From Lead PM to AI Product Leader: A 15-Year Roadmap

Fifteen years is enough time to be honest about which bets paid off, which were premature, and which I would do again without thinking. The arc from junior software engineer in Mumbai to Lead Product Manager running multi-country retail platforms in Dubai was not linear. It was, in retrospect, more reproducible than it felt while I was inside it.

This is the version of the journey I'd hand to a peer trying to walk the same road in 2026 — what compounded, what I'd skip, and the bets I'm making now as the field shifts toward AI product leadership.

The four eras.

The career broke into four distinct phases. Each had a different unit of work, a different feedback loop, and a different skill that needed to harden before the next phase was unlocked.

2010–2014 — Software engineer. The unit of work is a build.

Mobile engineer at NIIT Technologies and HCL Technologies, working on the Emirates iPad app and a medical-journal app for The Lancet. Early career was about building taste for what "done" actually means. The feedback loop was tight: the build either crashes or it doesn't, the search either returns in 200ms or it doesn't.

What compounded: pattern matching for what tends to break, comfort with the debugger, a habit of reading the framework source instead of guessing.

What I'd skip if I were doing it again: the time I spent reinventing solved problems out of pride. The ego cost of using someone else's library is small. The time cost of writing one badly is enormous.

2014–2018 — Senior engineer. The unit of work is a release.

Senior Technical Engineer at Emirates Airlines, leading a redesign of the Emirates app check-in flow. Migrated to Swift alongside Obj-C, cut crash rates by 20%, automated tests to take 25% off QA cycle time.

This was the era where I learned that shipping is a discipline, not a milestone. The release calendar is a living thing, every release degrades or compounds the next one, and the engineers who care about the second derivative are the ones whose teams quietly outpace everyone else.

What compounded: instinct for release-readiness, comfort with cross- functional reviews, the ability to estimate without flinching.

What I'd skip: the months I spent over-investing in personal productivity tooling at the expense of team-level practices. Solo productivity gains compound only until you hit a team-shaped ceiling.

2018–2021 — Technical PM. The unit of work is a rollout.

Technical Product Manager — Platform & Operations at Carrefour (Majid Al Futtaim Retail). Rolled new digital tools to 400+ stores, held platform reliability at 99.5% by tightening vendor SLAs, improved release cadence by 30%, and brought Dynatrace and New Relic into the observability stack.

This is the bridge phase. The transition from "I write the code" to "I am responsible for whether this thing works at scale across countries" requires letting go of identity-as-builder and growing identity-as-orchestrator. The hardest skill is to stop measuring my contribution by my own output.

What compounded: vendor governance, SLA muscle, post-incident discipline, the patience to debug an outage and the political ear to debug an organisation.

What I'd skip: the over-correction into pure management work. The PMs who keep their hands close to the technical surface have a long-term advantage that becomes obvious in the AI era.

2022–Present — Lead PM. The unit of work is a portfolio.

Lead Product Manager — Digital Transformation & Innovation at Carrefour. 19-person cross-functional team. 15 countries operated. Scan & Go scaled to 9 countries in 12 months. FacePay and Carrefour City+ launched as region-firsts. AI-driven planning system cut roadmap cycle time by 70%.

This phase is about operating a portfolio, not a product. Multiple bets, different time horizons, different audiences, different country contexts. The job is no longer "what to build next"; it is "what to start, what to accelerate, what to harvest, and what to sunset" — every week, in parallel.

The skills that compounded across all four phases.

1. Writing. Clear writing is the most underrated PM and engineering skill of the last fifteen years. It compounds monthly. I have never regretted spending more time on a written decision and frequently regretted spending less.
2. Reading dashboards skeptically. Every dashboard lies a little. The PMs and engineers I trust most can look at a metric and ask three questions before they react: how is this measured, what's the denominator, and what does it not capture.
3. Operational discipline. SLAs, runbooks, on-call, post-mortems. Not glamorous. Compounds enormously. Every senior role I have held has rewarded this disproportionately to what the job description implied.
4. Cross-cultural and cross-country fluency. Working across MENA, Asia and Africa for a decade taught me to read what an objection actually is, separate from how it's phrased. This is the hardest skill to study and the most valuable to have.

The bets I'm making for the next era.

The transition from Lead PM to AI Product Leader is not a title change. It is a re-organisation of how I spend the week. Concretely, the bets I'm making in 2026:

• Build, don't admire. I keep my hands on the tools — Claude Code, MCP servers, Cursor, the Anthropic and OpenAI APIs. Not because the team needs me to. Because the leaders who will shape AI product strategy in 2027 are the ones whose intuitions are calibrated by current build experience, not last year's.
• Trust over throughput. The new tools made throughput cheap. Trust did not get cheaper. Most of my time investment is shifting toward the relational and judgement work that does not have a model in the loop.
• Bet on agentic operating models. Internal systems that produce documents and decisions as outputs, not documents and decisions written by hand. Every quarter I am replacing one workflow that used to need a person with one that needs a person to review. The compounding effect of that across a year is big.
• Pick the next role for what it forces me to learn. Director of Product, Head of Digital, Head of Platform, AI Product Leader — the framing of the role matters less than which one will compress my learning curve fastest in the next two years. That filter changes how the conversation with a hiring team feels.

What I'd tell a 25-year-old me.

Three things, in order of how much they would have helped:

Choose the operating environment, not the title. The seat that makes you sit next to senior operators who are slightly above your league will compound faster than the seat with the bigger title and worse company. I made the right call on this twice and the wrong call on it once. The wrong call cost a year.

Do the boring layer well. The unsexy parts — observability, on-call, vendor SLAs, post-mortems, configuration layers — are where senior people spend their time. Do those parts well early and you skip a phase later.

Write more than you think you should. Specifically, write the things you wish someone had written for you. The act of writing them shapes your own thinking faster than reading does.

The bottom line.

The path from engineer to AI Product Leader is not a ladder. It is a series of phase transitions where the unit of work changes and the skill that has to harden is different each time. Every era rewards going deep on one skill while keeping a hand on the previous era's. The leaders who keep that combination — current builder instincts plus operator discipline plus portfolio thinking — are the ones I want to work with, and the ones I'm trying to become.

Building toward the same arc? Happy to compare notes.