Shubhendu Tripathi: AI and ERP Strategy Consultant
Shubhendu Tripathi is an AI strategy consultant in Toronto specializing in ERP consulting, digital transformation advisory, SAP S/4HANA implementations, and enterprise AI strategy for mid-market and large organizations.
Insights
- Field Note from India: Nobody Asked What Your AI Can Do. They Asked Whether It Can Be Certified.
Field note from the Team Canada Trade Mission to India: aerospace, agriculture and space buyers ask if AI can be certified and controlled, not what it can do.
- Headless ERP Removes the Screen. Some of Your Controls Only Existed There.
Agents that call ERP APIs bypass rules built into the screen: page layouts, client scripts, form rules. The Screen Test finds them before an agent does.
- Your GL Renewal Just Excluded Generative AI. Your Vendor's Cap Is Twelve Months of Fees. Nobody Is Holding the Loss.
Insurers can now exclude generative AI from general liability, and AI vendors cap liability at 12 months of fees. Here is who is left holding the loss.
- The Safest Small Model Is the One You Distilled Yourself. Everyone Is Doing the Opposite.
Most enterprises adopt small open models with no chain of custody. Distilling your own from a licensed frontier model gives you the records auditors want.
- Your ERP AI Is Metered by the Action and You Have No Circuit Breaker
Enterprise software is moving from per user to per action pricing for AI agents, and most buyers cannot stop a runaway agent. How to build a circuit breaker.
- The EU Delayed the AI Act Obligations Your Vendor Builds. The Ones You Deploy Landed on August 2.
The AI Omnibus moved high risk deadlines to December 2027, but deployer duties such as Article 50 transparency did not move. What enterprises must do now.
- Open Weights Are a Decoy. The Moat Was Never the Model.
Capable AI models now cost nothing while compute, power and data centres get scarcer. Why open weights do not decide who controls AI, and what does.
- Your ERP Vendor Refused to Build Its Own Model. Take the Hint.
SAP, Oracle and the other ERP vendors chose to rent frontier models rather than build one. What that decision tells enterprises about their own AI strategy.
- Your AI Rollout Is a System of Inspection. That Is Why Nobody Is Using It.
AI rollouts fail the way CRM did: they ask staff to feed management and give nothing back. A test from four technology waves for tools people actually use.
- The Model Stopped Being Worth Money. The Harness Started Collecting Instead.
xAI open sourced Grok Build after a researcher showed it uploading whole repositories. The lesson: the model is cheap, and the harness is the risk surface.
- The Most Expensive Knowledge in Your Company Is About to Be Free.
When company knowledge becomes searchable in plain language, a newcomer works like a veteran by day two. What that means for teams, careers and hiring.
- AI Is a Compression Engine, Not an Invention Engine. That One Rule Explains the 95 Percent.
One test predicts whether an AI project will work: compare the output you want with the input you can give. Why AI compresses well and invents badly.
- SAP Wants to Be Your Agentic Workforce. Your ERP Cannot Be the Whole Story.
SAP is selling an agentic workforce inside S/4HANA, yet most enterprise work lives outside the ERP. A practitioner reality check on SAP and AI agents.
- What Is the A2A Protocol? How AI Assistants Are Learning to Work Together
The Agent2Agent (A2A) protocol lets AI assistants talk to each other and share work. A plain English guide to what A2A is, how it works, and why it matters.
- What Is Model Context Protocol? The Simple Idea That Lets AI Actually Do Things
Model Context Protocol (MCP) is the universal plug that lets AI connect safely to your tools and data, like USB-C for AI. A plain English guide to MCP.
- Quantum-AI in 2026 Is a Press-Release Market, Not a P&L Market
Quantum AI headlines describe real engineering, but nothing a CFO can model yet. The two clocks to separate: quantum advantage and post quantum migration.
- Stateful Agents Are a Lock-In Mechanism, Not a Feature
An agent that learns your business accumulates switching cost on the vendor's platform. Four contract clauses and four questions to ask before you sign.
- The Lights-Out Factory Is a Labor Theory in a SKU Complexity World
Direct labor is 10 to 15 percent of COGS, while changeover, yield and forecast error cost far more. Why AI in manufacturing should target complexity.
- Field Note from Sapphire 2026: The 'Almost Right' Doctrine
At SAP Sapphire 2026, Christian Klein said AI must be accurate, not almost right. Why that line makes governance, not the model, the AI on ERP battleground.
- SAP's New API Policy Isn't Lock-In. It's the First Honest Architecture Spec for Agents on ERP.
SAP's API Policy v.4.2026a has been called lock-in. Read it as the first honest spec for running AI agents on ERP, and what CIOs should build next.
- Your CFO Will Kill Your AI-in-ERP Project Before Regulators Do. They Should.
High risk AI inside an ERP can carry $400K to $1.2M per system per year in compliance cost most business cases never model. The cost stack and the fix.
- The ERP AI Payoff Is Real. Your CFO Will Sign Off When You Show Them Three Controls.
Only 7 percent of CFOs see high ROI from finance AI. The gap is three SOX grade controls that let finance sign off on AI inside the ERP.
- Your Quebec Law 25 Exposure Is Hiding in Three ERP Defaults. You Have Not Been Audited Yet.
Quebec Law 25 invalidates three default settings in most ERP deployments, and AI makes the exposure worse. What the defaults are and how to fix them.
- The Mid-Market ERP Playbook Was Written Before AI Existed. You're Still Following It.
The mid-market ERP playbook was written when ERP was a ledger. Five selection criteria that matter now that AI makes ERP the operating spine of the company.
- Your AI Safety Framework Is 200 Pages. The Real Risk Is One Word: Confidence.
Enterprise AI safety frameworks keep growing while the real risk is confident wrong answers. A one page, 12 question audit that replaces 200 pages.
- Open Weights Just Won. Data Governance Is the Only Moat You Have Left.
With capable open weight models free to download, model choice no longer differentiates. Data quality, lineage, ontology and governance are the moat left.
- You'd Terminate Any Employee Who Operated Like Your AI Agents
Enterprises apply SOX, least privilege and separation of duties to people, then give AI agents more access than the CFO. How to govern agents like staff.
- The Most Dangerous Person in Enterprise AI Is the One Who Says 'We Can Do This Ourselves'
Most enterprise AI pilots never reach production, and going it alone is a leading reason. Why the DIY instinct backfires on AI and what to do instead.
- Three CEOs Quit Because of AI. The Real Question Is Why More Haven't.
Three Fortune 500 CEOs cited AI when they stepped down. Why it signals a structural leadership crisis, and why boards are reaching for the wrong playbook.
- AI Compliance Frameworks Are Making You Less Compliant
Enterprises spend millions on AI compliance frameworks that feel safe but leave real risks open. Five blind spots no framework covers and what to build.
- AI Agents Are the New Microservices: Most Enterprises Need Workflows, Not Autonomy
Vendors sell AI agents the way they sold microservices in 2016. Most enterprises need structured workflows with human checkpoints. How to tell the difference.
- Beyond AI Pilot Purgatory: A 5-Pillar Strategy for Enterprise AI That Actually Ships
Most enterprise AI fails because it is run like an IT project. Five pillars that separate companies shipping AI to production from those stuck in pilots.
- RAG Is a Crutch, Not a Solution
Enterprises default to retrieval augmented generation for every AI problem, but RAG cannot fix bad data. What CTOs should focus on instead.
- Your AI Strategy Is Incomplete Without an ERP Strategy
Your ERP is your largest structured data asset, so your AI ambitions depend on it. Why AI strategy and ERP modernization must be planned together.