Intent
Quarterly to annual
Human-authored outcomes, trade-offs, boundaries, and the decisions the organisation refuses to delegate.
MAIN CTO Day · Utrecht · 1 October 2026
Engineering Organisations at the Speed of AI
The product can now change faster than the company can decide. The answer is not more agents. It is an operating model that can hold their speed: human-authored intent, executable structure, bounded execution, and evidence that changes the next decision.
Execution became continuous. Management stayed episodic.
AI compresses the distance between customer intent and working product. That is useful until inherited decisions, policy, ownership, and review still travel at meeting speed. The resulting gap is not a tooling problem. It is an organisational design problem.
The ISEE framework
Four layers. Four different cadences.
Organisations become brittle when every decision is forced to move at the same speed. Intent should be durable. Structure should evolve deliberately. Execution can be continuous. Evidence must travel quickly enough to challenge all three.
Structure
Monthly to quarterly
The spine: policy, platform capabilities, context, interfaces, ownership, and encoded organisational memory.
Execution
Continuous
Cells and agents translate intent into working software, infrastructure, operations, and customer outcomes.
Evidence
Continuous to periodic
Telemetry, quality, security, cost, adoption, and drift challenge assumptions and trigger new decisions.
The operating model
A spine, cells, and an evidence loop.
Platform capabilities do not originate organisational judgement. The spine carries decisions into the paths agents and teams can actually take. Cells execute inside those boundaries. Evidence reveals where the structure is working, where it is drifting, and where humans need to decide again.
The spine
A human-authored system of intent, decisions, policy, context, platform capabilities, permissions, and non-negotiable boundaries.
The cells
Small human-agent teams with end-to-end ownership. They consume the spine instead of rediscovering the organisation's trade-offs in every task.
The evidence
Delivery, quality, security, cost, customer, and operational signals flow back into decisions. Drift becomes an input, not a surprise.
A practical first move
One product. One cell. Ninety days.
Do not begin with an enterprise reorganisation. Use one consequential product to test whether intent can become executable structure, whether agents can operate safely inside it, and whether the resulting evidence improves the next decision.
Name the Intent
Choose one product outcome and write down the trade-offs, constraints, and decisions that must remain human-authored.
Author the Spine
Turn those decisions into context, platform capabilities, policies, permissions, quality gates, and evidence contracts.
Run One Cell
Give one small, cross-functional human-agent cell end-to-end ownership for ninety days. Do not scale the org chart first.
Read the Evidence
Measure whether speed improved the outcome. Feed drift, exceptions, quality, security, cost, and customer evidence back into the spine.
Continue the argument
Ghost Decisions
The book behind the talk: how invisible inherited decisions shape AI-native engineering organisations, and how the ISEE framework turns intent, structure, execution, and evidence into an operating system for speed with accountability.
Products and platforms
The systems that make boundaries executable.
Tools are not the operating model, but they are how the operating model becomes real. These products provide the development, security, quality, identity, automation, and operational surfaces that agents and humans can share.
Agent control plane
Microsoft Agent 365
Microsoft
The control plane for inventory, identity, access, observability, security, and governance across an enterprise agent estate.
Open resource AI experienceMicrosoft Copilot
Microsoft
Microsoft's human-AI interaction layer, connecting individual assistance with agents that participate in organisational workflows.
Open resource Work systemMicrosoft 365 Copilot
Microsoft
AI assistance and agents grounded in Microsoft 365 work, data, and organisational context.
Open resource Developer platformGitHub Enterprise
GitHub
The governed software-production platform joining repositories, pull requests, automation, security, and AI-assisted delivery.
Open resource Agentic developmentGitHub Copilot
GitHub
AI-assisted and agentic software development across planning, implementation, review, and repository-based delivery.
Open resource Quality evidenceGitHub Code Quality
GitHub
Continuous maintainability and reliability findings that turn code quality into visible evidence inside the repository workflow.
Open resource Security guardrailsGitHub Advanced Security
GitHub
Code scanning, secret protection, dependency controls, and security evidence that can act as enforceable guardrails.
Open resource Governed automationGitHub Agentic Workflows
GitHub
Repository automation authored in natural language and compiled into auditable GitHub Actions workflows with explicit permissions.
Open resource OperationsAzure SRE Agent
Microsoft Azure
An operations teammate for investigation and governed mitigation, showing agents participating in production rather than stopping at code generation.
Open resourceOpen source
Working systems, not just diagrams.
Use these projects to inspect the mechanics: converting intent into specifications, packaging context, enforcing infrastructure guardrails, evaluating ISEE maturity, and running governed agentic workflows.
Execution
Git-Ape
Microsoft Azure
A reference implementation of agentic platform engineering: agents gather requirements, generate infrastructure, price trade-offs, run security and architecture gates, deploy, and detect drift.
Open resource Agent contextAgent Package Manager
Microsoft
Packages and distributes agent instructions, prompts, skills, agents, and dependencies so context can be versioned and reused.
Open resource StructureApe Context
Suzanne Daniels
Builds project-level context and instructions from organisational and repository knowledge, including governed decisions and context-quality gates.
Open resource EvidenceISEE Advisor
Suzanne Daniels
Assesses Intent, Structure, Execution, and Evidence maturity and makes organisational drift visible without confusing missing evidence with proof of absence.
Open resource Intent to specGitHub Spec Kit
GitHub
A spec-driven development toolkit that moves intent and constraints upstream before coding agents execute.
Open resource Engineering systemHVE Core
Microsoft
Hyper Velocity Engineering practices and reusable assets for increasing delivery speed while retaining disciplined engineering controls.
Open resource PatternsAwesome Copilot
GitHub
A community catalogue of instructions, agents, prompts, and configurations for extending GitHub Copilot responsibly.
Open resource Workflow engineGitHub Agentic Workflows
GitHub
The open-source implementation behind governed repository agents, including compilation, permissions, safe outputs, and workflow controls.
Open resourceResearch and reports
The evidence behind the talk.
The presentation synthesises organisational, software-delivery, platform, context-quality, security, and AI-adoption research. The dual-track maturity roadmap is Suzanne's synthesis of these sources, not an official combined model from Microsoft, Accenture, or Carnegie Mellon SEI.
Research report
2026 Work Trend Index: Agents, Human Agency, and Opportunity
Microsoft WorkLab
Finds that organisational factors account for roughly twice the reported AI impact of individual factors. Tool adoption is not operating-model transformation.
Open resource Research report2025 Work Trend Index: The Year the Frontier Firm Is Born
Microsoft WorkLab
Introduces the Frontier Firm and the shift toward human-agent teams, redesigned workflows, and new management capacity.
Open resource Industry analysisHow Will AI Change Software Organizations?
Bain & Company
Explores smaller human-agent pods, blurred role boundaries, end-to-end ownership, and the need to redesign workflows and decision systems.
Open resource Industry researchDigital and AI Leaders Outcompete
McKinsey & Company
Shows that sustained advantage comes from rewiring the operating model around technology, data, talent, and customer value, not isolated adoption.
Open resource Survey reportThe State of AI in Early 2024
McKinsey & Company
Tracks the rapid rise of generative-AI adoption, emerging value, risk practices, and the organisational capabilities separating high performers.
Open resource Readiness researchAI Readiness Landscape
Microsoft + Accenture
A readiness perspective spanning strategy, organisation, data, technology, talent, governance, and the transformation required to become a Frontier Firm.
Open resource Maturity modelAI Adoption Maturity Model v1.0
Carnegie Mellon SEI + Accenture
A maturity model for evaluating AI adoption across strategy, governance, engineering, operations, workforce, and measurable outcomes.
Open resource Research synthesisThe New Future of Work Report 2025
Microsoft Research
Synthesises research on generative AI's effects on knowledge work, collaboration, management, and organisational systems.
Open resource Engineering researchDORA 2024 Accelerate State of DevOps
Google Cloud DORA
Shows why local productivity gains do not automatically improve system outcomes and why strong platforms, feedback loops, and culture matter.
Open resource Industry reportState of Platform Engineering, Volume 3
Platform Engineering
Industry evidence on platform-team maturity, developer self-service, golden paths, and the continuing evolution of internal platforms.
Open resource White paperPlatforms White Paper
CNCF TAG App Delivery
Defines platform engineering as a sociotechnical discipline and grounds internal platforms, interfaces, and platform-as-product practice.
Open resource Research paperCAFE(S): Your Agent Is Only As Good As Its Context
ACM Queue
Defines Clarity, Actionability, Fidelity, Efficiency, and Security as practical properties of context supplied to agents.
Open resource Risk frameworkAI Risk Management Framework
NIST
An authoritative vocabulary for mapping, measuring, managing, and documenting AI risks across organisational systems.
Open resource Security guidanceOWASP GenAI Security Project
OWASP
Open threat models and security guidance for LLM and agentic systems, including prompt injection, excessive agency, and unsafe tool use.
Open resourceEssays and field reports
Follow the argument beyond the stage.
These essays, implementation reports, case studies, and practitioner perspectives connect organisational design to the daily work of engineering leaders, platform teams, and agent builders.
Essay series
Engineering Beyond Agile
Suzanne Daniels
The essay series behind Agentile, ISEE, cells, the changing engineering-management role, and operating beyond Agile's human-only assumptions.
Open resource Series part 1Platform Engineering for the Agentic AI Era
Microsoft Developer Blog
Why agentic interfaces collapse the traditional infrastructure interaction stack and what that changes for platform teams.
Open resource Series part 2Putting Agentic Platform Engineering to the Test
Microsoft Developer Blog
A working Git-Ape walkthrough: infrastructure generation, policy, security, cost, architecture review, approval, and deployment.
Open resource Series part 3Removing the Monkey Work of Migration
Microsoft Developer Blog
Using agentic platform engineering to extract intent, critique design, map services, and generate migration artefacts.
Open resource Engineering perspectiveHarness Engineering
Martin Fowler
Describes feedforward guidance and feedback sensors that constrain coding agents before and after execution.
Open resource Decision practiceArchitecture Decision Record
Martin Fowler
An established architectural pattern for preserving decisions and rationale instead of allowing them to become invisible inherited constraints.
Open resource Case studyAI Is Approving Our Pull Requests
Fin / Intercom
A production account of specialised review agents and bounded automated approval as implementation volume outgrows human review capacity.
Open resource Leadership perspectiveJudgment Is All You Need
Will Larson
Argues that as agents reduce execution scarcity, judgement and problem selection become the binding constraints.
Open resource Leadership perspectivePolicy for Strategy
Will Larson
Shows how strategy becomes executable through contextual policy and durable decision records.
Open resourceUse the framework
Take the next step.
Diagnose the operating model, explore the full ISEE matrix, or turn the talk into a ninety-day implementation plan.