From M-Scale Tools to Autonomous Agent Team Members
Transform engineering teams from human-only structures to hybrid human-agent teams where autonomous specialist agents function as actual team members—taking ownership of specific domains, participating in planning, and contributing to team success alongside human colleagues.
The classic M-scale covers M1–M5, where humans use AI as a tool. The extended scale goes further — where AI becomes a peer, then a domain owner.
TOOL USE TEAM MEMBER
◄─────────────────────────────────────────────────────────────────────────►
M1 M2 M3 M4 M5 │ M6 M7 M8
│ │ │ │ │ │ │ │ │
▼ ▼ ▼ ▼ ▼ │ ▼ ▼ ▼
Manual AI- AI AI AI │ Bounded Domain Autonomous
work assisted automates orchestrates native │ Agent Owner Teammate
tasks workflows processes teams │ Roles w/ Input & Peer
│
Augmented Teams ◄── New Vision ──►
| Level | Label | Human Role | Agent Role |
|---|---|---|---|
| M1 | Manual | Does everything | None |
| M2 | AI-Assisted | Leads, AI helps | Suggestion, draft, autocomplete |
| M3 | AI-Automated | Reviews, AI executes | Task execution within workflows |
| M4 | AI-Orchestrated | Approves, AI manages | Runs multi-step processes |
| M5 | AI-Native Teams | Directs strategy | Full workflow ownership |
| M6 | Bounded Specialist | Oversees, escalates | Owns scoped domain, reports back |
| M7 | Domain Owner | Sets direction | Makes tactical decisions, advises |
| M8 | Autonomous Peer | Strategic leadership | Proactive, self-directed, learning |
Current M-Scale (M1–M5): Humans use AI tools → AI augments work → AI automates workflows → AI orchestrates processes
Extended Vision (M6–M8): AI becomes actual team members with defined roles, responsibilities, and autonomous decision-making within bounded contexts.
North Star: Teams where product and engineering leads manage hybrid squads of humans and agents, each with specialised roles, working together as integrated teammates rather than humans directing AI tools.
Traditional Team (M1–M3) Hybrid Specialist Team (M6–M8)
───────────────────────── ──────────────────────────────
Product Manager Product Manager
Engineering Manager Engineering Manager
Engineer × 6–8 Senior Engineer × 2–3
Designer ┌─────────────────────────────┐
QA │ Autonomous Agents │
DevOps │ • SRE Agent │
│ • QA Agent │
10–12 humans │ • Frontend Agent │
Sequential, 40h/week │ • Backend Agent │
│ • Product Agent │
└─────────────────────────────┘
5–8 total (humans + agents)
24/7 operational capability
- M6: Task-specific agents (run tests, deploy code, monitor alerts)
- M7: Workflow-specific agents (handle complete feature development)
- M8: Domain-specific agents (own entire technical areas with strategic input)
- M6: Execute predefined actions within clear parameters
- M7: Make tactical decisions within bounded technical contexts
- M8: Make strategic technical decisions with business impact awareness
- M6: Report status and escalate issues to humans
- M7: Collaborate with humans and other agents on shared objectives
- M8: Lead technical initiatives with human strategic oversight
- M6: Follow established patterns and escalate exceptions
- M7: Adapt approaches based on team feedback and outcomes
- M8: Proactively identify improvements and optimise team performance
- 24/7 Coverage: Agents provide round-the-clock monitoring, incident response, and development work
- Consistency: Standardised approaches across all agent work, reducing human error and variation
- Scalability: Add specialised agents faster than hiring and training human team members
- Strategic Focus: Humans concentrate on high-value creative and strategic work
- Reduced Toil: Elimination of repetitive tasks and routine maintenance work
- Enhanced Creativity: More time for innovation, architecture, and complex problem-solving
- Faster Time-to-Market: Parallel agent work streams accelerate development cycles
- Higher Quality: Consistent agent testing and monitoring reduces production issues
- Cost Efficiency: Massive productivity gains with optimised team composition
- Agent framework development and tooling maturity
- Robust monitoring and observability for agent performance tracking
- Secure agent-to-system integrations with appropriate access controls
- Comprehensive testing frameworks for agent decision validation
- Clear communication about human-agent collaboration, not replacement
- Training programs for humans to work effectively with agent teammates
- Performance evaluation frameworks that account for hybrid team dynamics
- Cultural adaptation to viewing agents as team members, not just tools
- Gradual rollout with careful monitoring of agent decision quality
- Human override capabilities for all critical business decisions
- Comprehensive audit trails for agent actions and decision-making
- Regular review and optimisation of agent performance and boundaries
A vision for the future of work — where humans and AI agents collaborate as true teammates to deliver exceptional outcomes at unprecedented scale and speed.