Agile Frameworks - Embedding Agentic AI into Scrum
About The Author & The Article
Jonathan Bishop, Group Chairman, Bishop Phillips Consulting. [1]
Copyright 2020-2026 - Moral Rights Retained.
This article may be copied and reprinted in whole or in part, provided that the original author and Bishop Phillips Consulting is credited and this copyright notice is included and visible, and that a reference to this web site (http://RiskWiki.bishopphillips.com/) is included.
This article is provided to the community as a service by Bishop Phillips Consulting www.bishopphillips.com.
Introduction
Embedding agents into Scrum teams means integrating autonomous AI systems as bounded‑role collaborators that support backlog refinement, planning, development, testing, documentation, risk management, and flow optimisation while humans retain all strategic and ethical accountabilities. It is a mixture of both Agentic AI (AI coordinating and running other agents) and embedded AI Agents (humans coordinating and running specific AI Agents).
The obvious first step is that, at least in IT projects, the software developer or systems implementer roles become coordinators of a team of AI code‑writers, but this is only one small slice of what embedding agents into Scrum teams really means. What’s coming is far bigger:
Scrum teams will evolve into hybrid human–AI systems, where agents participate in every Scrum event, every artifact, and every workflow — not as tools, but as team members with bounded autonomy.
In this paper we explore this strategy and how to implement it.
The Core Idea
Embedding agents into Scrum teams means giving AI explicit roles, responsibilities, and boundaries inside the Scrum framework - without violating Scrum’s human‑centric accountabilities.
In this model:
- Humans still own the *accountabilities*.
- Agents take on *capabilities*.
This distinction is crucial. In the real world a human (manager) must own the risk and be accountable for the outcome of the team he or she manages. It is not acceptable for the manager to say "Ah, that mistake was one of the team members, so not my responsibility." He is responsible for the governance of the team and everything the team does because the manager:
- approve, re/defines and (possibly) selects the destination/ objective / strategy,
- selects the team & resources to deploy or how they are deployed,
- decides the training & skills the team requires to perform the work,
- inspires & motivates the team,
- prioritizes team activities,
- tunes & monitors performance (efficiency and effectiveness),
- owns the consequences of all these decisions made.
This is true whether the team is 100% human, a mixture of human and machine, or 100% AI. That is the manager's role.
What changes when we augment the human 'worker' team member with AI agents is that every human worker becomes a manager of a little automated workforce. Whereas before he or she just administered themselves and their own workload, now they are administering a team of separate AI minds to do that work at much higher speeds and depth than they could do before on their own.
The correct model going forward will be that every worker with agent augmentation will need the at least some of the skills of management that were previously reserved for those in human supervisory positions. Indeed, while it should be obvious that directing (prompting) the Agent is a required skill transported from the management layer to the worker, so also are the verification (testing) and monitoring skills - something that may need to be specifically trained and equipped into the humans agentic workflow.
With that shift in focus must come the recognition that, just as before, the human was responsible for his or her output, that does not change just because that output is now produced by a team of AI agents working for that human.
The Agent Skill Master
1. Where Agents Fit Inside the Scrum Roles
A. Product Owner (PO)
This role becomes AI‑Augmented, not AI‑Replaced. The PO remains human, but agents become a Product Owner’s staff.
Agents can:
- Secretarial functions:
- Schedule meetings and draft/read team, customer, Project Manager and Steering Committee communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with customer, project manager and steering committee AI liaison agents
- Analyse customer behaviour
- Interpret customer feedback and update product backlog details and prepare supporting materials
- Undertake ad-hoc research
- Propose backlog items
- Draft acceptance criteria
- Estimate value and risk
- Simulate ROI scenarios
- Detect dependencies
- Monitor regulatory changes
- Suggest backlog ordering
The human PO:
- Makes decisions
- Sets priorities
- Owns value
- Ensures ethics, strategy, and context
- Agent Skill Master
This turns the PO into a strategic decision‑maker, not a backlog administrator.
B. Scrum Master
The Scrum Master becomes an AI‑Enhanced Flow Coach, but remains human. The agents fill the roles of secretary, flow analyst and impediment detector.
Agents can:
- Secretarial functions:
- Schedule meetings and draft/read team and Project Manager communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with product owner, project manager and developer AI liaison agents
- Train staff in team processes and policies
- Monitor cycle time, WIP, throughput
- Detect bottlenecks
- Identify anti‑patterns (e.g., too much WIP, stalled PBIs)
- Suggest process improvements
- Analyse team sentiment (retrospective input)
- Monitor and flag risk events early
The human Scrum Master:
- Coaches
- Facilitates
- Inspires & protects the team
- Drives organisational change
The Scrum Master becomes a systems thinker, supported by AI diagnostics and secretarial functions.
C. Developers
This is the biggest transformation as this sees the developer transform into a Human–AI Hybrid Delivery Team. The previous two roles were already managerial functions, but this role is the one that transforms from a sole operator into a team manager.
Agents become:
- Secretarial functions:
- Schedule meetings and draft/read team and Scrum Master communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with user, scrum master and other developer AI liaison agents
- Skills trainer & and interactive help library
- Code generators
- Test writers
- Debuggers
- Documentation writers
- Refactoring assistants
- Architecture evaluators
- Security scanners
- Data analysts
Humans adopt the role of a senior dev:
- System designers
- Integrators
- Developer agent supervisor & debugging director
- Code Reviewer
- Decision‑makers
- Problem framers
- Ethical & standards overseers
Developers evolve into senior orchestrators of agentic work but with far more breadth. It is probable that the velocity metrics change under this model with higher story point throughput and greater complexity being able to be covered in the same Sprint as before. Team sizes should ideally stay approximately the same as before as this is a function of the human supervisor to staff ratio rather than the velocity to supervisor ratio.
2. Where Agents Fit Inside Scrum Events
Sprint Planning
Agents:
- Secretarial functions
- Schedule meetings and draft/read team communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with team AI liaison agents
- Propose Sprint Goal options
- Suggest PBIs based on value + capacity
- Estimate complexity
- Identify risks
- Generate initial task breakdowns
Humans:
- Choose the Sprint Goal
- Select PBIs
- Validate the plan
Daily Scrum
Agents:
- Secretarial functions
- Schedule meetings and draft/read team communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with team AI liaison agents
- Summarise progress
- Highlight blockers
- Predict whether the Sprint Goal is at risk
- Suggest adjustments
Humans:
- Decide what to do
- Coordinate
- Adapt the plan
Sprint Review
Agents:
- Secretarial functions
- Schedule meetings and draft/read team communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with team AI liaison agents
- Generate demo scripts
- Summarise metrics
- Analyse stakeholder feedback
- Propose backlog updates
Humans:
- Present the Increment
- Engage stakeholders
- Make decisions
Sprint Retrospective
Agents:
- Secretarial functions
- Schedule meetings and draft/read team communications
- Book facilities
- Prepare reports and summaries
- Record meeting transcripts
- Liaise with team AI liaison agents
- Analyse flow metrics
- Detect patterns in team behaviour
- Suggest improvements
- Identify systemic issues
Humans:
- Reflect
- Decide
- Commit to improvements
3. Where Agents Fit Inside Scrum Artifacts
Product Backlog
Agents maintain:
- PBI descriptions
- Acceptance criteria
- Estimates
- Dependencies
- Value scoring
- Risk scoring
- Refinement suggestions
The backlog becomes a **living, self‑maintaining system**.
Sprint Backlog
Agents:
- Break PBIs into tasks
- Estimate task effort
- Monitor task progress
- Update burndown automatically
Humans:
- Validate
- Adjust
- Execute
Increment
Agents:
- Generate code
- Test code
- Validate DoD compliance
- Produce documentation
- Perform static analysis
Humans:
- Review
- Integrate
- Approve
4. The Boundaries (what agents must NOT do)
To preserve Scrum’s integrity Agents must NOT:
- Make product decisions
- Prioritise the backlog autonomously
- Commit to Sprint Goals
- Approve increments
- Replace human accountability
- Override ethical or strategic judgement
Agents are capability providers, not accountability holders.
5. The Deeper Transformation
Under this model Scrum becomes a *coordination protocol* for humans + agents. Scrum’s structure of roles, events and artifacts becomes the coordination layer for a hybrid workforce.
Humans provide:
- judgement
- ethics
- creativity
- strategy
- context
Agents provide:
- speed
- analysis
- automation
- consistency
- continuous monitoring
This, I believe, is the future of Agile delivery.
6. The new hybrid team model
Human Team Members
- Product Owner
- Scrum Master
- Senior Developers / Solution Architects
Embedded Agents
- Backlog Agent
- Planning Agent
- Coding Agent
- Testing Agent
- Documentation Agent
- Risk Agent
- Flow Agent
- Compliance Agent
Each agent has:
- a defined scope
- a bounded autonomy level
- a human owner
- a clear interface
This is the **Agent‑Augmented Scrum Team**.
Next Steps
- Define the **Agent Roles** formally (Backlog Agent, Flow Agent, Risk Agent, etc.)
- Design the **Agent–Human interaction model**
- Create a **Scrum Team Operating Model with Agents**
- Map this into **SAFe / Enterprise Agile**
- Show how to build an **AI‑augmented Definition of Done**
- Show how to embed agents into **RiskManagement**
