Arun Singh

[Arun Singh @ArunsInbox]

@cars24india Abhishek from Cars24 called and explained. This is resolved for now.

[Arun Singh @ArunsInbox]

Most Indian enterprises are sitting on fragmented, siloed data — built for reporting, not inference. Then they drop an AI layer on top and wonder why the outputs are unreliable. You can't fix a foundation problem with a better roof.

[Arun Singh @ArunsInbox]

Two major consulting firms, back-to-back months, caught publishing fabricated research under their own brand. The reputational damage is real. But the scarier part is how many reports didn't get caught.

[Arun Singh @ArunsInbox]

EY pulled a report in May. KPMG pulled one in June. Both Big Four firms. Both retracted because AI hallucinated facts about real clients — UBS, the NHS, Transport for London — who all denied the claims publicly.

[Arun Singh @ArunsInbox]

KPMG's own policy requires human oversight to validate AI content. The report went out anyway. So the process existed — it just wasn't followed. AI didn't break the system. It made the existing failure mode invisible until it wasn't.

[Arun Singh @ArunsInbox]

Manufacturing leaders talk a lot about digital transformation. But when Amazon can undercut your entire merch supply chain with a free app feature overnight — what does your vendor dependency map actually look like?

[Arun Singh @ArunsInbox]

The hard part isn't fixing it. It's admitting that the 'source of truth' question wasn't answered before the project started — and that vendors never surface it because their product looks fine in isolation. It only breaks at the seams.

[Arun Singh @ArunsInbox]

Most factory 'digital transformation' projects don't fail because the technology doesn't work. They fail because nobody decided who owns the current state of a work order when three systems all hold a copy.

[Arun Singh @ArunsInbox]

The diagnostic I use: map your critical state objects — machine status, work order, quality event, material lot. For each one, list every system holding a copy. Then count the reconciliation points. Above five for any single object, you don't have a transformation problem. You…

[Arun Singh @ArunsInbox]

One mega-deal can make a bad quarter look like a milestone. Worth asking how many real signals we've been misreading the same way — celebrating aggregate numbers that mask what's actually happening at the seed and Series A level, where the pipeline gets built or doesn't.

[Arun Singh @ArunsInbox]

Strip out SambaNova's $350M Series E and Black founders are having one of their worst funding stretches in years. That's the actual story behind the 'record quarter' headline.

[Arun Singh @ArunsInbox]

Gené Teare at Crunchbase put it plainly: investor caution may be systematically filtering out first-time founders — who skew more diverse. Risk aversion doesn't distribute evenly. It compounds existing gaps.

[Arun Singh @ArunsInbox]

Digital sovereignty isn't about where the servers are. It's about whether you control your operational history when the business relationship ends. And most manufacturers I talk to have never stress-tested that assumption.

[Arun Singh @ArunsInbox]

Your SaaS vendor gets acquired. The acquirer has a competing product. Your contract renews in 90 days. What happens to 5 years of your operational data? That question — not geopolitics — is what digital sovereignty actually means for most manufacturers.

[Arun Singh @ArunsInbox]

What I'd actually push for in any industrial SaaS contract: open export formats by default, defined SLAs on data retrieval, explicit clauses on what happens to your data post-acquisition, and annual audit rights. Most vendors will push back. That resistance tells you something.