Gazal Garg

📊 The AI Money Movement Layer (Part 1): MCP, ACP & TAP launched but are autonomous payments really the future? 📊 The last month’s big talking point has been agentic payments. Four launches stand out, each solving the same core problem: payments from an LLM/chat interface 🚀 1️⃣ Razorpay × OpenAI x NPCI (UPI Reserve Pay) 👉 Flow: LLM → merchant MCP → merchant creates order → Razorpay uses merchant level reserve Pay to debit user ✅ Takeaway: Works for repeat merchants with pre-set payment blocks. Convenient but limited by merchant specific setup 🚀 2️⃣ Google AP2 👉 Flow: LLM → AP2 trust layer maps instruction to stored mandate → AP2 signals merchant/PA to execute ✅ Takeaway: A centralized mandate vault acts as a trust layer. Ecosystem + regulators need recognize it. Like UPI Reserve Pay, needs to be pre-set up, and may introduce additional end user friction 🚀 3️⃣ OpenAI × Stripe ACP 👉 Flow: LLM → Merchant ACP → merchant creates order → Stripe issues single use delegated token → processes using stored creds or triggers 2FA ✅ Takeaway: Practical and safer with token limits. No pre-set up required. Works autonomously for non 2FA flows; still needs SDK/redirection for 2FA 🚀 4️⃣ Visa TAP 👉 Flow: LLM → Trusted Agent (VISA registered) → references tokenized card via PA → TAP validates agent + credential → issuer authorizes ✅ Takeaway: Similar to ACP, which doesn't require pre-set up. Authorizes a third party to access stored credentials and process payment using creds + agent authority. My mental model: it's important to decouple agentic commerce into shopping and payment flows. The shopping experience on LLM chats could be exponentially better than siloed sites. But the payment experience is already fairly seamless, and enabled through deep links, SDKs, or redirection. Direct 'in-chat' payments may be a nice to have, not a necessity Also, payments may never be fully autonomous. Friction can be a feature. Authentication builds trust; too much seamlessness can reduce perceived security. With strict 2FA in India/SEA, autonomy will depend on pre set mandates or payment blocks - not practical for casual purchases. What seems most likely are hybrid models: 👉 LLM commerce for discovery + order creation 👉 Semi-autonomous payments (1 click for pre authorized merchants, redirections/embedded SDKs) 👉 AP2 mandate layers only scale if banks, issuers, networks, and regulators accept them (and customers accept this step) This space will evolve: both implementation & customer behaviour: What I've got my eye out for is: 1️⃣ Will banks treat 3rd party mandate vaults as equivalent to direct user auth? 2️⃣ Will mandate/payment block UX at a merchant level be too much friction for mainstream users? (category level is messy) 3️⃣ Will users trust autonomous commerce, or prefer visible auth steps? 🧠 Deep dive in the comments. with Chinmay Shah

64

6 Comments

Razorpay's Vulcan launch yesterday was not just another fintech press release. It was a massive opportunity hidden in plain sight. And if your AI roadmap is still wrapping general LLMs for operational workflows, you are building on borrowed time. In simple terms, Vulcan is the obvious next step in AI/LLMs, graduating from text, code, image and video LLMs to Domain Specific, aka, specialist foundation models. We have seen this earlier too. Stripe's Payments Foundation Model (May 2025), Bloomberg's BloombergGPT (2023), NVIDIA's ChipNeMo for semiconductor engineering, Nubank's NuFormer (July 2025) for credit and customer intelligence, Revolut's PRAGMA (April 2026) The lesson for us to take note is that --> if we are building Agentic systems with high-velocity event sequences (like logistics tracking SKU flows, financial reconciliation of ledgers or tax invoices), a generic chat LLM is fundamentally the wrong primary decision engine. Be prepared and look-out to your domain specific LLMs. Yes, it may not exist yet, but then is that NOT the OPPORTUNITY that you should be looking into? The real enterprise moat is proprietary operational data wired directly into a closed-loop decision system. Congratulations to Razorpay.

72

6 Comments

Your users are abandoning payments at the auth step. And most fintech PMs don’t know it yet. Reserve Bank of India (RBI)’s 2FA mandate went live April 1. Compliance got checked off. The UX damage is just starting. Every domestic payment in India now needs two authentication factors. UPI, cards, wallets, recurring mandates. Cross-border follows in October 2026. Teams at Razorpay Cashfree Payments JUSPAY and PhonePe are already sitting on drop-off data from the auth step. That number is the real story. Because here is what most PM teams got wrong: OTP is still the default. Reserve Bank of India (RBI) has moved toward cryptographic auth. Most products have not. Mandate and auto-pay flows were not audited before the deadline. They should have been the first thing. The October cross-border deadline is being treated like a future problem. It is a now problem. Compliance and good product are not the same thing. One keeps you legal. The other keeps your users from dropping off at the worst possible moment in their journey. If you are a PM in fintech right now, what does your auth flow look like post-mandate? #productmanagement #Fintech #RBI #UPI #India #digitalpayments

35

1 Comment

SDE and EBITDA are both important metrics — but they are not interchangeable. One of the questions I often see when discussing business valuation is: Which metric should be used? For smaller, owner-operated businesses, SDE (Seller’s Discretionary Earnings) is often the more relevant measure because it reflects the total financial benefit available to one working owner. For larger companies with professional management, multiple owners, or businesses being evaluated by private equity or strategic buyers, EBITDA is generally more appropriate. A key distinction is how owner compensation is treated. With SDE, we typically add back one owner’s compensation. With EBITDA, reasonable management compensation generally remains an operating expense because a buyer may need someone to continue running the business. That distinction matters. Using SDE when EBITDA is appropriate — or EBITDA when SDE better reflects the business — can significantly distort earnings and, ultimately, valuation. Understanding your numbers is not just about preparing to sell. It’s about understanding your business through the eyes of a buyer. #BusinessValuation #SDE #EBITDA #BusinessBrokerage #BusinessOwners #MergersAndAcquisitions #ExitStrategy #BusinessForSale

2

1 Comment

At a certain scale, companies outgrow founder centric execution. What they need instead is: –> institutional leadership –> clearer governance –> and separation between operations and exploration Deepinder Goyal's transition at Eternal fits this pattern. The legacy is already visible: –> a category created –>a platform scaled –> a public company stabilized Stepping back from day to day execution doesn’t reduce impact. In many cases, it changes the altitude at which decisions are made.

9

#BookRecommendation #ByteDance I am a staunch believer that the India technology ecosystem can benefit immensely from key learnings ("What to do?" and also, "what NOT to do?") from mature markets like China (one of the key reasons PROMAFT exists today!). 'Attention Factory', the book does a brilliant job of deeply researching the evolution of ByteDance - a company that has been rather obscure for it scale and impact. The book immaculately captures - 1. Thinking hard about the 'Consumer need' - Social network or Content Platform - using data to prove the same. 2. Evolution to 'Product-Market-Fit' - they were not the first to market! 3. Investing behind scalable technology - a recommendation system that 2x'ed' retention rate and usage in a matter of weeks 4. Rapidly scaling post 'Product-Market-Fit' - customer acquisition, M&A, offline activation - leaving gargantuan competitors like Tencent behind Super valuable for any entrepreneur looking to build & scale globally. Can't wait to re-read this again. Matthew Brennan

23

2 Comments

Why Downstream #SCF in India Breaks So Easily — A Human Story Backed by First Principles A founder once told me, “We did everything right — anchor agreed, bank approved, tech worked.” Two months later, the pilot collapsed. Not because the code broke. Because the people did. Chapter 1: The Hope The anchor’s CFO nods. The procurement head half-smiles. The bank signals interest. The fintech team works evenings building APIs, scoring, dashboards. Everyone believes liquidity is about to flow downstream — finally. This is the most dangerous moment: when hope outruns truth. Chapter 2: The Math Nobody Likes to Discuss An average invoice: ₹2 lakh Tenor: 30 days Discount: 6% p.a. → monthly 0.5% Funding cost: ₹1,000 Platform fee (0.35%): ₹700 Ops + dispute cost: ₹250 Net ≈ ₹450 per invoice before risk. But onboarding an anchor costs ₹2–4 lakh. One dispute delays 15–20 invoices. One inconsistent confirmation wrecks lender confidence. One unexpected GRN mismatch reopens legal review. This is where SCF stops being a financial product and becomes a probability puzzle. Chapter 3: The Psychology That Decides Everything The CFO isn’t trying to sabotage innovation — he’s trying to protect predictability. Treasury yields, cash-cycle stability, relationship capital… that’s his survival. Procurement isn’t resisting tech — they fear losing negotiation leverage. Dealers aren’t reluctant — they’re loyal to anchors, not fintech dashboards. Banks aren’t slow — they’ve simply learned to trust patterns, not promises. Downstream SCF fails because humans optimise for safety, not efficiency. Chapter 4: The First Principles That Matter • Liquidity isn’t the problem — behaviour* is. • Digitisation isn’t the answer — discipline is. • Data isn’t a moat — governance is. • Funding isn’t scarce — trust is. • Substitutes aren’t weak — trade credit is powerful, emotional, relational. Porter would say anchors have all the power. Economists would say margins are too thin. Psychologists would say incentives are misaligned. Founders know all three are true. Chapter 5: What Actually Wins Not speed. Not AI. Not dashboards. Not “MSME empowerment” messaging. What wins is alignment: 1. Show CFO measurable ROI (better procurement economics). 2. Integrate into anchor ERP so deeply that opting out becomes harder than staying in. 3. Price rails — validation, duplicate checks, dispute automation — not just spreads. 4. Prove six months of anchor discipline before demanding lender scale. 5. Build operations like a war-room — because in SCF, ops are underwriting. This is how trust forms. This is how scale happens. This is how the pilot survives month three. The Final Human Truth Downstream SCF is not broken because India is complicated. It is broken because humans are. We optimise for comfort before change, for relationships before rules, for predictability before efficiency. Liquidity solves problems. But understanding people solves liquidity. #SCF #Supplychainfinance

3