
Hours returned
Measured against the baseline, review time included.
- 03
Capacity
Time your people control again, priced at what you pay for it.
- 04
Unsolved problems
Spent on work no playbook covers.
- 05
New value
Customers, products, and markets you couldn’t reach before.
Three questions you’ll be able to answer
- What work now runs without a person doing it?
- How much human time came back, and what is it worth?
- What did your people do with it?
Hours saved are the proof. They aren’t the point.
After the program
The program ends. The redesign doesn’t.
Your people leave with a method, and with a place to use it. When the models improve, the people who built the workflows rebuild them, and each new person who learns the method adds more capacity.
The method
A way of working that holds up whatever AI you use next year.
The platform
Where your people’s workflows run, where they review and correct the work, and where every run is recorded.
The people
They know how to do it again, on the next piece of work and the next model.
For companies that want AI to change the economics, not the tool count.
You already believe AI should change how your company runs. You suspect licenses and encouragement won’t get you there. You want a number. And if every meaningful call still routes to you for sign-off, there’s nothing for your people to run. It’s the first thing we test on a call.
A strong fit
20 to 200 people, and a CEO, COO, or CTO who will sponsor it
You’ve spent on AI and can’t point to what it changed
You can name work with a playbook that eats your people’s weeks
You have capable people whose time goes to execution
You know what you’d do with the capacity if you had it
Not a fit
The plan is to give everyone ChatGPT and see what happens
You want a workshop, or better prompting
You want someone to build automations for you
Participants would work on practice exercises
Every meaningful call still needs your sign-off
Recognize your company in the strong-fit column?Request a free 30 min working session
Terms
If nothing ships, 100% of the fee comes back.
“Ships” means running on real company data, in real use, by week eight. If it isn’t, the full $3,000 fee is refunded.
Program fee$3,000Per person · eight weeks
Shipping guarantee100%back if nothing ships
Executive sponsorAccess to real systems8 protected hours a week
The first workflow should pay for the program. Every one they build after that is capacity you keep.
Bring one piece of work. Leave knowing what it’s worth.Request a free 30 min working session
Who runs it
Ryan York
Co-founder & Chief Product and Technology Officer at Willow Education
View LinkedInOperator first
Twenty years building teams, institutions, and software people actually depend on.
Ryan has led from the classroom, the principal's office, the executive team, and the product seat. He co-founded and co-led a 650-student public charter school, hired and managed more than 200 people, and ran an annual budget exceeding $8 million.
500%User growth in his first 90 days at Willow, with 100% uptime
10,000+Students reached by platforms and programs he led
100+Public schools adopted the computer science program he built
Today he builds AI products at Willow and runs his own work through agentic workflows, rebuilding them each time the models change. This program teaches the method he uses himself.
A working session
Bring one piece of work. Leave knowing what it’s worth to hand off.
Thirty minutes. We’ll pick one piece of work someone on your team owns, sketch how an agentic workflow would take it over, and estimate the hours and dollars at stake. You leave with that whether or not you work with us.