Public philosophy and professional identity
AI adoption is nearly finished. Capability development has barely started.
This is the public argument about what AI-native work is, what it demands of a professional, and why the vocabulary we use for it is currently doing more harm than good.
If you are deciding what to train, what to delegate to a machine, or who remains answerable for the result, this is the distinction those decisions turn on.
01 · Literacy versus capability
Knowing the tool is not the same as being able to do the work.
AI literacy is knowing what the tools are, roughly how they fail, and how to prompt them competently. It is necessary, it is teachable in an afternoon, and it is the part most organisations have already solved.
AI-native capability is the ability to operate inside a workflow where a human and a machine share the work and the human remains accountable for the result. That requires knowing where the machine's output must be verified, where judgment cannot be delegated, what has to be escalated, and how the reasoning is preserved for whoever inherits the work.
02 · Three terms
Three terms, three jobs. They are not synonyms.
AI-enabled
AI assists isolated tasks while the underlying workflow remains largely unchanged.
A property of tool use.
AI-native
The human and AI operate as one deliberately governed and accountable performance unit inside a defined workflow.
A property of the workflow.
AI-augmented
The professional expands the scope of work they can responsibly perform by operating one or more AI-native workflows.
A property of the professional.
The sequence runs one way. A professional becomes AI-augmented by operating AI-native workflows. Using the two words interchangeably as category labels collapses the distinction between how work is designed and what a person can then responsibly take on.
03 · The accountable performance unit
One unit, one accountable human.
The useful unit of analysis is neither the person nor the model. It is the pairing, inside a defined workflow, with one human accountable for the output. Once you adopt that unit, the design questions become concrete:
- Which steps belong to the human, which to the machine, and on what basis?
- Where are the handoffs, and what has to be true at each one?
- What is verified, by whom, against what?
- What must be escalated, and to whom, and at what threshold?
- What reasoning is preserved so the next person can pick the work up?
Escalation is not an optional courtesy in this design. It is a mandatory control.
04 · Nine capability domains
Nine domains that define AI-native capability.
These are the capabilities that determine whether a professional remains genuinely accountable for work a system helped produce. They are developed together, because in practice they fail together.
01
Task decomposition
Breaking professional work into components fine enough to decide, part by part, what may be delegated and what may not. Most poor delegation decisions are made at too coarse a grain.
Without it: you delegate a whole job because it was too much trouble to separate the parts.
02
Human–AI role allocation
Deciding deliberately which party does what, on what basis, and under what conditions the allocation should change. The alternative is allocation by convenience, which drifts in one direction only.
Without it: allocation drifts toward whatever is convenient, and only ever in one direction.
03
Contextual grounding
Supplying the situation, constraints, standards and history a system does not have and cannot infer. Much of what looks like model failure is unsupplied context.
Without it: you blame the model for not knowing what nobody told it.
04
Verification
Establishing whether output is correct, sufficient and appropriate, at a cost proportionate to the consequence. Verification that costs more than doing the work is not verification.
Without it: checking costs more than the work, so it quietly stops happening.
05
Professional judgement
The capacity to decide well in conditions that are ambiguous, contested or unprecedented, and to recognise when a situation is one of those. Built by practice, eroded by its absence.
Without it: the unusual case is handled like the routine one.
06
Workflow redesign
Rebuilding the sequence of professional work around the new allocation, rather than inserting a tool into a process designed for its absence.
Without it: a fast tool is bolted onto a process built for its absence.
07
Ethical reasoning
Working out what should be done where the rules do not reach: disclosure, consent, attribution, bias, and the limits of automated involvement in decisions about people.
Without it: the first hard question arrives without anyone having decided who answers it.
08
Capability transfer
Developing capability in others when the work no longer produces it as a by-product. This is the domain most organisations have not begun, and the one with the longest lead time.
Without it: your juniors never acquire what your seniors know.
09
Accountability
Holding and being seen to hold responsibility for jointly produced work, including the parts you did not personally perform and could not fully inspect.
Without it: everyone signs off work nobody can fully explain.
This page teaches. It does not measure. These nine domains are a way of organising a conversation about practice. Nothing here produces a score, a profile or a recommendation, and no response is collected. How the domains are weighted inside an instrument is a separate discipline, handled by Edney Learn.
05 · Judgment and accountability
Accountability was never delegable.
Accountability has always stayed with the professional. What has changed is that judgment, the thing that made accountability survivable, is now harder to acquire, because the work that produced it is the work most readily delegated.
You can delegate the drafting. You cannot delegate being the person who has to stand behind it.
06 · What changes inside a role
The shift is in what the role is for.
Specifying and verifying it
Being able to say what good looks like before seeing it
Making reasoning legible
Documenting judgment, not just decisions
Governing throughout
Designing checkpoints rather than inspecting output
Workflow design
Treating the pairing, not the person, as the unit
07 · Two professions
Where the argument is worked through.
The educator
In developmentNo Edney Learn educator programme is running yet.
The project professional
PilotThe Edney Learn Project Manager offer is at pilot stage.
08 · Measuring capability responsibly
Wanting to know where you stand is reasonable. Getting an answer that holds is a different problem.
Wanting to know where you stand is reasonable. Getting an answer that holds is a different problem, and it is where this field does most of its damage.
A measure that survives contact with a real decision needs the construct defined before anything is scored, items tested rather than assembled, calibration against real people, and a plain statement of what the result may and may not be used for. Most capability assessments skip all of it and go straight to the profile.
That discipline is operated by Edney Learn, as a single developmental profile rather than a second opinion competing with the first.
Pilot / controlled developmental releaseResults are developmental and do not establish verified competence.
09 · Evidence and open questions
How far the evidence goes.
The argument on this page is conceptual. Where an Edney Learn instrument exists it is at a controlled developmental stage. Nothing here establishes a measurement model, a population norm, a percentile, a ranking or verified competence, and no result from it is suitable for hiring, promotion, grading, certification or eligibility.
Putting this to work.
Reading it changes how you see the problem. Building it changes how the work runs. Edney Learn does the second part.