Waterfall-to-Agile Was the Easy Part. Agile-to-AI Is the Real Test.

5 MINS

I've spent half my career helping financial-services teams move from waterfall to agile. We celebrated when the standups got shorter, when the backlog got groomed, when the burndown started looking like an actual burndown. Mission accomplished, we said. Then AI showed up and made all of those rituals look quaint.

What agile got right

Before I criticise it, credit where it's due. Agile gave heavily-regulated organisations three things that mattered:

Smaller batch sizes — the cost of being wrong dropped

Cross-functional pods — risk and engineering finally talked

A regular cadence of demos — stakeholders stopped guessing what was happening In a SAFe environment, even the planning rituals — for all their corporate weight — created a shared rhythm. That's not nothing.

What agile is now hiding

But the same rituals that liberated us in 2018 are now the bottleneck. Two-week sprints made sense when human throughput was the constraint. In an AI-assisted team, two weeks is a long time.

The PR I shipped this morning had three AI-generated drafts before the human review. The user research synthesis I did last week happened in 90 minutes instead of 9 days. The ticket template we used to spend 30 minutes on now writes itself in two.

The constraint has moved. It's no longer how fast we can do the work — it's how fast we can decide what work is worth doing.

The PM job is shifting harder than the eng job

Most of the AI-and-PM conversations I see online are about engineers. They're about Copilot, code generation, agentic dev tools. Fine. But the bigger shift is on the product side.

Discovery is collapsing. A solid AI research loop produces in a day what a research sprint used to take a fortnight to deliver

Specs are drafts, not deliverables. Anyone can generate a spec; the value is in the disagreement and the trade-off conversation that follows

Backlog grooming is becoming taste-curation. AI will fill the backlog. Knowing which 20% to keep is the skill The product manager who treats AI as "another tool in my stack" is going to be replaced by one who treats it as **a co-thinker who needs sharper questions**.

What I tell people moving from BA to PM right now

Half of my career started in business analysis. The route I see opening up for BAs and POs in financial services has nothing to do with adding "AI Product Manager" to your title. It has everything to do with getting closer to the customer pain than the AI can.

Here's why: AI is fantastic at synthesising what's already known. It's mediocre at finding what nobody has bothered to ask. The PM with first-hand customer empathy — the one who's actually listened to a credit-card-collections call, sat with a fraud agent, watched a small business owner abandon an onboarding flow — that person becomes more valuable, not less.

The differentiator is not your tool stack. It's your proximity to the truth of how the product gets used. That's the part the model can't generate.

The next ten years

I'm bullish on financial services as a place to build AI products. Not because banks are first-movers — they aren't — but because the stakes are real, the data is rich, and the trust gap is huge. Whoever closes that trust gap responsibly will have built something that matters.

That's the bet I'm making.

In Banking, the Customer Journey Is the Product

Shipping AI Inside a Bank Is a Different Sport

Komal skipped presentations and built real AI products.

Komal Sikka was part of the March 2026 cohort at Curious PM, alongside 17 other talented participants.