Performance Management

AI-assisted development fundamentally changes what "good performance" looks like for software engineers. The developer who writes 200 lines of clean, well-tested code per day using AI is not necessarily outperforming the one who writes 50 lines of critical algorithm code by hand. Traditional metrics like code output, velocity points, and bug counts need reinterpretation. This section provides a revised performance framework that recognizes the new skills and behaviors required for AI-augmented engineering, aligned with Pillar 5: Organizational Alignment.

Revised Competency Framework

The following competency areas should replace or augment your existing engineering competency model for roles that involve AI-assisted development.

Core Technical Competencies (Unchanged)

These fundamentals remain essential regardless of AI tool usage:

• Software design and architecture
• Data structures and algorithms
• Testing and quality assurance
• System debugging and troubleshooting
• Domain knowledge and business understanding

warning

AI tools make it possible for developers to produce code beyond their understanding. Ensure that performance evaluations assess comprehension, not just output. A developer who generates a microservice architecture with AI but cannot explain the trade-offs has a performance gap, not an achievement.

AI-Augmented Competencies (New)

Add these competencies to your evaluation framework:

Competency	Definition	Beginner Indicators	Proficient Indicators	Expert Indicators
Prompt Effectiveness	Ability to communicate requirements to AI tools clearly and precisely	Simple prompts, high iteration count	Structured prompts, 1-2 iterations	Template-quality prompts, first-attempt success
AI Output Evaluation	Ability to assess AI-generated code for correctness, security, and fit	Catches obvious errors	Catches subtle logic and security issues	Identifies systemic AI patterns and weaknesses
Workflow Integration	Effectiveness of AI integration into daily development workflow	Uses AI sporadically	Systematic AI usage with clear workflow	Designs AI workflows for the team
Security Vigilance	Awareness and mitigation of AI-specific security risks	Follows the checklist	Proactively identifies risks beyond the checklist	Defines security standards for AI code
Knowledge Sharing	Contributing to collective AI-assisted development knowledge	Asks good questions	Contributes to prompt library, shares tips	Mentors others, creates training content
Tool Adaptability	Ability to learn and evaluate new AI tools effectively	Proficient with one tool	Proficient with 2-3 tools, provides feedback	Evaluates tools, shapes tool strategy

Map these competencies to the four skill levels defined in Skill Development.

Evaluation Criteria

What to Evaluate

Dimension	What Good Looks Like	Red Flags
Code Quality	AI-assisted code meets team quality standards; low defect rate; clean reviews	Consistently high defect rate in AI-assisted code; rubber-stamps AI output
Productivity	Delivering more value (not just more code); cycle time improving	Volume increase with proportional quality decrease; gaming metrics
Judgment	Knows when to use AI and when to code manually; appropriate tool selection	Uses AI for everything indiscriminately; refuses to use AI for anything
Collaboration	Shares AI knowledge, contributes to prompt libraries, helps others improve	Hoards effective prompts; dismisses colleagues' AI approaches
Growth	Progressing on the competency matrix; actively learning	Stagnant skill level; not engaging with training or practice
Responsibility	Reviews AI code thoroughly; flags concerns; follows security practices	Commits unreviewed code; dismisses review feedback; bypasses security checks

What NOT to Evaluate

Do not use the following as performance indicators:

• AI usage frequency. Some tasks are better done manually. Penalizing low AI usage discourages good judgment.
• Lines of code generated. This incentivizes volume over value.
• Number of AI suggestions accepted. A high acceptance rate may indicate insufficient review.
• Speed of AI-assisted task completion. Fast is only valuable when it is also correct and secure.
• Comparison to AI non-users. Different developers have different optimal workflows. Comparing AI users to non-users is not meaningful.

Career Conversations

Reframing the Narrative

AI adoption triggers career anxiety for many developers. Address this directly in career conversations:

For junior developers: "AI tools give you the ability to be productive faster, but the fundamentals still matter. Your growth path is about building understanding -- not just building output. We will invest in your skill development alongside AI proficiency."

For mid-level developers: "Your value is shifting from 'writing code' to 'engineering solutions.' AI handles the routine work; your judgment, design skills, and domain knowledge are what make the output valuable. The Skill Development path shows how this translates to career growth."

For senior developers: "Your role in mentoring, architecture, and quality oversight is more important than ever. AI amplifies both good and bad engineering practices. You are the force multiplier that ensures AI makes the team better, not just faster."

Growth Conversations by Level

Developer Level	Career Focus	AI-Related Growth Goals
Junior (0-2 yrs)	Build fundamentals, learn to evaluate AI output	Reach Level 2 on competency matrix within 6 months
Mid-Level (2-5 yrs)	Deepen domain expertise, start mentoring	Reach Level 3, contribute to prompt library, lead AI code review sessions
Senior (5-8 yrs)	Architecture, technical leadership, team impact	Reach Level 3-4, evaluate tools, define team AI practices
Staff+ (8+ yrs)	Organizational impact, strategy influence	Level 4, influence organizational AI strategy per CTO Guide

Recognition

Behaviors to Recognize and Reward

• Quality catches. Developer identifies and prevents an AI-generated vulnerability from reaching production.
• Knowledge multiplication. Developer creates a prompt template that improves the entire team's output quality.
• Honest limitation discovery. Developer identifies a scenario where AI tools produce consistently poor results and documents the finding.
• Effective mentoring. Developer helps a colleague advance one level on the competency matrix.
• Process improvement. Developer suggests a workflow change that measurably improves team metrics.

Recognition Formats

Format	Frequency	Audience
Verbal acknowledgment in standup	As it happens	Team
Written recognition in team channel	Weekly	Team + org
Performance review documentation	Quarterly	Developer + manager
Nomination for organizational awards	As warranted	Organization
Conference talk or publication sponsorship	As warranted	Industry

Performance Review Adjustments

Review Cycle Considerations

During the first year of AI adoption, consider these adjustments to your standard review cycle:

1. Increase check-in frequency. Move from quarterly to monthly performance conversations during the first 6 months of AI adoption. Reduce back to quarterly once patterns stabilize.
2. Weight learning over output. During the first 3 months, weight growth and learning metrics higher than productivity metrics.
3. Adjust expectations. There is a natural productivity dip during the first 2-4 weeks of AI tool adoption as developers learn new workflows. Do not penalize this transition period.
4. Include AI competencies. Add the AI-augmented competencies to your formal review template starting with the first review cycle after AI adoption begins.

info

Coordinate performance management changes with your Scrum Master to ensure that sprint expectations are aligned with individual development goals. The Sprint Adaptation section addresses how to account for learning time in sprint capacity.

For the organizational perspective on performance management in AI-augmented teams, see Organizational Design in the CTO Guide. For executive-level reporting on team development, see Board-Ready Metrics.