Slow down to speed up

2026-03-13 · Source: The Engineering Manager · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

The article, "Slow down to speed up," argues that deliberate slowness in planning and design phases is crucial for effective AI-assisted development, despite AI's capacity for rapid execution. It frames this using Daniel Kahneman's System 1 (fast, automatic, AI-like) and System 2 (slow, deliberate, human judgment) thinking, asserting that AI's power makes System 2 decisions *more* important. The "illusion of speed" arises when AI rapidly implements flawed requirements, leading to significant technical debt and costly rework. The author emphasizes that while AI excels at fast pattern-matching, human judgment remains essential for defining the *right* problems. The piece suggests using AI itself to accelerate deliberation, offering methods like pre-mortems and throwaway prototypes to clarify requirements and surface risks early, ultimately ensuring that speed is deployed deliberately and effectively.

Key takeaway

For AI/ML Directors and Software Engineers facing pressure to accelerate development with AI, you should actively integrate deliberate "slow phases" into your workflow. Prioritize clarifying requirements, running pre-mortems, and building throwaway prototypes using AI to validate direction. This approach, making invisible planning visible through shared artifacts and timeboxing, prevents costly rework and ensures your teams build the *right* solutions efficiently, ultimately achieving faster long-term throughput.

Key insights

AI amplifies the cost of flawed initial decisions, making deliberate "slowness" in planning more critical for true speed.

Principles

• AI shifts leverage to pre-execution decisions.
• Catch mistakes early to minimize rework costs.
• Deploy AI's speed deliberately, not blindly.

Method

The "thinking first" protocol involves clarifying requirements, success criteria, and constraints before offloading work to AI, ensuring critical skills are retained and mistakes caught early.

In practice

• Use AI for pre-mortems on design approaches.
• Build throwaway prototypes to validate understanding.
• Ask AI to generate edge cases early.

Topics

• AI Development Workflow
• Kahneman's System 1/2
• Technical Debt Prevention
• Software Design Principles
• AI-Assisted Prototyping
• Requirements Engineering

Best for: Director of AI/ML, AI Architect, Software Engineer

Related on AIssential

• AI Coding Assistants Haven’t Sped up Delivery Because Coding Was Never the Bottleneck — AI coding tools shift bottlenecks from coding to specification and verification, demanding new team structures and workflows.
• Thoughts on slowing the fuck down — Agentic engineering's speed accelerates error accumulation and cognitive debt, making codebases unmanageable without human oversight.
• Fragments: May 5 — AI's rapid evolution demands adaptable development practices, responsible deployment, and strategic infrastructure choices.
• Coding Agents Meet Data Science — AI coding agents boost developer productivity but introduce new bottlenecks and require specialized adaptation for data science workflows.
• Ideas: slow down to speed up when working with AI agents — AI coding tools are rapidly adopted, prompting a "slow down to speed up" approach.
• You Shipped It Fast. But Did You Ship It Right? — AI-generated code creates an "illusion of correctness" by hiding critical unstated assumptions that lead to production incidents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Engineering Manager.