AI in development

Turn individual AI use into a secure, measurable development process with shared practices, appropriate governance, and clear next steps.

Discuss this solution

Animated process from ticket through planning and implementation to quality assurance and release

Maturity check

AI in development needs a reliable operating frame

Coding tools and agents can accelerate development work. Their impact becomes sustainable when context, responsibilities, quality boundaries, and risk handling are agreed across the team.

Context

Repositories, architecture, and tasks give AI a reliable basis for work.

Workflow

Tasks, approvals, and stop conditions make agentic work controllable.

Quality

Tests, reviews, and measurable goals keep speed and reliability together.

Where does your development team stand today?

Place your current way of working across twelve dimensions. The result shows more than an average: it highlights the areas where the next improvements can make the biggest difference.

AI-assisted software development maturity explorer

The matrix makes visible how consistently AI is already embedded in your development process – and where shared practices can make the greatest difference.

How to use the explorer

1
Assess each of the twelve dimensions on a scale from 0 to 5 – from no use through to AI-native development.
2
Choose the level that best describes your current practice. This is a shared, honest snapshot, not a test.
3
After the final selection, explicitly open the results to see your overall maturity, the profile across all dimensions, and the three most important next steps.

The result clarifies your starting point and makes the most useful levers for standards, governance, quality, and collaboration easier to prioritize.

0 of 12 dimensions assessed

Dimension	0No use	1Assisted	2Integrated	3Agentic	4Orchestrated	5AI-native development
Typical use	0 · No useAI is not used in this area yet.	1 · AssistedQuestions, explanations, and code snippets.	2 · IntegratedIDE copilots, tests, refactoring, and documentation.	3 · AgenticAn agent handles complete tickets.	4 · OrchestratedSeveral specialized agents work in parallel.	5 · AI-native developmentSoftware development is largely designed as an AI-driven process.
Task scope	0 · No useAI is not used in this area yet.	1 · AssistedIndividual lines of code or functions.	2 · IntegratedFiles and clearly bounded tasks.	3 · AgenticFeatures spanning several files.	4 · OrchestratedEpics and complex changes across several systems.	5 · AI-native developmentFrom requirements to operations across the full lifecycle.
Context understanding	0 · No useAI is not used in this area yet.	1 · AssistedOnly the prompt and pasted code.	2 · IntegratedRepository and open files.	3 · AgenticCode, documentation, tickets, and tests.	4 · OrchestratedCross-project context, architecture, and historical decisions.	5 · AI-native developmentContinuous organizational knowledge with maintained long-term memory.
Workflow integration	0 · No useAI is not used in this area yet.	1 · AssistedA separate chat interface.	2 · IntegratedIDE, CLI, or pull requests.	3 · AgenticTickets are moved into a workspace and branch.	4 · OrchestratedAutomatic planning, delegation, execution, and merging.	5 · AI-native developmentDynamic end-to-end processes with human control points.
Autonomy	0 · No useAI is not used in this area yet.	1 · AssistedPeople perform every step.	2 · IntegratedAI proposes changes.	3 · AgenticAI implements, tests, and opens a pull request.	4 · OrchestratedAgents plan and coordinate subtasks independently.	5 · AI-native developmentWithin defined limits, AI also optimizes processes, tools, and agents.
Tool use	0 · No useAI is not used in this area yet.	1 · AssistedNo tools or manually copied results.	2 · IntegratedGit, terminal, search, and tests.	3 · AgenticStructured tools for code, browsers, databases, and APIs.	4 · OrchestratedDifferent agents with specialized toolsets.	5 · AI-native developmentDynamic tool selection and automatic provisioning of required infrastructure.
Quality assurance	0 · No useTraditional reviews.	1 · AssistedManual review of AI output.	2 · IntegratedAutomated tests and linting.	3 · AgenticAn agent runs tests, static analysis, and self-review.	4 · OrchestratedIndependent implementation, review, and security agents.	5 · AI-native developmentContinuous risk-based validation with measurable quality objectives.
Human role	0 · No useDeveloper.	1 · AssistedExecutor with AI assistance.	2 · IntegratedAccountable practitioner.	3 · AgenticReviewer and decision-maker.	4 · OrchestratedOrchestrator and approver.	5 · AI-native developmentShapes goals, rules, architecture, and exceptions.
Governance	0 · No useNo AI rules.	1 · AssistedInformal rules.	2 · IntegratedApproved tools and data-protection rules.	3 · AgenticPermissions, sandboxing, and audit logs.	4 · OrchestratedPolicies per agent, project, data source, and action.	5 · AI-native developmentAutomatically enforced governance with risk and escalation models.
Infrastructure	0 · No useStandard development setup.	1 · AssistedSaaS chat.	2 · IntegratedIDE integration and central models.	3 · AgenticIsolated agent runtimes.	4 · OrchestratedKubernetes, agent factory, model routing, and observability.	5 · AI-native developmentDemand-driven, self-optimizing agent infrastructure.
Measurement	0 · No useAI is not used in this area yet.	1 · AssistedSubjective time savings.	2 · IntegratedUsage and adoption are measured.	3 · AgenticLead time, cost, and success rate per task.	4 · OrchestratedComparison of agents, models, prompts, and skills.	5 · AI-native developmentContinuous optimization using business and engineering KPIs.
Organizational adoption	0 · No useIndividual experiments.	1 · AssistedIndividual power users.	2 · IntegratedTeam-wide standards.	3 · AgenticSeveral production teams.	4 · OrchestratedCompany-wide platform.	5 · AI-native developmentThe development organization is structurally designed for human–AI collaboration.

Complete every dimension to see the overall score, maturity band, and prioritized next steps.

From individual use to a controlled development process

A maturity check is not a competition or a certificate. It creates a shared view of how AI is used today and which boundaries should apply to repositories, data, architecture, and release processes.

That foundation separates standards from experiments: teams can try new models and agents while review, testing, security rules, and human responsibility remain explicit.