Skip to main content

2026 AI series: When AI becomes a team member

Rethinking team design and delivery

Authors: Julene Marr and Mette Mikkelsen

It’s not about using AI tools, it’s about how we lead and design teams where AI is an embedded team member — not just a plug-in.

In our latest article, we explore:

  • How to design for human-AI collaboration
  • Where governance needs to evolve beyond AI approvals
  • Why AI stewardship, prompting, quality assurance and fluency are the new must-have skills across tech delivery roles
  • And what questions every tech delivery leader should be asking right now

It’s not hype. It’s about how we scale responsibly, adapt faster — and lead smarter.

The rise of AI-enabled delivery teams

Constantly evolving across industries, AI is rapidly becoming more than just a tool — it’s becoming part of the team. From intelligent testing and code generation to warehouse orchestration and predictive logistics, AI is changing how delivery happens, who delivers it, and how success is measured.

Reflecting on a workshop series we recently ran on the incorporation of GenAI into the way technology teams work, it became clear that we need to start having the discussion today about how to shape tech teams in the future. We also need to support our people to make that transition.

While most conversations in 2025 have focused on AI’s impact on productivity or fears of role replacement, a more fundamental shift is underway: What does a high-performing delivery team look like when AI is embedded into the core of how it operates?

Traditional vs. AI-enabled teams: A new architecture for delivery

We’re witnessing a structural evolution in how delivery teams form, operate, and make decisions:

Team element

Traditional delivery team

AI-enabled delivery team

Roles

Stable, human-defined (developer, tester, ops, scrum master)

AI tools take on select delivery tasks (code, test, plan) alongside humans (who focus more on strategy, quality assurance and prompting)

Tooling

Orchestrated by humans

Orchestrated with and by AI (e.g., GitHub Copilot, ML observability platforms)

Planning

Sprint-driven, backlog-oriented

Dynamic, predictive, AI-influenced prioritisation

Decision-making

Role-led, collaborative

Shared between humans and AI agents, with growing reliance on model outputs, and clear governance for ‘who does what’. In the US we are seeing a ratio of one human to four agents and forecast to be one to 30 by 2026 and one to 100 in 2030.

Structure

Persistent, cross-functional pods

Human-AI hybrid teams, flexibly augmented with services or AI agents.

This new model isn’t about replacing humans, it’s about rethinking team composition and capability.

What organisations are not yet talking about
While AI use cases are proliferating, key gaps remain in how organisations design and govern AI-augmented teams:

  • Team operating models and organisational design are not yet integrating AI – how do you design a hybrid human/AI team?
  • New roles and responsibilities like prompt engineers, model reviewers, and AI delivery leads are still emergent – what roles need to be reshaped, and how?
  • Accountability and oversight are unclear when AI tools make delivery-impacting decisions – who governs AI, and what is the best way to do this?
  • Trust and fluency with AI in team settings remain low outside of early adopters – how do you upskill a workforce for AI?

Designing high-performing, AI-augmented delivery teams
To stay ahead, delivery leaders should intentionally redesign how teams work with AI.

This can be done with four key steps:

1. Define the role of AI in the team
Is AI a junior assistant, a peer reviewer, or a specialist? And what does this mean for human roles – do they focus more on prompting or quality assurance? Being clear about AI’s function creates clarity about human versus AI roles and helps align expectations, workflows, and validation processes.

2. Create hybrid roles to bridge AI-human interactions
Introduce responsibilities such as:

  • AI delivery lead: ensures integration aligns to value delivery
  • Prompt architect: curates effective, safe prompts for consistent AI outputs
  • AI risk steward: oversees governance, bias, and ethical use of AI in delivery
  • AI quality assurance specialist: provides quality assurance of AI outputs.

3. Embed governance into delivery rituals

  • Ensure that you have fit-for-purpose AI governance in place
  • Include model validation in the definition of ‘done’
  • Review AI-generated content in sprint retros and QA gates
  • Document AI logic and decision inputs during planning.

4. Upskill for fluency, not just usage
Success doesn’t just depend on knowing the tools, it relies on knowing how to work with AI:

  • Prompt crafting
  • Interpreting AI outputs
  • Quality assuring AI outputs
  • Managing uncertainty
  • Knowing when not to use AI.

Where to start
Consider these steps as starting points:

  • Pilot AI as a team “member” in one delivery cycle — review in sprint retro
  • Run an “AI Role Canvas” exercise as part of team chartering
  • Add AI validation to your CI/CD or release gates
  • Measure outcomes beyond speed: track decision quality, trust, and transparency

Anecdotal insights from our recent AI workshop series identified that one of the most common challenges organisations face was change management for teams who are impacted by AI in addition to governance and controls of AI use. Traditional tech teams may not be well-versed in change management and governance design. However, when managed well, and with thoughtful team design and fit-for-purpose governance, AI has the potential to significantly uplift productivity and supercharge teams.

Final thought
AI isn’t replacing delivery teams, it’s reshaping them. Those who thrive will be the teams that intentionally redesign how humans and AI collaborate, make decisions, and deliver value together.

AI is here to stay. The question is, are our teams ready to work alongside it?

If you’d like support discussing change management or teams of the future, please connect with Mette Mikkelsen or Julene Marr.

Did you find this useful?

Thanks for your feedback