Prashant Chandel

📌 Trends don't make business history. It's made by decisions. This account studies those decisions. Forget the fluff. We provide compressed, judgment-focused breakdowns of the choices that defined companies. It’s like a mini-MBA in your feed, distilled by advanced AI agents that plan, critique, and rewrite until only the truth remains. Follow to learn how to think, not just what to think. 🧠 👇 Start reading the threads below.

The real question isn't who orchestrates AI workflows — it's who owns the write-back. Every enterprise agent eventually needs to update a financial record, change inventory, or process payroll. That happens in ERP, not CRM. SAP and Oracle own that last mile.

Jevons Paradox might apply to software engineers too - we might end up needing more of them to build and ship products using AI

@brucedando_ The data readiness framing is the right one. Joule is only as good as the master data underneath it - and most SAP customers have years of inconsistent vendor records and undocumented hierarchies.

The 3% production adoption stat is the number that matters. 40+ agents and 2,400 skills is impressive engineering. But the gap between capability and customer readiness is the real story. The reason most customers aren't deploying Joule agents in production: the data underneath isn't ready. An AI agent making cash positioning decisions needs clean, governed, contextualised master data. If your vendor records are duplicated, your bank account mappings are inconsistent, or your cash flow hierarchies are undocumented, the agent will automate the wrong decisions. This is why the Reltio acquisition matters. Golden records and entity resolution are the prerequisite for production-ready AI agents. And it's why data products in Datasphere need to exist before Joule can reliably consume them. The capability is there. The data foundations aren't. That's the 97% gap.

Oracle launched 22 production AI agents in March. It got almost no coverage. Everyone's watching Salesforce vs. ServiceNow fight for the "enterprise AI OS." They're both playing the wrong game.

@stats_feed I'm still not over Covid and dont think it started 6 years back so maybe yes

Do you believe time feels faster as you get older?

Built an agentic system - strategy researcher and analyst who talk to each other discussing latest developments then a complier takes their discussion and writes a blog post and posts it to my personal blog and x. Checkout the latest blog: prashant-chandel.org/blog/2026-04-1… It talks about one of the biggest problems in agentic AI systems for enterprises which is currently unsolved i.e. how do agents across multiple vendors SAP, Oracle, Salesforce, Servicenow communicate to each other and who holds the best cards to win and why.

What are you building this weekend? Drop your project URL Let’s drive some traffic

Biggest question in enterprise AI. The default architecture of most large enterprises. Financial System of record: @SAP Customer system of record: @salesforce IT workflow system: @ServiceNow HR system: @Workday You have four "systems of record" that are authoritative for different domains, none of which talk to each other natively. When an agent needs to process a customer return - which requires reading CRM data, updating inventory in ERP, triggering an IT workflow for exception handling, and logging a finance entry - which system is the agent's home? Whose orchestration layer governs the cross-system write-backs? This is the actual enterprise AI OS question, and it is not answered by any of the companies currently competing for the title. ServiceNow has the most plausible answer: it's already the company that sits above all these systems for governance and workflow management. But ServiceNow has never had authoritative write-back access to SAP or Oracle. It manages about those systems. The moment of truth for the "AI Control Tower" thesis is whether ServiceNow can get ERP-level write-back access or whether SAP and Oracle will foreclose that path to protect their own agent strategies. This is a classic platform perimeter battle, and it will be decided not by product features but by enterprise contracts and API access terms. The company that controls the APIs to the systems of record controls the orchestration layer. That is not a technology question. It is a business development and legal question, which means it will take longer to resolve and create more durable outcomes when it does.

@DavidKPiano Not sure if strawberry test or turing test

This is bad

Full deep-dive at prashant-chandel.org/blog. Follow for weekly strategy threads. #AI #Strategy #EnterpriseAI

My take: this ends like the middleware wars — domain oligopolies, not a single winner. The orchestration tax gets collected. The question is whether it goes to Salesforce, ServiceNow, Oracle, SAP, or a company nobody is watching yet. The database always wins. Eventually.