The AI-SDLC Maturity Assessment is designed to measure the gap between feeling AI-driven and actually being AI-driven. A developer installs an AI coding assistant and accepts a few autocomplete suggestions. A business analyst pastes meeting notes into ChatGPT and summarises them. A QA engineer generates a handful of test cases from a prompt. The team reports, "We're using AI." These are real activities. But they don't add up to a mature, AI-assisted way of working.
When teams self-assess based on feeling rather than evidence, the result is predictable. Misallocated investment. Unrealistic timelines. Missed opportunities for genuine improvement.
The AI-SDLC Maturity Assessment does this through two complementary lenses. Individual tool proficiency is necessary but not sufficient — team-level infrastructure, governance, and cross-role integration are what determine whether AI adoption is real or performative.
Where AI-assisted sits in the maturity model
| Level | Name | Core relationship |
|---|---|---|
| 1 | Traditional | Human-only; no AI involvement |
| 2 | AI-supported | Human-led; AI provides reactive suggestions |
| 3 | AI-assisted | Human-led; AI contributes proactively |
| 4 | AI-native | Human-guided; AI as active collaborator |
| 5 | AI-autonomous | AI-initiated; human oversight for strategic decisions |
Each level of the AI-SDLC maturity model represents a different relationship between human expertise and machine capability.
Why focus on Level 3? Most teams today operate between Level 1 (Traditional) and Level 2 (AI-supported). Very few genuinely approach Level 3 (AI-assisted). Calibrating the assessment to be AI-assisted makes it practical. It measures the realistic adoption frontier. It also provides a clear signal when a role or team is ready to advance toward being AI-native.
What AI-assisted really means in practice
The defining shift from Level 2 (AI-supported) to Level 3 (AI-assisted) is the move from reactive to proactive AI participation.
- At Level 2, AI responds only when prompted. It offers autocomplete in a single file, answers direct questions, and runs surface-level checks. The person initiates every interaction.
- At Level 3, AI anticipates needs across a broader context. It works with multi-file and cross-artefact awareness. It generates substantial outputs. It participates in planning, review, and quality assurance. This applies equally to a developer refactoring across services, a BA drafting requirement specs from stakeholder notes, a QA engineer building regression suites from acceptance criteria, or an architect producing ADR drafts.
How to recognise an AI-assisted team
- Human-led work with proactive AI contributions. AI participates actively across role tasks, but every decision remains with the human.
- Cross-artefact context awareness. AI understands how artefacts relate: code to tests, requirements to user stories, and technical design to implementation.
- Substantial AI-generated artefacts. Function implementations, test suites, requirement summaries, review recommendations, and deployment strategies.
- Mandatory human review. All AI-generated output must be validated before integration.
The critical constraint: trust does not replace verification. Every AI-generated artefact still requires human validation. The artificial intelligence suggests, generates, and recommends. The human decides.
Why teams misjudge their AI maturity
- Tool presence does not equal process adoption. Having AI tools available, even using them daily, does not mean the team operates in AI-assisted mode. If individuals prompt AI for isolated tasks with no shared practices, no configured workspace, and no governance, the team is at Level 2. That holds regardless of tool usage volume.
- Individual champions do not equal team maturity. One developer with an advanced AI workspace does not make the team AI-assisted. If that person's practices aren't shared, documented, and adopted across other roles, the team has pockets of adoption, not maturity.
- Speed without repeatability does not equal maturity. Producing more output with AI, without being able to explain or verify it, is not a sign of maturity. Some call this "vibe coding". Velocity without verification creates risk, not value.
- The perception gap is real. Developers consistently overestimate AI's impact on their productivity. This often leads teams to believe they are operating at a higher level of AI maturity than they are. As a result, planning decisions, productivity forecasts, and adoption strategies can become misaligned with real outcomes.
Why you need both role-level and team-level scores
A team can have several individuals scoring 7 to 8 on the role-level scale while scoring only 4 to 5 as a team. The reason: no shared knowledge base, no documented workflows, no governance, and no cross-role integration. Conversely, a team with strong governance and shared practices but low individual proficiency won't see the benefits either.
Individual skill and team infrastructure are both necessary. Neither alone is sufficient.
The model also defines a clear ceiling for AI-assisted maturity. Reaching 10/10 on either scale means the role or team has fully realised AI-assisted capabilities. Beyond this point, standalone AI agents outside the IDE, autonomous workflows, and organisational context awareness indicate progression toward AI-native (Level 4).
What a 10/10 AI workspace looks like
The highest score on the role-level scale represents a fully configured AI workspace. This workspace is the primary work interface for the role. It includes:
- A dedicated IDE (Cursor, Copilot, Claude Code, or equivalent)
- Configured rules (project-level, role-specific conventions and constraints)
- Created and used skills (reusable prompt-driven workflows for recurring tasks)
- Configured subagents for routine multi-step flows
- MCP (Model Context Protocol) integrations connecting to relevant external systems (Jira, Confluence, Git, CI/CD, databases)
- Knowledge base/context pointing to project documentation, standards, and prior work
- Structured planning (task decomposition, "plan then execute" discipline, retrospective review of plan vs. outcome)
- Repeatable, validated workflows covering core role responsibilities
Role-level AI maturity scale (1–10)
| Score | Phase | Key characteristics |
|---|---|---|
| 1 | No usage | AI tools not used or access is blocked |
| 2 | Awareness | Tool available; occasional ad-hoc usage; no established pattern |
| 3 | Basic individual use | Simple, isolated tasks; generic prompts; single file scope; no project context |
| 4 | Task-level support | Regular use for small scope tasks; limited verification; no shared practices |
| 5 | Structured use | Context-aware prompts; multi-file work; reviews AI output before applying |
| 6 | Repeatable workflows | Personal repeatable workflows for recurring tasks; non-trivial work covered; personal prompt templates |
| 7 | Codified workflows | Project rules configured; reusable skills created; conventions documented and reproducible |
| 8 | Orchestrated automation | Multi-system integrations; subagents for routine flows; MCP connections; cross-system workflows |
| 9 | Optimised mastery | All core tasks AI-assisted; quality gates enforced; curated skill library; traceability; metrics |
| 10 | Mastery (AI-assisted ceiling) | AI Workspace is the primary work interface; continuous improvement loop; measurable impact |
The critical transition is from 4 to 5. It marks the shift from isolated, ad hoc AI usage to structured, context-aware work with repeatable workflows. This boundary is the entry point into AI-assisted (Level 3). The second key transition is from 6 to 7, where personal workflows become shared team infrastructure.
Team-level AI maturity assessment
Team-level maturity measures capabilities that no individual can provide alone: shared knowledge, common practices, governance, metrics, and cross-role integration. The AI-SDLC Maturity Assessment scores six independently weighted dimensions and maps the result to a named maturity level.
6 building blocks of team AI maturity
| Dimension | What it measures |
|---|---|
| 1. Tooling alignment | Common AI tooling selection and project-level configuration across the team |
| 2. Knowledge base | Shared knowledge infrastructure: Git versioned rules, skills, specifications, ADRs, templates, RAG under knowledge index |
| 3. Practices & workflows | Documented, followed AI-driven workflows across roles |
| 4. Guardrails & governance | Data handling policies, review expectations, quality gates for AI output |
| 5. Metrics & feedback loop | Collection and use of AI adoption metrics; continuous improvement |
| 6. Cross role integration | AI-enhanced handoffs and workflows spanning multiple roles |
Team-level maturity scale (1 to 10)
The full team-level scale provides observable evidence for each score. This makes it easier to calibrate where a team actually stands:
| Score | Level | What it looks like |
|---|---|---|
| 1 | No adoption | Traditional SDLC; no AI tooling in use |
| 2 | Individual experiments | Scattered ChatGPT/Copilot usage by a few individuals; no team coordination |
| 3 | Emerging awareness | AI topic raised in retros/planning; 2 to 3 people use tools regularly |
| 4 | Pockets of practice | Multiple roles use AI weekly; personal templates exist; no shared knowledge base or rules |
| 5 | Shared foundations | Team agreed on primary AI tool; shared repo for AI practices; initial rules configured |
| 6 | Active knowledge base | Shared repo with rules, skills, templates; regular updates; some automated flows |
| 7 | Standardised practices | Documented AI workflow guidelines; guardrails for data handling; AI onboarding; cross role AI workflows |
| 8 | Integrated governance | AI usage metrics tracked; quality gates for AI output; automated SDLC flows; regular review of practices |
| 9 | Optimised operations | Data driven optimisation; team experiments and iterates; cross team knowledge sharing; ROI evidence |
| 10 | Mastery (AI-Assisted ceiling) | Near universal AI adoption for applicable tasks; self sustaining improvement loop; replicable model; measurable business impact |
How scores map to AI-SDLC maturity levels
Both the role-level and team-level scales map to the broader AI-SDLC maturity model:
| Score Range | AI-SDLC level | What it means |
|---|---|---|
| 1 | Level 1 (Traditional) | No AI usage in work |
| 2 to 4 | Level 2 (AI-supported) | Passive AI assistance: autocomplete, basic prompts, reactive suggestions |
| 5 to 10 | Level 3 (AI-assisted) | Proactive AI participation, workspace concept, shared knowledge base |
| Beyond 10 | Level 4 (AI-native) | Full AI agents outside IDE, autonomous collaboration |
The boundary between Level 2 and Level 3 (score 4 to 5) is the most important transition. It marks the shift from ad hoc, isolated AI usage to structured, context-aware work with repeatable workflows.
The boundary between Level 3 and Level 4 (beyond 10) is aspirational for most teams today. When a role or team consistently demonstrates standalone AI agents operating outside the IDE, autonomous workflows across IDE phases, and organisational (not just project) context awareness, they have moved beyond what the 1 to 10 scale measures.
Signs your team is not yet at AI-assisted
Teams often confuse activity with maturity. These patterns indicate Level 2 (AI-supported) or below, even when teams believe they are further along:
- Scattered tool usage. Multiple people use AI tools, but there are no shared practices, no common configuration, and no shared knowledge base.
- Individual champions without team infrastructure. One or two advanced users carry AI adoption while the rest of the team operates traditionally.
- Ad-hoc prompting without repeatable workflows. AI is used for one-off tasks with generic prompts. There is no configured workspace, no project rules, and no skills.
- No governance or quality gates. AI-generated output has no special review process. Team members decide individually whether to check it.
- No metrics beyond anecdotes. The only evidence of AI's impact is "it feels faster". There is no time-saving data, no quality comparisons, and no adoption tracking.
- No cross-role integration. Each role uses AI independently. AI artefacts don't flow from one role to another. For example, BA-generated stories aren't used as context for development.
If a team recognises three or more of these patterns, the AI-SDLC Maturity Assessment provides a structured way to understand exactly where they stand and what to improve first.
Signs a team is ready to move beyond AI-assisted
A team that consistently scores 9 to 10 across both scales and demonstrates the following indicators is approaching AI-native (Level 4):
- Standalone AI agents outside the IDE: Agents that operate as services or bots with access controls and audit trails.
- Autonomous agent workflows across SDLC phases: Requirement, code, test, and deploy flows where agents handle substantial execution.
- Agents with session memory: AI agents that maintain context across sessions and work as persistent teammates.
- Organisational context awareness: AI understands not just the project but also broader business rules, compliance requirements, and strategic priorities.
- Sprint planning includes AI agent capacity: Teams plan work with explicit consideration of what agents will handle.
- Tiered approval model: Routine changes can be auto-merged when checks pass. Complex changes require human review.
These indicators represent a qualitative leap. AI moves from being a tool within the IDE to a collaborator with defined responsibilities.
How to run the assessment across your team
- Role-level assessment. Each team member completes the individual questionnaire. It takes 10 to 15 minutes.
- Team-level assessment. The tech lead or PM evaluates six team dimensions.
- Review results. The assessment provides scores, maturity-level mapping, and priority improvement areas.
- Reassess quarterly. Track progress over time and adjust the action plan based on evolving scores.
Common pitfalls of AI adoption
Adopting AI at the team level is harder than it appears. Honest assessment requires acknowledging real constraints.
- Inconsistent adoption between roles. Developers often adopt AI tools faster than BAs, QAs, or designers. A team with developers at score 7 to 8 and other roles at score 3 to 4 has a fragmented adoption profile. The team-level score will reflect this.
- Shallow or tool-driven usage without process depth. Installing an advanced AI tool and using its basic features does not constitute maturity. Many teams plateau at scores 3 to 4 because they never invest in workspace configuration, rules, or skills.
- Lack of measurable outcomes. Without metrics, teams cannot distinguish between "AI makes us feel productive" and "AI measurably improves our delivery". The perception gap is well-documented. Developers believe AI makes them faster even when measurements show otherwise.
- Context drift and poor knowledge inputs. AI tools are only as good as the context they receive. Outdated documentation, incomplete project rules, or poorly maintained knowledge bases produce low-quality AI output. Teams then blame the tools rather than the inputs.
- Quality risks are hidden by perceived speed. AI-generated code suggestions may contain security vulnerabilities. Without proper security checks and validation processes, faster delivery can lead to the accumulation of hidden technical debt over time.
- The delivery stability paradox. The DORA 2024 report highlights that AI adoption is associated with mixed effects on software delivery performance. While teams report improvements in individual productivity and workflow efficiency, increased AI usage is also associated with a measurable decrease in delivery stability.
- Code quality degradation. According to GitClear's 2025 AI Copilot Code Quality Report — an analysis of 211 million lines of code written between 2020 and 2024 — refactoring declined from 25% of changed lines to less than 10%, while duplicated code blocks increased 8-fold. It reinforces the idea that producing more code does not necessarily lead to better code quality or maintainability.
- Dependency on a few AI-heavy individuals. When AI adoption concentrates in one or two people, the team's capability is fragile. If those individuals leave, the practices leave with them.
These constraints don't argue against AI adoption. They argue for structured, measured, honest adoption. That is precisely what the AI-SDLC Maturity Assessment enables.
Conclusion
AI-assisted (Level 3) represents the practical frontier for most software delivery teams today. It's the level where AI moves from reactive suggestions to proactive contributions. Real productivity gains become achievable through structured adoption.
But reaching this level requires more than installing tools. It requires configured workspaces. Shared knowledge bases. Documented workflows. Governance guardrails. Measurable outcomes. Cross-role integration. The gap between using AI and being AI-assisted is a gap between individual activity and team capability.
The AI-SDLC Maturity Assessment provides a practical way to answer the question honestly: Where are we, and what should we improve next? It evaluates both role-level proficiency and team-level infrastructure. It provides the structured evidence needed to turn AI adoption from an aspiration into a measurable, improvable capability.
Related Insights
The breadth of knowledge and understanding that ELEKS has within its walls allows us to leverage that expertise to make superior deliverables for our customers. When you work with ELEKS, you are working with the top 1% of the aptitude and engineering excellence of the whole country.
Right from the start, we really liked ELEKS’ commitment and engagement. They came to us with their best people to try to understand our context, our business idea, and developed the first prototype with us. They were very professional and very customer oriented. I think, without ELEKS it probably would not have been possible to have such a successful product in such a short period of time.
ELEKS has been involved in the development of a number of our consumer-facing websites and mobile applications that allow our customers to easily track their shipments, get the information they need as well as stay in touch with us. We’ve appreciated the level of ELEKS’ expertise, responsiveness and attention to details.
