When AI Enters Stock Broking

https://ift.tt/465xqVO

Welcome to The AI Shift by Inc42, our all-new newsletter that delves deep into the world of artificial intelligence, LLMs, big tech giants and the major trends sweeping the Indian startup and tech ecosystem. Here’s the fourth edition; do send us your feedback and suggestions so we can improve as we go along!

For years, retail investors have craved one thing — a cheat code that could help them maximise returns by cutting through noise, timing trades better, and managing risk more efficiently. From Telegram tips to YouTube gurus, the search for this edge has been unending.

AI promises to change this equation even if that promise isn’t entirely new. There were robo advisors before built on rule-based models, based on static data, and pre-defined risks. They automated for use-cases like upselling, but didn’t really “think” on investors’ behalf.

With the advent of generative AI, the likes of Groww, Zerodha, INDMoney and other digital brokers have evolved. They are increasingly using GenAI models and tools to summarise market information, surface relevant signals, answer portfolio-level questions, and reduce the cognitive overload that comes with modern trading.

But this shift raises uncomfortable questions.

If AI can analyse data faster and spot patterns humans miss, how close can it get to actual trading decisions? Where does assistance end and automation begin? And in a market as tightly regulated as India, how do platforms ensure that AI-powered insights do not nudge users into risky behaviour?

And finally, as AI embeds itself deeper into the retail investing stack, how much control are humans willing to give up?

Embedding AI Into The Discount Broking Stack

Retail trading is cognitively demanding. Investors juggle multi-asset portfolios, shifting intraday margin, volatile option chains, earnings data, and an endless stream of news. AI co-pilots are emerging as the layer that can condense this complexity into insights that traders can absorb and act on faster.

At INDMoney, AI has been designed to sit squarely in the insight layer. It synthesises signals, summarises information, and answers portfolio-level questions in natural language, but never crosses into execution.

“We draw a hard architectural line between probabilistic insight and deterministic execution,” said Kausal Malladi, CTO at INDMoney. “AI reduces the time-to-insight, but trading is ultimately about risk appetite, which is a subjective, human variable.”

The thesis is gaining ground across brokerages — let AI handle cognitive load, not capital.
“If traders follow AI signals without thinking, it can cause big mistakes. AI should help you make decisions. It should not make decisions for you.” a Groww spokesperson added.

Also, instead of rolling out generic chatbots or surface-level assistants, Indian online brokers are embedding AI deep into their core systems. The goal is not conversation, but context.

Various platforms like Zerodha, Groww, INDMoney, and FYERS have different approaches to how they plug in AI systems. For instance, both Zerodha and Groww offer model context protocol (MCP) support, which lets you connect to Claude or ChatGPT for AI-assisted workflows.

At INDMoney, Malladi explains that AI is woven tightly into data-retrieval pipelines, internal engineering processes, and customer operations. However, its AI models do not ‘know’ market data. They are simply fed structured, real-time inputs from internal systems and reason only within that context.

A similar infrastructure-first approach is visible at investment tech startup FYERS, which also offers an AI trading assistant. However, the startup avoids positioning the AI assistant as an all-knowing oracle.

“The biggest risk is blind trust,” said Yashas Khoday, cofounder and CPO of FYERS.

Bad or incomplete data can quietly corrupt AI output. Plus, as Khoday added, markets are shaped not only by numbers but also by policy, sentiment, and events that models struggle to capture all the time. “Many AI tools don’t explain why they give a suggestion… If you don’t know what’s behind it, it’s very hard to trust it when real money is involved.”

This explains why brokers are choosing to keep intelligence, data, and workflows tightly governed within their own platforms, instead of letting users rely on generic external AI tools with no market or regulatory context.

Execution Control Becomes The Moat

As AI moves closer to money, the source of defensibility in stock broking is changing. Pricing and user experience still count, but they no longer form a sufficient moat on their own.

“Execution control and regulatory compliance create the deepest moats in India. Data and models will get commoditised. Foundational models are already becoming accessible. What’s harder to replicate is the trust infrastructure,” said Arjun Malhotra, general partner at Good Capital.

Malhotra pointed to KYC rails, compliance architecture, reliability of settlements, and operational discipline as the real entry barriers. As such, defensibility will sit with platforms that can reliably execute trades at scale inside a tightly regulated market.

This new reality is shaping how brokerages are designing AI within their tech stacks. At INDMoney, for instance, AI is intentionally stateless by design. Each session is grounded using real-time market inputs, and the model is not meant to retain memory or make assumptions beyond what is explicitly provided.

Crucially, execution is kept separate. “The AI has zero write-access to the order management system… A trade can only be executed via a manual, user-initiated action,” Malladi added.

These controls are not just product choices but regulatory necessities. SEBI has consistently emphasised explainability, auditability, and disclosure when AI-generated insights are involved in the Indian markets.

Early Fault Lines

Even with guardrails in place, early fault lines are starting to show. One of the biggest risks is contextual hallucination, especially in fast-moving segments like F&O, where a small mismatch between the prompt and the live market can translate into real losses.

To contain this, Malladi explains that INDMoney’s architecture treats AI purely as a reasoning layer. Rather than letting the model draw from its pre-trained memory, the platform constrains the AI to structured, real-time inputs.

Execution also remains strictly segregated – the model cannot place or route orders, and the user must still verify and manually click to execute. This keeps both control and accountability with the trader.

FYERS takes a similar conservative stance. “AI is just a tool,” Khoday said. “Use it like a chart or a screener. Let it support your thinking, not replace it.”

As a result, online brokers are favouring explicit user-triggered execution, clear disclaimers, and explainable outputs over predictive bravado.

“In the near term, faster research and context-building will see the earliest adoption. AI can help surface what’s relevant and connect insights quickly, without replacing traders’ judgment,” Groww’s spokesperson told Inc42.

What Can Startups Build?

This pivot to AI is creating a meaningful whitespace for Indian startups. “The real opportunity is not in prediction; it’s in decision support,” Khoday said.

He pointed to personalised insights, trader self-analysis, and tools that help users understand what works and what doesn’t. He believes that traders want clarity and learning, not magic signals.

Malhotra sees two divergent paths. Building a full AI-native broker offers greater upside but demands solving distribution, trust, and compliance upfront. Selling AI infrastructure to brokers is faster, but risks commoditisation unless the solution is deeply India-specific.

“Unless you’re solving something uniquely Indian, like vernacular intelligence for tier II and tier III investors or compliance-first AI built around SEBI rules, you’re competing with global tools,” he said.

All in all, AI is not turning retail investors into super-traders. Its real impact lies in reducing friction, sharpening judgment, and helping humans see markets more clearly. As AI moves closer to capital, the winners will be platforms that can keep execution, accountability, and trust firmly human, while letting machines handle the noise.

The Weekly Buzz: End Of Human-Written Code?

An AI CEO declares coding is dead. A founder nods gravely. Twitter argues. Engineers roll their eyes and get back to writing code.

But this time, the pattern may break, because the warning didn’t come from an AI evangelist or a slide deck. It came from Ryan Dahl — the creator of Node.js, one of the most widely used runtime environments in the world.

So when someone of Dahl’s pedigree claims the era of humans writing code is over, it suddenly feels less like hype and more like an uncomfortable reality — not because software engineers are obsolete, but because typing syntax is no longer the main job.

Dahl argues that while engineers still have plenty of work, directly writing code is becoming secondary to higher-level thinking about systems and intent.

This definitely prompted some reflection. Some saw this as the rise of designers and product thinkers. Others called it the latest compression in a long history of abstractions. A few went further, suggesting AI could even verify its own code, shifting human effort toward specifications instead of fixing bugs.

The takeaway wasn’t that software engineering is ending. It’s evolving. Coding, like writing machine language before it, is becoming a niche. What’s taking its place is software design, clarity of intent, and systems thinking, with machines increasingly handling the syntax.

Startup In The Spotlight: Dashverse

Media businesses have historically faced disruption from demand-side shifts like streaming and mobile, but the actual production of content remained resource-intensive. Dashverse believes the next wave of disruption is on the supply side, where Generative AI will fundamentally change how media is created.

Founded in 2023 by Sanidhya Narain, Soumyadeep Mukherjee, and Lalith Gudipati, Bengaluru-based Dashverse is building an AI-native entertainment ecosystem designed to democratise storytelling. Starting with digital comics and expanding into video, the startup positions itself as a ‘supply-side disruption’ engine, enabling creators to produce high-quality serialised content without traditional studio constraints.

Dashverse’s core thesis is that AI can solve the ‘content starvation’ problem by allowing creators to build faster and deeper. Rather than mitigating the shortcomings of current AI models, it focusses on an application layer that assumes models will improve, building depth in consistency and controllability that raw models lack.

The startup’s platform rests on two proprietary pillars — a specialised ‘image stack’ that ensures consistency and control over props and composition, and an ‘orchestration layer’ that intelligently selects the best video models (like Google’s Veo) for specific shots. This stack powers their consumer app, DashReels, allowing them to monetise via subscriptions rather than a typical per-token AI business model.

In theory, Dashverse aims to make micro-dramas the next massive social format — comparable to TikTok or Instagram — by compressing production complexity so that anyone can eventually create a short drama from home. With over 5 Mn monthly active users, $13 Mn in funding last year and $30 Mn in annualised revenue as of December, the company claims that it is rapidly validating this new media landscape.

Prompt Of The Week

What prompts and hacks are CTOs, CEOs and cofounders using these days to streamline their work?

Here’s Praveer Kochar, cofounder and CPO of KOGO AI, with a prompt he uses to quickly convert ambiguous business problems into deployable AI workflows, while keeping feasibility, risk, and timelines front and centre:

“You are an enterprise AI product lead. Given this problem statement (add problem statement), break it down into:

What Can Be Automated End-To-End

Where Human-In-The-Loop Is Mandatory,

Key Risks Or Blind Spots

The Simplest V1 Workflow That Can Be Deployed In Under 2 Weeks.

Respond In Clear, Decision-Ready Bullets”

Editor’s Note: Some prompts may need to be adjusted by users for best results or may not work as intended for certain users.

The post When AI Enters Stock Broking appeared first on Inc42 Media.

from Inc42 Media https://ift.tt/JPzjbgi
via

When AI Enters Stock Broking