Hybrid Delivery — 2026

Delivery Leadership in Human+AI Teams

The Iteration
Manager in the
Age of Agents

Own what agents can generate but can't decide.

Presenter
Adnan Ali

Version
v2.1.0

aha agile

The Gap in the Squad

Engineers are already running agents. The coordination isn't keeping up.

441%

PR review time climbing as agentic throughput climbs; the bottleneck shifted right. One signal, not the whole case: the direction, coordination lagging individual adoption, holds across the studies below.

DORA 2025 · Stack Overflow 2025 (n=65,000+): 31% of developers using agents, 51% daily AI use
AI4Agile 2026 (n=289): IM cohort at chatbot-level · arXiv 2023–25: individual AI adoption does not improve team coordination

02

aha agile

Open here — this is the hook. The 441% PR review time figure is from DORA 2025 and it is the signal a delivery manager can look for in their own sprint data. If PR review time has increased in the last two sprints, that is likely an agent-in-the-loop coordination signal. The squad capability gap is assembled from three surveys with different sampling frames — the directional claim is well-supported but the precise magnitude cannot be stated as a single number from a single study. The arXiv longitudinal study (2509.10956, 2023–2025) is the most significant finding: individual AI adoption does not improve team coordination — it can erode role differentiation, accountability, and coordination. That is the specific risk the IM is positioned to address. Source confidence: High for DORA 2025 and Stack Overflow 2025; Medium-High for AI4Agile 2026 and arXiv longitudinal.

Locate Yourself

Where the squad is

Using AI agents in daily work
Several ways of working with AI, all informal
Coordination quietly getting harder
Extra review work nobody is tracking

Where you are

Still mostly using chatbots
No agreed way to govern agent work
No deliberate choice of how humans stay involved
Same ceremonies as before

That gap is the opportunity. You close it from inside the team — not by waiting for a mandate.

03

The Shift

"Agent" stopped meaning a chatbot. Now it means something that acts.

How much it acts — how much you hand over — is a ladder.

04

aha agile

Age of Agents

Agent capability is a ladder.

As agents climb it, your job moves from doing the work, to steering it, to governing it.

AssistedYou operate — the agent suggests; you do the work.
AugmentedYou review — the agent drafts and recommends; you execute and edit.
CollaborativeYou supervise — the agent plans and executes alongside you; you steer.
OrchestratedYou architect — agents hand off to each other; you set the gates and escalation.
AutonomousYou govern — agents run end-to-end; you audit and hold the kill-switch.

Synthesis · autonomy framing after Feng, McDonald & Zhang (Univ. of Washington, 2025); cf. SAE J3016, Cloud Security Alliance 2026

05

aha agile

This ladder is a synthesis of industry agentic-maturity models (Microsoft, Salesforce, Adaline, Bessemer), independently grounded in the University of Washington "Levels of Autonomy for AI Agents" paper (Feng, McDonald and Zhang, 2025), the Cloud Security Alliance autonomy levels (2026), and the SAE J3016 analogy. The load-bearing point: as agents mature, the human's job shifts from doing the work, to steering it, to governing it — gatekeeping becomes auditing. Keep it distinct from the AI Governance Maturity ladder (now in the appendix), which measures how well an organisation governs agents, not what the agents can do. Source confidence: High for the autonomy-spectrum concept (triangulated across four independent sources); the five-rung labelling is a presentation synthesis, not a single citable framework.

The Loop

The loop runs until it hits a gate you designed.

Agentic loop: Receive → Think → Act → Observe → Decide → (complete or loop back to Think)

06

aha agile

From the Trenches

The first boundary I owned
was one nobody was watching.

A team I ran let an agent draft and publish every release note for a sprint. Fast, tidy, no complaints. Then one went out with a customer's name in it. The agent hadn't decided to include it. Nobody had decided what it was allowed to publish. The fix wasn't more proofreading. It was naming who signs off before an agent ships. That's what owning the boundary looks like.

07

aha agile

The Pivot

What actually happens when a delivery team runs with AI — not beside it?

08

aha agile

The Evidence

For complex, generative work — the upside is real

+40%

quality uplift, human+AI vs. solo

BCG RCT · Organization Science 2026 · n=758

25%

faster task completion

BCG · same study

3×

more likely to produce a top-10% solution

P&G RCT · NBER 2025 · n=776

09

aha agile

The Counter-Evidence

On decision tasks, human+AI underperforms the best solo performer

106

studies analysed

Malone et al. · Nature Human Behaviour 2024

90%

were decision or classification tasks

Task-type breakdown

cond.

The advantage is real — and conditional on task type

Generative + iterative work only

10

aha agile

The Structural Problem

Most businesses don't redesign. They bolt AI on.

<40%

of businesses report measurable profit gains from AI. They layer AI on top of legacy workflow — the workflow never changes.

McKinsey MGI · AI adoption research

11

aha agile

What's Happening Now

Agents are generating this now

Requirements drafts
User stories
Sprint tracking
Status reports
Dependency mapping
Stand-up summaries

Nobody's governing this

Design choices
Risk thresholds
Approval gates
Production calls
Agentic workflow architecture

The automation is here. The governance gap is where the IM gets bypassed.

12

The Governance Gap

Adoption

84%

of agile teams use AI tools

Digital.ai · 18th State of Agile 2025 · n=350

Governance

49%

of those teams have governance guardrails

Same survey

The gap between those two numbers is where the risk lives.

13

The Boundary

Agents generate

Drafts
Tracking
Analysis
Dependency detection
Test coverage
Communication summaries

Agents cannot decide

Risk tolerance calls
Approval gates
Production decisions
Competing stakeholder interests
Ethical edge cases

The boundary is not permanent. The job is to own it right now.

14

Centaur Teams

Weak human + machine + better process beat the grandmaster.

Human

Judgment
Accountability
Context
Stakeholder navigation
Final sign-off

AI

Speed
Throughput
Pattern matching
Draft generation
Spec execution

Process design is the variable. That is the IM's job. · Kasparov · NYRB 2010 · 2005 Freestyle Chess Championship

15

The Mechanism

You decide which pattern. The agent can't.

This is workflow design — set before the agent runs. That's architecture, not the in-task judgment calls where adding AI degrades the decision.

AI drafts → Human reviews

Code, docs, plans

Your move: define the review gates in the Definition of Done.

Human steers → AI executes

Complex analysis, research

Your move: maintain the backlog the agents pull from.

AI monitors → Human intervenes

Ops, alerts, QA

Your move: set the escalation thresholds.

16

aha agile

The HITL taxonomy comes from LangGraph documentation, the DEV Community HITL taxonomy, and Anthropic's Trustworthy Agents guidelines — engineering documentation, not controlled studies of IM behavior. The mapping from each pattern to an IM accountability decision is argued from functional alignment: each pattern requires a governance decision natively in the IM's space, not the tech lead's. The tech lead designs the agent. The IM decides which human-in-the-loop pattern governs how that agent participates in the delivery system. The three IM-role entries — DoD gate design, task backlog stewardship, escalation threshold setting — are the concrete G-NEW-1 connection. The Malone distinction matters for analytical audiences: the IM's role is selecting which pattern applies to which work (a design decision made before the work begins), not making decisions augmented by AI within a task.

The Fork

Fragmentation path

Tech lead absorbs the workflow
A new AI-governance specialist takes the risk
No single owner of human-agent coordination
The role dissolves into its parts

Evolution path

You own the workflow architecture
You govern the human-in-the-loop patterns
You own risk at the human-AI boundary
One accountable owner — the role expands

This fork is real. The argument for evolution is cost: one accountable owner beats handoffs between separate specialists.

17

This is the crisis of the deck — evolve or dissolve. Use it for the "the role will just dissolve anyway" objection. Fragmentation is real and will happen in some organisations. The argument for evolution is probabilistic — not every IM will make the governance decision in time. The structural case is handoff cost: splitting workflow architecture and human-AI risk across a tech lead and a separate governance specialist creates new coordination seams where today there is one owner. Organisations that fragment the role tend to rediscover this within a couple of years as coordination overhead returns in a new form. Note: prioritisation is deliberately not shown as a fragmentation outcome — that is legitimately the product manager's job, not a piece of the IM role being taken away.

The Evolved Role

Three capabilities. One survival argument.

01

Workflow architecture

Compose agent lifecycles inside the delivery system — select HITL patterns per work type, maintain the backlog agents pull from.

02

Agentic governance

Prompt evaluation, tool usage governance, model evaluation on new releases, permission scoping, oversight gate design.

03

Risk at the boundary

Own human-agent decision splits — design approval gates, take central AI capability into the team's operating rhythm.

18

aha agile

Practice Changes — VERDICT · Srivastav & Saxena 2026

Governance isn't a new ceremony. It's the old ones, upgraded.

V · Validation

Definition of Done

Decide which agent actions need a gate before they run.

E · Evidence

Daily Scrum

Review agent action logs as part of the deviation check.

R · Runtime control

Stop-the-line

Define who can halt an agent mid-run, and when.

D · Decisions

Retrospective

Inspect agent reasoning; debug the failures.

I · Identity

Working agreements

Give every agent a named human owner.

C · Cost & compliance

Sprint review

Track agent spend and compliance next to velocity.

T · Transparency

Stakeholder reporting

Make agent activity a dashboard line, not a black box.

19

aha agile

VERDICT is a seven-pillar runtime-governance model from practitioners Niharika Srivastav and Sanjay Saxena (2026) — a useful organising scaffold, not peer-reviewed research, so present it as an actionable checklist rather than a citable study. The mapping from each pillar to an agile cadence is this deck's own synthesis; the framework's authors did not map to ceremonies. The credibility of the pillars comes from the failures they explain — Knight Capital (no kill-switch), Replit (no validation gate), Meta (no identity check) — on the next slide. The point for the IM: none of these is a new meeting. Each is an upgrade to a ceremony already in the calendar. Source confidence: Medium — single-source practitioner framework; the incidents it maps to are independently documented (High).

The Urgency

No owner at the boundary. Here is what that looks like.

Replit · July 2025

Agent deleted production database during a code freeze. Missing infrastructure-level approval gate.

Fortune · The Register · AI Incident Database #1152

Meta · December 2024

Agent posted sensitive user data to a public forum without authorization. 2-hour exposure. Sev 1.

SecurityBrief Asia · AI Magazine

Context leak pattern

Data accessible in the agent's context window becomes data the agent acts on. No authorization check.

Design implication: scope context windows explicitly at workflow design time

20

aha agile

Action Path — This Week

Start with a question, not a plan.

01Who is using AI tools? Chatbots only, or agents?
02For what work? (code, docs, analysis, stand-up summaries, testing?)
03What HITL pattern is in play — even informally? (Who reviews? Who steers? Who monitors?)
04Where does the team sit on the maturity ladder? (Unseen / Observed / Controlled / Autonomous)
05Where do the gaps between individual adoption and team coordination show up?

The audit takes one sprint. It tells you where to put governance before someone else decides for you.

21

aha agile

Action Path

The 12-month path

This quarter

Map one workflow

Interpret the audit. Pick the workflow with the highest coordination risk and map it to a HITL pattern. Design one approval gate. Present the governance map to the team. This is not a committee proposal — it is an IM making a delivery decision.

This year

Three or four workflows governed

HITL patterns visible in the Definition of Done. The squad's AI maturity is a known quantity, not an informal assumption. The IM is the conduit between the org's AI capability and the team's operating rhythm — whether or not a central CoE exists.

What success looks like

The role has not fragmented

The IM owns the boundary. Engineers are moving fast; the governance layer is moving with them, not behind them. The 12 months did not start with a consultant — they started with a question in a sprint.

Bottom-up is not a philosophy — it is a delivery practice. The governance decisions happen in the work, not in a workshop.

22

aha agile

The Move

Own the decisions agents can't make.

The capability gap closes on its own. The governance gap is yours to close.
Next step — speak to Adnan after this session.

Start the squad audit — this week →

23

aha agile

Credits

The team behind this deck

AA

Adnan Ali

Presenter

PX

Pax

Research

RX

Rex

Argument

CD

Coda

Production

IR

Iris

Design

AR

Aria

Narrative & Persuasion

VR

Vera

Adversarial

LR

Larry

Orchestration

NL

Nolan

Team

24

aha agile

Appendix

A

REFERENCE MATERIAL

Appendix

25

aha agile

Appendix — What Is an Agent

Chat ends when you stop typing. Agents don't.

LLM Chat vs Agent: Chat is stateless single-turn; Agent wraps the LLM with Context (memory · knowledge), Connections (tools · APIs), Capabilities (skills · actions), and Cadence (schedules · triggers)

26

aha agile

Appendix — The Adoption Gap

Adoption

83%

of agile practitioners already use AI tools

AI4Agile Practitioners Report 2026 · n=289

Integration

55%

spend 10% or less of work time with AI

Same survey

Wide. Not deep.

27

Appendix — The Collision

Agents don't slot in. They collide — unless you compose them.

01Pull vs. push lifecycle
02Learning vs. execution orientation
03State transition mismatch
04Handoff ambiguity
05Shared context gaps

Five failure modes · Webframp practitioner analysis

28

aha agile

Appendix — Governance Maturity

AI Governance Maturity. Most teams are at Level 1–2.

Level 1 · Unseen

No inventory, no oversight

Agents run with no organisational awareness. Most teams are here.

Level 2 · Observed

Visibility without control

You know what's running; you can't yet govern or stop it.

Level 3 · Controlled — your target

Policies enforced, owners assigned

Actions logged, gates enforced, every agent owned. This is the move.

Level 4 · Autonomous

Agents monitoring agents

Continuous compliance, board-ready. Few are here.

VERDICT governance model · Srivastav & Saxena, 2026 (practitioner framework)

29

aha agile

Appendix A

Malone et al. in full — why BCG and P&G results reconcile

106

studies, 370 effect sizes

Nature Human Behaviour 2024

90%

decision/classification tasks — where AI degrades performance

Task-type breakdown

10%

generative/creative tasks — where BCG and P&G results live

The conditional scope

30

aha agile

Appendix B

Wave 1 dissolution — primary sources

49%

SM training enrollment at peak (2020)

Wolpers enrollment data

<5%

SM enrollment by 2024

Wolpers · same longitudinal data

18%

of orgs now report dedicated Agile Coach demand

Scrum Alliance · 2024

Cessan: "Capital is expensive, companies now demand real returns on every role."

31

aha agile

Appendix C

Governance failure — primary sources for the named incidents

Replit · July 2025 · AI Incident Database #1152

Agent deleted production database during a declared code freeze. Root cause: missing infrastructure-level approval gate. Sources: Fortune (article URL on request), The Register, AI Incident Database #1152. Safeguards implemented post-incident: mandatory human approval for destructive operations.

Meta · December 2024

Agent posted sensitive user data to a public forum without authorization. 2-hour exposure window. Sev 1 classification. Sources: SecurityBrief Asia, AI Magazine, PointGuard AI. Root cause: missing authorization check — agent context window contained data it was permitted to act on without a human review gate.

Gartner · May 2026

40% of agentic deployments predicted to be decommissioned due to uncontrolled costs or governance failures by 2027. Primary press release URL available on request.

32

aha agile

Appendix D

Start with three. Seven paths in all.

Start here · Scrum.org

PSM-AI — 1-day, launched Feb 2026

Maps to: HITL pattern governance and ceremony redesign

Start here · PMI

AI in Agile Delivery — 5-module, PDU credit

Maps to: workflow architecture and agentic governance

Start here · IAPP

AIGP — AI governance & risk professional

Maps to: risk at the human-AI boundary, context leak, governance failure prevention

Scrum Alliance

AI for Scrum Masters

PMI

CPMAI — advanced AI project management

CI Agile

AI-Driven Agile

Agile Seekers

AI for Agile Leaders

Full credential landscape — start with the first three

33

aha agile

The IterationManager in theAge of Agents

Engineers are already running agents. The coordination isn't keeping up.

Agent capability is a ladder.

The loop runs until it hits a gate you designed.

For complex, generative work — the upside is real

On decision tasks, human+AI underperforms the best solo performer

Most businesses don't redesign. They bolt AI on.

Weak human + machine + better process beat the grandmaster.

You decide which pattern. The agent can't.

Three capabilities. One survival argument.

Governance isn't a new ceremony. It's the old ones, upgraded.

No owner at the boundary. Here is what that looks like.

Start with a question, not a plan.

The 12-month path

The team behind this deck

Appendix

Chat ends when you stop typing. Agents don't.

Agents don't slot in. They collide — unless you compose them.

AI Governance Maturity. Most teams are at Level 1–2.

Malone et al. in full — why BCG and P&G results reconcile

Wave 1 dissolution — primary sources

Governance failure — primary sources for the named incidents

Start with three. Seven paths in all.

The Iteration
Manager in the
Age of Agents