AI-Native Ways of Working

As AI absorbs the mechanical work of building software, what's left is the part that still has to be human: the judgment. This is my working canon for that shift — what I believe, how I work, and the practices I keep testing on my own systems before I'd ask anyone else to trust them. It changes as the evidence does.


Someone has to go first. Then mark the route.

The three shifts

fromIN THE LOOPtoON THE LOOPsteer the flow, don't gate it
fromTRUSTtoVERIFICATIONreliability from checking,
not model magic
fromDISCOVERY WHILE BUILDINGtoDISCOVERY BY DESIGNput back the learning AI skips

The four values

Outcomesoveroutput
Earned autonomyoverstanding permission
Human judgment where it mattersoverhuman hands on everything
Verified workovertrusted work

There's value on both sides. When they pull against each other, choose the left.

The seven principles

The depth behind the card — how the shifts and values get practiced.

  1. Shift the human up the loop — doing, reviewing, coaching, setting intent — as trust is earned.
  2. Gate only the irreversible and the high-blast-radius; let the rest flow.
  3. Give every agent a role, constraints, and an escalation path.
  4. Pair workers with checkers — reliability comes from checking.
  5. Make delivery-state visible; pay down comprehension debt; never rubber-stamp.
  6. Preserve discovery — put back, on purpose, the learning AI skips.
  7. Adopt AI as change management, on an owned and durable substrate, matched to your risk.

The practice areas

Seventeen recurring problems of human-agent teams — the field guide's table of contents. The four marked ◆ FOUNDATION run underneath all three shifts above. Each area becomes an essay as the evidence comes in, tested here first.

Where the human sits as agents earn trust.Essay — Aug 28
02Agent roles & team structure
Roles, constraints, and escalation paths — for agents and the humans around them.TO COME
03Orchestration & dispatch
Agents dispatching agents, and keeping the routing honest.TO COME
04Interfaces & harnesses
The working surface: how a person actually supervises the work.TO COME
WIP-limiting the human; finishing over starting.Essay — coming late Sep
06Comprehension debt
Work getting done faster than you understand it.TO COME
07Verification & quality
Checkers for the workers; testing the tests.TO COME
08Measurement & outcomes
Better than what, by how much — signals you can run a team on.TO COME
09Incidents & recovery
What the organization does after its first agent-caused failure.TO COME
10Coaching agents vs. coaching people
The same craft on different fuel.TO COME
11Discovery & planning
Putting back the learning that building used to force.TO COME
12Adoption & change management
How teams actually shift — fear and identity included.TO COME
13Skill formation & the expertise pipeline
Where senior judgment comes from now.TO COME
14Governance, trust & risk◆ FOUNDATION
Boundaries, audit, and agent security that let teams move.TO COME
15The durable substrate◆ FOUNDATION
Owning the layer that survives model and vendor churn.TO COME
16Memory & continuity◆ FOUNDATION
What agents read, and what survives the session.TO COME
17Leadership & sponsorship◆ FOUNDATION
The leader's own practice, stance, and sponsorship.TO COME

The Companion Instrument

AI-Native Engineering Maturity Model

Where your organization stands across ten dimensions, and what to move next. Free, no gate.

Take the assessment →

Created and maintained by Ed Schaefer — executive coach & strategic advisor.
Host of the Leadership Explored podcast.