Build an AI-enabled company that keeps getting better on its own.
Work Redesigned is a workforce AI enablement practice for growing companies. It builds the ownership, documentation, evaluation, and measurement that turn scattered AI activity into a capability the organization runs on its own.
For HR and People leaders.
Why AI rollouts stall
AI activity can be high while organizational capability stays low.
One team has built a useful workflow. Another is paying for a tool nobody opens. Employees use personal accounts because the approved system does not fit the real work. The documentation is outdated, so every AI output starts from a different version of the truth.
The result is scattered progress. Underneath it sits an older problem: the work was built around how the job is supposed to be done, so the judgment you hired for was never part of the process, and AI is now making that expensive.
Uneven use
Capability depends on the person, team, or tool rather than a company-wide way of working.
Isolated workflows
Useful builds remain local, are recreated elsewhere, and often have no documented owner.
Inconsistent source material
AI works from outdated documents, personal files, and competing versions of the truth.
Unequipped managers
The people responsible for quality and performance were never given a way to evaluate AI-assisted work.
An AI-enabled company meets six conditions at once.
Six simultaneous conditions. Each answers a different question from a different person in the room.
Together, the conditions put judgment back where the work happens, which is what keeps a company improving after the initial push ends. Employees can improve real work. Managers can judge quality. Leaders can make defensible decisions. Technical teams can see what is being built and who maintains it.
A half-day workshop is the smallest starting point.
Applied to your real organizational questions and your own working examples. Your team leaves with shared language, practical decisions, and a next step it can use.
Leading AI Sustainably
The AI spend keeps growing, and nobody in the leadership room can say who owns the outcome or what better looks like. Clarify ownership, spending, checks and balances, and the next decision the group has to make.
Participants leave with a clearer ownership decision, an agreed definition of better, and a practical starting point.
Guiding Your Team Through AI
Your managers know their people are using AI, and each one is improvising. Learn what to encourage, what to slow down, what to escalate, and how to make room for useful experimentation.
Participants leave with clearer leadership responsibilities and a shared approach to guiding their teams.
Think Like An Engineer
Most workflows people build work once, then quietly become someone else's maintenance problem. Build ones that survive contact with reality: testing, troubleshooting, token costs, edge cases, and maintenance in plain language.
Participants leave with a practical build standard and an applied exercise based on their own work.
Evaluating AI Work
AI-assisted work is crossing every desk, and quality depends on who happens to review it. Establish what quality looks like for AI output, agents, and new ways of working, and how to keep it current.
Participants leave with a manager evaluation approach they can apply to real work.
The work happens in three acts.
Read the whole company. Prove it in one function. Extend what works.
The organization gains visibility before choosing where to begin. One function proves the new way deeply enough to build internal capability. The last act turns that capability into a company-wide rhythm.
Act 1 · 4 weeks
Whole company
The Fog Clears
You can see the whole board, and someone’s name is on it.
What is actually happening across the company: how people use AI, where shadow AI sits, which tools are paid for and never opened, where work is duplicated, and which function has the strongest combination of opportunity and leadership support.
Act 2 · 8 weeks
One function, deep
The Work Changes
One function works a different way and can build the next thing itself.
The team documents the real work, exceptions and failure paths included, then converts the right procedures into Automated Operating Procedures. Every task gets the right level of automation. Nontechnical builders learn to test, troubleshoot, cost and maintain what they create. The team watches one real build happen, then makes the next one itself.
Act 3 · 6 weeks
Whole company
The Pull Begins
The work spreads, it is measured, and it maintains itself.
The first function becomes the company-wide way of working. Ownership distributes across leaders, the registry and spending view extend across functions, and measurement becomes an internally run habit. The company can find the next opportunity and extend the method without waiting for another external push.
From shadow AI and outdated documentation to a shared way of building.
Vinfen, a human-services organization of roughly two thousand employees, operating across several states in the northeast.
The situation
The HR team was working across a range of personal AI tools. That shadow usage produced inconsistent material and unclear sources of truth, and the approved AI environment could not become useful until somebody understood how people were actually working.
The work
An assessment of the team’s capability and of how people felt about using AI at work showed which HR subteams were best positioned to start.
Outdated SOPs were rewritten around the real process, including the knowledge that had only ever lived in experienced employees’ heads. Where it made sense they became Automated Operating Procedures, structured so AI could execute the work without that unwritten expertise.
Then the team learned to build with the discipline of a development process: context, testing, edge cases, troubleshooting, maintenance. The last piece reconsidered how the reclaimed hours got spent.
What changed
Role redesign focused reclaimed capacity on more thoughtful and valuable work.
The organization gained a shared view of current capability and AI usage.
The strongest starting teams were identified deliberately rather than selected by volume or enthusiasm alone.
Critical work was documented in a form useful to both people and AI.
The team gained a repeatable approach for building workflows that address edge cases and can be maintained.
Guidance that leaves the capability with you.
Keith Anderson, founder of Work Redesigned
I work at the intersection of AI, learning, behavior change and business. Across Google, YouTube, Facebook, DoorDash, Uber and Calibrate, the same thing was true: capable people were hired for their judgment and then managed as though they had none, and the technology only ever paid off when that changed.
I tend to see the pattern before the language for it exists. My job is to make that pattern clear and turn it into a method, a standard, or a decision your organization can use.
I guide without becoming the permanent source of every answer. The work should leave your people with better judgment than the job has been letting them use, and the confidence to continue without me. Serious change can be rigorous without becoming heavy.
Follow the work behind Work Redesigned.
Practical field notes on AI-enabled work, manager judgment, organizational capability, and what companies are learning as AI moves from individual tools into everyday operations. Written from real engagements, in plain language you can use on Monday.
Frequently asked questions
Usage is not capability. Individual wins that stay local do not become a reliable way of working. This work builds the ownership, documentation, evaluation, and measurement that turn scattered activity into an organizational capability.
All three, in sequence. Act 1 reads the whole company. Act 2 builds alongside one function: real automations, real documentation, real skills. Act 3 hands the rhythm to your organization so it can continue without an external push.
Good. A named owner is one of the six conditions. The work equips that owner with visibility, a method, and a measurement rhythm rather than replacing them.
No. Nothing needs to be prepared for Act 1. Documentation happens in Act 2, only for the work being redesigned, and only to the level a person or an agent needs to execute it.
A standard operating procedure rewritten around the real process, exceptions and failure paths included, and structured so AI can execute the work without relying on unwritten background knowledge, with human-review rules attached.
Not very. Nontechnical builders learn testing, troubleshooting, token costs, edge cases, and maintenance in plain language, then make the next build themselves.
Every task gets the right level of automation, with human-review rules written into the automation map. Managers get a rubric and an evaluation method for judging AI-assisted work, and every build is registered with an owner.
Act 1 is mostly interviews and an assessment. Act 2 asks real time from one function. That is where the capability is built. Act 3 folds into existing leadership rhythms.
Four weeks for the Diagnostic, twelve weeks through one function, eighteen weeks end to end. A half-day workshop is the smallest starting point.
With a conversation. The first call establishes which starting point is proportionate to the problem.