Craig Horton Advisory
A Paper on the Human-Agent Operating Model

The End of the Org Chart

How the human-agent operating model extends a 100-year arc from Taylor to Hamel.

By Craig Horton, AI Transformation Advisor April 2026  ·  brief.craighortonadvisory.com
Executive Summary

The 100-year operating system, with AI layered on top.

Gary Hamel calls it inheritance bureaucracy. I call what the industry currently calls transformation, decoration.

Most organisations are still running a 100-year-old operating system, and have now added AI tools on top of it. As most of you know, I've been a big fan of Gary Hamel for several years, he calls this “inheritance bureaucracy”, I call what the industry currently calls transformation, “decoration”.

The real transformation sits underneath the technology, in the operating model, and when AI agents can reason, decide, and act, the question has to change.

Is your org chart fit for the future, or an archaeological record of decisions made by people who never imagined AI would be their colleague?
Craig Horton, opening question
94%
of leaders are not making progress on designing human-AI interactions.
Deloitte, 2026
The Question Cascade

One strategic question, one governance question, three operational questions.

Answer these honestly, and the shape of the redesign starts to show itself.

Strategic
How do we maximise the value of humans and machines working together?

The question every leader I speak with is circling, and the one Deloitte's 2026 Global Human Capital Trends puts at the centre of the next era of work. The governance question follows immediately.

Governance
Who has the authority to decide when algorithms act and when humans intervene?

And then the three operational questions, the ones that force the redesign itself.

Operational
What work still needs a human on it?
What is a team, when half the team is an agent?
What is hierarchy for, now that coordination is no longer the scarce resource?

The answers require a different design, and you cannot get there by layering agents on top of the architecture you already have.

The 100-year Arc

From Taylor to Hamel, a century of assumptions that broke.

Each era of management design was rational for its moment. None of them anticipated a world where some of the workers would not be human.

Era
Innovation
Assumption That Broke It
Scientific Management1911, Taylor
Task specialisation and measurement
Only humans work, managers must control
Knowledge Worker1960s, Drucker
Worker as asset, not cost
Hierarchy still controls knowledge flow
Matrix1960s+
Dual reporting for flexibility
More reporting lines, more coordination cost
Business Units1980s+
Decentralised accountability
Silos without integration
BPR1993, Hammer and Champy
Radical process redesign
Redesign stopped at process, power structure intact
Agile2000s+
Team autonomy and iteration
Teams agile, structures not
Humanocracy2020, Hamel and Zanini
Bureaucracy as the tax on human capability
Incentive to defend bureaucracy underestimated
Human-Agent Model2025+, Horton
Humans and agents as one system
The one we are writing now
The Human-Agent Operating Model

Four modes your workflows sit in, whether you designed them that way or not.

Mode 3 is where the step-change lives, Mode 4 is the frontier, and most organisations are stuck at Mode 2, the copilot ceiling.

01
Mode 01
Human Solo

Humans do all the work, AI is absent or cosmetic. The org chart is the operating system, and for workflows needing deep human judgement, ethical reasoning, or relationship management, this is the right place to stay.

The legacy state, still correct for some work.
02
Mode 02
Human-led,
Agent-assisted

Humans decide and act, AI tools suggest, draft, or accelerate. Copilot territory, where most AI transformation programmes stall. A 10 to 15 percent gain, and then the ceiling.

The copilot ceiling, where most of you sit today.
03
Mode 03
Agent-led,
Human-governed

AI agents execute end-to-end, humans set goals, handle exceptions, and hold accountability. The workflow has been redesigned around what agents can do, and willingness to redesign is what separates the organisations that get here.

The step-change, where the real value lives.
04
Mode 04
Fully
Autonomous

Agents operate within defined guardrails. Humans design the system, the system runs the steps. Real-time governance, continuous monitoring, automated escalation. Operational today in procurement and healthcare.

The frontier, already shipping.
The Diagnostic

Eight dimensions of readiness, score yourself honestly.

To move from Mode 2 to Mode 3, you need to see where you sit across eight dimensions of the operating model, dimension by dimension, without the aspiration that typically hides in strategy decks.

01
Workflow Design
Are your processes designed for AI, or is AI layered on legacy?
02
Decision Rights
Who decides, and how fast do decisions move?
03
Team Structure
Are your teams composed for human-agent collaboration?
04
Governance and Accountability
Can you govern agents that operate continuously?
05
Data Foundation
Is your data infrastructure ready to fuel agentic operations?
06
Talent and Roles
Are you developing people for the human-agent future?
07
Culture and Leadership
Do your leaders enable or block the transition?
08
Measurement and Value
Are you measuring coordination cost, agent contribution, and outcomes?
Run the diagnostic
Each dimension scores Mode 1 to Mode 4. The aggregate maps to a maturity stage from Legacy Hierarchy through to Agentic Enterprise, with recommendations for the four lowest dimensions and an organisational drag estimate.
Open the interactive diagnostic
The First 90 Days

From a diagnostic score to a shipped redesign, in ninety days.

The playbook the research firms do not give you. A concrete sequence from score to redesigned workflow, measured and documented.

What this is

Strategic frameworks describe the agentic future in detail, and the ones I return to are McKinsey's Six Shifts, BCG's Now, Next, Always, and Deloitte's 2026 Human Capital Trends. This sits one layer downstream, the sequenced ninety-day path I use with leadership teams inside the building, from diagnostic score to a workflow redesigned, shipped, and measured.

Days 1 to 30
Map
Run the diagnostic across the senior team, honestly, and identify the two lowest dimensions.
Pick one workflow inside those dimensions, high-volume, high-friction, and owned by a leader who will stay close to the work.
Map the gap between where that workflow is today and where it needs to be, and turn it into a redesign brief the team can work from.
Output Scoring document, gap map, one workflow picked.
Days 31 to 60
Redesign
Redesign the workflow AI-first, with agents executing core steps and humans governing at pre-flight, not at runtime.
Recruit or assign the agent orchestrator, give it budget authority and a named outcome.
Wire the governance into the workflow itself, embedded monitoring, automated escalation, clear human boundaries.
Output Redesigned workflow in staging, governance playbook, orchestrator in post.
Days 61 to 90
Prove
Ship the workflow to production, with the governance live and a rollback path defined.
Measure coordination cost reclaimed, decision cycle time, and the hours returned to human work that actually needs a human.
Document the pattern in a one-pager the next function can lift straight off the shelf.
Output Production deployment, before and after metrics, pattern library entry one.
The Reframe
If you were designing your organisation from scratch today, knowing that AI agents could handle 40 percent of current tasks, would you build the structure you have now?

If the answer is no, stop optimising what exists and start designing what should.

After Gary Hamel, Q:A Ratio, November 2023
“Managers default to answers and leaders default to questions. In his framing, a manager's Q:A ratio sits below 1 and a leader's sits above 1.”