Three Core AI Transformation Skills
Choose the right work, build the solution, and enable people to adopt it
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 95%
The model separates AI transformation into three skills that are often incorrectly treated as one. The first is diagnostic: determine which work should be automated or augmented and where AI belongs in the existing system. The second is constructive: decide what application, workflow, or integration to build and create it around genuine operational needs. The third is enabling: help employees understand, trust, and consistently use the new capability. Strong transformation requires all three because a technically impressive build can solve the wrong problem, while a correctly chosen and well-built system can still fail through poor adoption. Systems and process thinking connects the stages by keeping technical decisions grounded in how work actually moves through the business.
Origin
Extracted from Marketing Against The Grain during a discussion of why businesses need human specialists to implement AI successfully.
Core principles
- 01Automate only after identifying valuable work
- 02Treat integration choices as business-design decisions
- 03Build solutions around actual workflows
- 04Adoption is a separate capability from development
- 05Combine technical curiosity with systems thinking
How to run it
- 1
Choose what to transform
Study current processes and identify where AI automation or augmentation would materially improve cost, speed, quality, or capability. Decide explicitly what should not be changed.
Pro tip Prioritize recurring work with a clear trigger, outcome, and owner.
Watch out Do not automate a process merely because it is technically possible.
- 2
Define what to build
Translate the selected opportunity into a concrete workflow, internal application, custom tool, or integration. Specify inputs, outputs, systems, users, and success measures.
Pro tip Design the smallest useful version that can test the central assumption.
Watch out An undefined business workflow cannot be rescued by a sophisticated model.
- 3
Build and integrate
Implement the solution inside the systems and constraints where employees already work. Test reliability, permissions, handoffs, and exception handling.
Pro tip Include real users and representative data in validation before broad deployment.
Watch out A standalone demonstration is not a working transformation if it cannot operate within the business.
- 4
Enable the users
Train users, explain the new workflow, provide support, and establish feedback loops. Observe whether people use the system correctly and whether it improves the intended outcome.
Pro tip Teach the changed job and decision process, not just the interface.
Watch out Deployment without enablement commonly produces low adoption and shadow processes.
In the wild
A services firm first identifies proposal drafting as a high-friction process, then designs an assistant connected to approved case studies and pricing rules. The project team trains account staff, monitors edits and acceptance rates, and adjusts the workflow from user feedback.
→ The firm addresses opportunity selection, solution construction, and employee adoption as separate but connected responsibilities.
A transformation consultant identifies repetitive defect classification as suitable for augmentation, builds a review workflow around existing quality systems, and trains inspectors to confirm or reject model suggestions. Exceptions remain under human control.
→ The solution improves review speed without mistaking model deployment for complete operational adoption.
Common mistakes
Building before choosing the problem
Technical teams may create impressive tools that do not address valuable work. Diagnose the process and desired outcome first.
Ignoring systems integration
A prototype that sits outside normal operations creates extra work and weak adoption. Design around existing data, permissions, and handoffs.
Treating launch as adoption
Making a tool available does not mean people can or will use it. Training, support, measurement, and workflow change are separate requirements.
Is it for you?
Best for
It is best for consultants, transformation leaders, and cross-functional teams responsible for practical AI adoption.
Not ideal for
It is not ideal for narrowly scoped technical work where the use case and user behavior are already fully defined.
From the transcript
“You really need to be an incredible systems and process thinker to figure out how to like integrate AI across things that are actually useful…”
“It's the two hard things are the knowing what to automate with AI and what to integrate AI into, knowing what to actually build, and…”
“They're really the three core skills.”
From the episode
The #1 Job Needed in the AI Era Right Now