Intro

I've held leadership roles that have scaled organizations to over 100 people, and within that, I have built engineering teams to just under 50. It was enlightening, humbling, and a supremely educational chapter of my career. I carry the lessons of reorgs, hiring mistakes, processes, and cultural blind spots.

This article isn't about the entire company, but rather the engineering function specifically. And it's not a manifesto about inventing "AI-first scaling." Many people are already talking about AI augmentation and AI-native organizations. This is my take: if I had the chance to scale engineering again, in today's world where AI is table stakes, here's how I would apply those lessons differently.

What I Learned Without AI

• Don't split too small, too early: sub-4-person teams implode under load.
• Always anchor with an EM (or substitute) and a process role: even a 10-person startup needs light structure.
• Reorgs are a pain: someone is always unhappy, and resistance to change is real.
• Metrics matter early: cycle time and delivery health scale linearly with team size if you start capturing them before you "need to."
• Hiring is imperfect: you will get some wrong; so hire fast and part ways fast.
• Culture compounds: Engineering-led product development and outcome-driven coaching pay dividends in the long run.

The do-over is about pairing them with AI to scale more smoothly, faster, and with fewer setbacks.

Stage 0–10: One Team, One Goal

At this size, there is no such thing as "multiple teams." You are a single, tight-knit group chasing one outcome: finding product-market fit. Splitting into sub-teams creates artificial drag and dilutes energy.

Composition (Learnings):

• 1–2 Founding engineers (Staff or Principal-level, setting architecture and standards)
• 2–3 Senior Engineers (execution + mentoring)
• 2–3 Mid-level Engineers (feature velocity)
• No managers yet as the CTO or Head of Engineering leads directly

The Do-Over With AI:

• AI as glue: Every meeting, decision, and PR is auto-summarized into a central knowledge base. Context stays alive.
• AI as first PM: Cycle time, throughput, and delivery health are tracked automatically from GitHub/Linear because you don't wait until you're "bigger" to measure.
• AI pair programmers everywhere: Every engineer runs with an AI assistant to reduce burnout and enforce discipline (tests, docs, telemetry).
• Customer loops, codified: Use AI to digest all customer calls and feedback into structured insights. Early on, this creates discipline around shipping to user needs.

Anti-Pattern: Splitting too small, too early. Sub-4-person teams lose alignment and end up building silos before the company even finds product-market fit.

Stage 10–25: Hiring, Role Clarity, and Efficiency

Once you cross into double digits, the problem shifts. Chaos isn't a result of a lack of effort; it stems from unclear roles, fuzzy goals, and mismatched hires, often causing it. Hiring quickly and parting ways just as fast becomes the discipline.

Composition:

• 1 Principal Engineer (chief architect)
• 4–6 Staff Engineers (mentorship + architecture)
• 6–10 Senior Engineers (execution + ownership)
• 6–8 Mid-level or Junior Engineers (feature development)
• 2–3 Engineering Managers (each with ~6–8 reports, focused on people + delivery)

The Do-Over With AI:

• Hiring process & rubric at center stage: Use structured role clarity questionnaires + AI to draft rubrics and evaluate fit consistently. Even with AI augmentation, you will make hiring mistakes—speed in correcting them is critical.
• Clear goals and roles: Every hire should be mapped to a measurable outcome. AI can help translate company goals into team-level objectives, so no role feels ambiguous.
• Efficiency metrics reviewed weekly: Use AI to track cycle time, bug rates, anddeployment frequency. AI can highlight where velocity is slowing or where teams are uneven.
• AI-assisted management: EMs utilize AI to write performance summaries, prepare one-on-one notes, and track delivery health. Managers stay focused on outcomes, not admin overhead.

Leadership Principles:

• Err on the side of fewer managers, but never compromise on quality. The best managers are force multipliers who can extract the best from large groups, connect with engineers at any career stage, and operate as 360-degree leaders.
• Lean on your leads. Some engineers will become 100 times more productive solo contributors with AI, but the best will be 1,000 times more productive because they multiply the work of others. Coach your Staff and Senior engineers to be technical coaches themselves. Their ability to elevate peers is the real unlock at this stage.

Anti-Pattern: Adding more people without adjusting roles or process. At this stage, unclear responsibilities kill momentum faster than a lack of headcount.

Stage 25–50: Respect Velocity, Lead With Persuasion

By this stage, you're operating as a "team of teams." The trap is overestimating your true velocity and committing to more than the organization can realistically deliver. That's where chaos creeps in.

Composition:

• 2 Principal Engineers (cross-team architecture + platform direction)
• 5–8 Staff Engineers (anchor pods, mentor seniors, steward AI adoption)
• 10–15 Senior Engineers (drive delivery and technical ownership)
• 10–15 Mid-level Engineers
• 5–8 Junior Engineers
• 4–6 Engineering Managers
• 1–2 Directors of Engineering

The Do-Over With AI:

• Be honest with velocity: Utilize AI-powered analytics to establish a baseline for cycle time, throughput, and bug rates. Don't inflate promises and set realistic goals first.
• Stretch through leadership, not fiction: Once the baseline is clear, use persuasion, motivation, and a compelling vision to inspire growth and stretch goals. Let managers and tech leads rally engineers around ambitious outcomes without manufacturing false deadlines.
• Light process, still: Keep Kanban/sprint ceremonies simple and AI-supported (automated grooming, summaries, insights). The lighter the process, the more leaders can flex their persuasive strengths.
• Engineering-led product development: By now, your culture should be outcome-driven. AI copilots can help teams frame PRs, designs, and retros in terms of user impact, not just code.

Anti-Pattern: Overpromising by assuming AI velocity multipliers will magically make teams superhuman. When goals don't match reality, morale collapses—and no AI bot can repair trust.

Stage 50–100: Leverage Metrics, Guard the Culture

At scale, you finally have levers. Metrics are abundant, dashboards are live, and data can inform decisions. But scale also introduces cultural risk: toxic engineers, know-it-alls, and "poisoners" in both management and IC ranks.

Composition :

• 3–4 Principal Engineers (platform + strategy)
• 10–12 Staff Engineers (cross-product technical ownership)
• 25–35 Senior Engineers (mentorship + delivery drivers)
• 25–30 Mid-level Engineers
• 10–15 Junior Engineers (with strong mentorship structures)
• 8–12 Engineering Managers
• 2–3 Directors of Engineering
• 1 VP of Engineering
• Dedicated AI/ML Platform Team (5–8 engineers)
• Developer Productivity Team (5–8 engineers)

The Do-Over With AI:

• Metrics as levers: AI turns raw telemetry into actionable decisions. Use it to simulate "what if" scenarios before making changes to your organization's structure or headcount.
• Guard against cultural poisoners: AI can flag delivery anomalies, but it won't spot the engineer who undermines peers or the manager who corrodes trust. Leaders must stay vigilant. Toxicity at this scale spreads fast.
• Onboarding and context at speed: With AI copilots, every new hire gains access to the full organizational memory on day one. Ramp-up becomes weeks instead of months.
• Leadership fluency with AI: Managers and tech leads must model AI adoption to drive effective implementation. At this scale, inconsistencies in AI usage lead to fragmentation.

Anti-Pattern: Believing metrics alone guarantee health. Velocity and bug rates can appear impressive on paper, while morale collapses due to cultural decay.

Closing: The Real Do-Over

The first time I helped scale the engineering team to nearly 50 people within a company that grew beyond 100. If I had a do-over, I would scale with AI, not because it is fashionable but because it is necessary. This is not about me inventing a new philosophy of scaling. It is about acknowledging that the baseline has shifted, and AI is no longer an experiment. It is infrastructure, and many engineering leaders are already embracing it. My contribution is personal because I learned from doing it the hard way, and I now know what truly works.

And here is what really works:

• Err on the side of fewer managers, but seek the highest quality ones. The best managers are force multipliers who elevate entire groups. They connect with engineers wherever they are in their career journey and pull the best out of them.
• Lean on your leads.Some engineers will thrive as AI-augmented 100x solo contributors, but the best will be 1000x because they help countless others become 50x.Coaching them to be coaches is the most powerful lever you have.