Managing Engineers Who Use AI: New Manager Skills

Managing AI-augmented engineers requires new skills around code review, output evaluation, and productivity measurement.

What AI does well here

• Calibrate code review for AI-augmented code (different patterns, different errors)
• Discuss AI tool choices with team (Cursor vs Claude vs custom workflows)
• Evaluate productivity by output not by activity (AI changes the activity-output mapping)
• Address quality concerns when AI use becomes a crutch instead of a tool

What AI cannot do

• Substitute AI productivity for actual engineering judgment
• Skip the engineering fundamentals AI doesn't replace
• Make the team faster than the team's actual capability

AI can draft AI engineering manager scorecard memos that summarize delivery, quality, and team-health signals on one page for the skip-level.

What AI does well here

• Pull metrics from delivery dashboards into a consistent template
• Draft narrative explanations linked to each metric movement

What AI cannot do

• Capture the team-health signals only the manager senses in 1:1s
• Decide which people moves to flag to the skip-level

Skip-level meetings reveal team health; AI rehearses the answers that show calibration without spin.

What AI does well here

• Rehearse honest answers about team health
• Draft talking points that show calibration
• Suggest questions to ask upward

What AI cannot do

• Replace genuine candor with phrasing
• Predict the skip's specific concerns

Skip-level conversations: what your boss's boss is actually listening for

Skip-level meetings — where a director or VP meets one-on-one with each of their reports' direct reports — are designed to surface signal that the direct management layer may not be passing up. The skip is listening for calibration: does this manager see their team clearly, name their team's risks honestly, and take accountability for both the wins and the gaps? Managers who spin — who present their team as higher-performing than the data suggests, or who blame the team for misses that belong to the manager — quickly lose credibility with skips who have been doing this for years. The questions that tend to be most revealing are about team health indicators: who on the team is at risk of leaving and why, what the manager is worried about that nobody else knows, and how the team makes decisions when the manager is not in the room. AI can help an engineering manager prepare for a skip by simulating the likely question set and providing feedback on draft answers — specifically checking whether the answers show honest calibration or self-protective spin. The preparation is valuable not because it produces scripts, but because it surfaces the manager's own blind spots: the topics they avoid, the glossy language they use when the reality is messier, the wins they over-explain because they feel defensive about the underlying story.

• Skips evaluate manager calibration — do they see their team clearly and take accountability?
• AI can simulate skip-level question sets and critique draft answers for spin versus honest calibration
• Name a real risk before you name a real win — skips who hear only wins know they are being managed
• Preparation is about surfacing your own blind spots, not scripting answers you'll sound like you're reciting