The Advice Layer Is the Next Battleground – And AI Just Made It Cheap

For two decades, Indian fintech competed ferociously on the transaction layer. Payments became free, onboarding became instant, mutual fund purchase became a three-tap affair. But the advice layer stayed stubbornly human, expensive, and gatekept by relationship managers who had more clients than time.

That moat is now draining – and not because of a better app.

The Signal

Bengaluru-based PowerUp Money has raised $10 Mn led by Flourish Ventures, with Peak XV, Accel, Blume and others participating. It is moving beyond mutual fund advisory to become a full-stack financial planning platform – launching Elite Prime (an AI wealth coach) and scaling Infinite, its managed advisory arm, which reportedly crossed ~₹500 Cr in AUM within nine months. It tracks over ₹2.5 Lakh Cr across 15 lakh portfolios.

Strip away the funding headline and what remains is an architectural claim: that a digital agent can substitute for a wealth manager. That claim deserves scrutiny.

From Read-Only Insight to Write-Access Agency

Here is the shift I find most interesting as an architect. Legacy wealthtech was essentially read-only: dashboards, analytics, portfolio X-ray, SIP calculators. Valuable, but low-risk. An AI “wealth coach” that recommends allocation changes and supports execution is write-access agency over someone’s life savings.

That is an entirely different threat model. The failure mode is not a slow page load – it is a household losing its retirement corpus because a model extrapolated from a bull run.

In my experience building regulated systems, the discipline that separates advisory-grade AI from chatbot-grade AI is unglamorous: deterministic policy engines wrapped around the probabilistic core, retrieval grounded in the user’s actual portfolio data, immutable audit trails for every recommendation, and explainability a compliance officer can sign off on. Hallucination is annoying in a content tool. In a fiduciary context, it is legal liability.

Which brings me to the hardest part: who is accountable when the AI is wrong? SEBI’s boundaries between advice and execution are not decorative. Every “AI-first” wealth platform will eventually be judged not on its model quality, but on its governance architecture – escalation paths, human-in-the-loop checkpoints, documented decision lineage.

The Real Asset Is behavioural, not technical

The ₹2.5 Lakh Cr tracking base is being described as a distribution win. I see it as a dataset win. Genuine labelled data on how Indian investors actually behave across market cycles – panic-selling, SIP pausing, risk-profile drift – is extraordinarily rare. That corpus, not the LLM, is the durable moat.

But scale also exposes operations. ₹500 Cr AUM in nine months means rebalancing engines, reconciliation pipelines, order routing and exception handling must survive 100x. In wealth management, the model is the easy build. The plumbing is where companies quietly die.

The Bharat Question

I’ve argued this repeatedly in STPI and state technology forums: the Account Aggregator framework, UPI and ONDC have given us the rarest thing in the world – open, consent-based financial data rails. Combine that with AI, and high-quality advice stops being a metro luxury.

In the Northeast, advisor density is abysmally low. A student in Silchar or a first-generation saver in Shillong has almost no access to a competent planner. An AI-native advisory layer could genuinely equalise that – if it handles Assamese and Bengali, if it works on patchy connectivity, and if it is honest about its uncertainty rather than confidently wrong. That last condition is the whole ballgame.

Takeaways

• Design for accountability first, intelligence second. Model capability is commoditising; governance is not.
• Treat every recommendation as a legal artefact. Audit trails, versioning, and replayability are architecture, not features.
• Hybrid beats autonomous in fiduciary domains – dedicated human advisers remain the escalation substrate, not a legacy cost.
• Your behavioural dataset is your moat. Collect it with consent, govern it with sovereignty.
• Operations scale is the silent killer. Test the rebalancing and reconciliation layers before the model.

Closing Thought

Technology has always been excellent at making access cheap. It has never once made judgement cheap. The winners in this next decade will be the ones who understand that distinction – and build accordingly.

---

About the Author: Sanjeev Sarma is the Founder Director and Chief Software Architect at Webx Technologies. With a core focus on Generative AI integration, Cloud-Native Scalability, and Enterprise Software Architecture, he has spent over two decades driving digital transformation across Northeast India and beyond. Beyond his corporate leadership, Sanjeev is deeply invested in shaping the future of the IT industry. He serves as an Industry Expert on the Board of Studies for Assam Don Bosco University’s School of Technology, advises state technology committees, and actively mentors emerging tech startups at STPI. He brings a unique, dual perspective of high-level enterprise execution and future-ready academic curriculum development.