AI Is Not Just a Brain: Building an Operating Architecture for Reliable Research | Mumbai
Session Overview
AI is easy to start using. Using it reliably for serious research is much harder. It is probably closer to learning Excel or financial modelling than learning to use a smartphone. You can get started quickly, but getting consistently good results takes an understanding of the underlying concepts, a disciplined way of working, and repeated practice.
This session is designed for equity research analysts and portfolio managers who want to use AI more effectively in their research workflow. It is practitioner-led and grounded in actual use.
Models, tools and features will keep changing. So rather than spend most of the session on features that may look different a few months from now, the focus will be on the more durable ideas behind using AI well—how to give it the right instructions and context, how to coordinate its work, how to reduce errors, how to verify what it produces, and where human judgment still matters
Recommended pre-read
Participants are strongly encouraged to review the earlier Delhi CFA presentation before the session:
AI Is Not Just a Brain — The Analyst’s Operating Architecture
https://contrarianvalueedge.substack.com/p/ai-is-not-just-a-brain-the-analysts
The Delhi presentation is fairly detailed and provides the broader foundation for many of the ideas we will discuss. In the October session, I plan to pick up a smaller number of important concepts and spend more time on practical examples, application and Q&A, rather than repeat the full presentation.
Reading it beforehand should therefore make the live session more useful: participants can come with better questions, engage more deeply with the examples, and we can spend more of the available time discussing how these ideas actually work in research.
Topics to be covered
- The AI Brain — what LLMs do well, where they struggle, and why a fluent answer is not necessarily a reliable one.
- Prompt Engineering — giving AI clearer objectives, assumptions and instructions.
- Context Engineering: Eyes and Memory — controlling what the AI can see, know and use.
- Clarify → Confirm → Execute: The Nervous System — making sure the AI understands the task before it starts doing the work.
- Reliability Layer: The Immune System — guardrails, verification and ways to reduce avoidable errors.
Human Judgment — what AI can assist with, and what should ultimately remain with the analyst.
EVENT DETAILS
Date- 10th October 2026
Time- 10:00 am to 4:00 pm
Venue- Jio Convention Centre, BKC
Anil Tulsiram
Anil Tulsiram is based in Bangalore. He is a Chartered Accountant (2001 batch) and a CFA Charterholder with experience in Auditing, Financial Due Diligence, and Equity Research. He has been a full-time investor in public equity markets since early 2012.
Anil’s journey with AI began in early 2023. As an individual investor with no team support, he saw AI as a way to work faster and better. Early results, however, were not particularly encouraging. The turning point came after the launch of Deep Research in February 2025, which significantly accelerated his use of AI. In July 2025, he presented “How to Use ChatGPT Deep Research to Produce Analyst-Grade Reports” at the FLAME Alumni Meet, followed by “Trustworthy AI for Equity Research — How LLMs Think and How to Reduce Errors” at PPFAS OctoberQuest in October 2025.
A pivotal moment in shaping his approach came from a talk by the CEO of a leading NBFC, who shared that he needed to “understand AI to the nth level” before deploying it in his business — otherwise, he would remain unaware of where mistakes could creep in. That insight reinforced a key conviction: focus on the first principles of how AI and its tools work. Tools will keep changing, but first principles will not. It is for this reason that my presentation will focus on fundamentals and first principles.
His writings on AI are available at contrarianvalueedge.substack.com.