ASPICE Guide

ASPICE Levels Explained: CL0 to CL5 Capability Levels

Automotive SPICE (ASPICE) defines six capability levels — from CL0 (Incomplete) to CL5 (Innovating) — that benchmark how well an organization performs each process in the V-cycle. This guide explains every level, the Process Attribute (PA) scoring rules behind them, and where AI-driven automation accelerates the climb.

What are ASPICE Capability Levels?

ASPICE assesses each process (e.g. SYS.2 System Requirements Analysis, SWE.3 Software Detailed Design, SUP.10 Change Request Management) against a capability dimension. Each level adds a layer of discipline on top of the previous one: you cannot reach CL2 until CL1 is fully achieved, and so on. OEMs commonly require CL2 across critical processes for series production, with Tier-1 suppliers targeting CL3.

CL0

Incomplete

The process is not implemented or fails to achieve its purpose. There is little or no evidence of any systematic achievement of the process outcomes.

Expected outcomes

No defined or repeatable activities
Outcomes are accidental, not engineered
Audit evidence is missing or contradictory

Process Attributes (PA)

• No PA rated — the baseline 'not achieved' state

AI shortcut: AI helps by quickly auditing existing artifacts, surfacing what is missing, and proposing a minimum process backbone.

CL1

Performed

The process achieves its defined purpose. Work products exist and base practices are carried out, even if not yet planned or controlled.

Expected outcomes

Base practices executed for each process
Required work products are produced
Traceability between inputs and outputs is visible

Process Attributes (PA)

• PA 1.1 — Process Performance

AI shortcut: LLM-assisted generation of requirements, test cases, and review checklists makes 'Performed' reachable in days rather than weeks.

CL2

Managed

Performance is planned, monitored, and adjusted. Work products are appropriately established, controlled, and maintained.

Expected outcomes

Objectives, responsibilities, and resources are defined
Performance is measured and corrected
Work products are version-controlled and reviewed

Process Attributes (PA)

• PA 2.1 — Performance Management
• PA 2.2 — Work Product Management

AI shortcut: Agentic AI tracks plan deviations, flags missing reviews, and auto-generates configuration management evidence from ALM/PLM systems.

CL3

Established

A managed process is deployed from a defined organizational standard, tailored per project, and capable of consistent outcomes.

Expected outcomes

Standard process exists with tailoring guidelines
Roles, competences, and infrastructure are deployed
Process performance data is collected centrally

Process Attributes (PA)

• PA 3.1 — Process Definition
• PA 3.2 — Process Deployment

AI shortcut: AI templates standardize tailoring, propose role assignments, and keep organizational process assets aligned with project reality.

CL4

Predictable

The established process performs within defined quantitative limits. Variation is understood and managed statistically.

Expected outcomes

Quantitative goals for quality and performance
Sub-processes are statistically controlled
Cause-effect of variation is analyzed

Process Attributes (PA)

• PA 4.1 — Quantitative Analysis
• PA 4.2 — Quantitative Control

AI shortcut: ML models on historical defect, effort, and review data predict variance, recommend control limits, and detect drift early.

CL5

Innovating

The predictable process is continually improved to meet current and projected business goals through innovation and change management.

Expected outcomes

Continuous improvement is institutionalized
Innovations are evaluated and deployed
Process changes are measured against business goals

Process Attributes (PA)

• PA 5.1 — Process Innovation
• PA 5.2 — Process Innovation Implementation

AI shortcut: Generative AI proposes process experiments, simulates impact, and rolls out improvements with automated change management.

PA Scoring: How Process Attributes Are Rated

Each Process Attribute is rated on a four-point ordinal scale derived from the percentage of generic practices and work products evidenced during assessment. To claim a capability level, every PA up to and including that level must be rated Largely or Fully achieved, and all lower-level PAs must be Fully achieved.

Rating	Achievement range
N — Not achieved	0% to ≤ 15%
P — Partially achieved	> 15% to ≤ 50%
L — Largely achieved	> 50% to ≤ 85%
F — Fully achieved	> 85% to 100%

How AI Helps Teams Reach Higher ASPICE Levels Faster

Requirements & traceability

LLMs score INCOSE quality, flag ambiguity, and auto-link requirements to design and tests — strong PA 1.1 / PA 2.2 evidence with minimal manual work.

Test generation & coverage

Prompt chains derive functional, boundary, and negative cases from specs, giving auditable test design rationale for SWE.4 / SWE.5 / SWE.6.

Process metrics & dashboards

Agents pull data from Jira, Polarion, and Git to compute defect density, review effectiveness, and rework rates — the quantitative basis of CL4.

Continuous improvement

Generative AI proposes process experiments and simulates impact, supporting PA 5.1 / PA 5.2 with measurable improvement cycles.

Benchmark your ASPICE maturity with AutomotiveAI

Our training and AI tooling tracks help engineering teams move from CL1 to CL3 on critical processes — and lay the data foundation for CL4. Explore the full training catalog or talk to us about an assessment.

Browse training tracks Talk to us