Inc42 Hosts Playbook Session On Agentic AI ROI
At the Inc42 AI Summit 2026 session on June 10, "The Operator Playbook For Building And Scaling Agentic AI Systems," panelists said fear of data leaks, not model capability, is the biggest blocker to agentic AI ROI. Skyflow CPO Amruta Moktali said "everyone is blocking data and then they say they are not getting the value out of the agent," arguing firms should protect sensitive fields rather than lock agents out of context entirely. The panel, moderated by Ideaspring Capital's Naganand Doraswamy and including executives from Razorpay, Wingify, Shipsy, and Avalara, cited a Deloitte 2025 survey of 1,854 executives finding that only 10% of organizations report measurable returns from agentic AI. Avalara's Job Sam Koshy said the operator playbook means working backwards from workflows rather than chasing the newest model.
The panel's most concrete, actionable claim is not that agentic AI underdelivers, it is why: according to Skyflow's CPO, organizations sabotage their own ROI by blocking agents from the business context they need, then blaming the technology when it underperforms, a self-inflicted failure mode that is fixable with better data-access design rather than better models.
What happened
At the Inc42 AI Summit 2026 on June 10, a panel titled "The Operator Playbook For Building And Scaling Agentic AI Systems That Deliver Real Business ROI" brought together Ideaspring Capital founder Naganand Doraswamy (moderator), Wingify cofounder and CPTO Ankit Jain, Razorpay CPO Khilan Haria, Skyflow CPO Amruta Moktali, Shipsy cofounder and COO Dhruv Agrawal, and Avalara's APAC partnerships head Job Sam Koshy, per Inc42's report. The session opened by citing a Deloitte 2025 global survey of 1,854 executives that found only 10% of organizations have seen meaningful, measurable returns from agentic AI deployments.
For practitioners
Moktali's central point was that data-security fears, not agent capability, are the primary ROI blocker: "Everyone is blocking data and then they say they are not getting the value out of the agent." Her recommendation was to protect specific sensitive fields within documents rather than withhold entire datasets, so agents retain the business context they need. Koshy framed Avalara's approach as working backwards from what a model can reliably deliver for speed and accuracy, rather than chasing every new model release, while Shipsy's Agrawal argued the highest-ROI teams will be those with the clearest operational boundaries, not the newest models.
What to watch
Because two panelists (Skyflow and Avalara) sell data-security and compliance tooling that benefits from an enterprise-adoption narrative centered on data governance, their specific framing is worth weighing alongside independent evidence. The Deloitte figure itself, however, comes from a named, citable global survey rather than the panel, giving the core ROI claim a verifiable anchor independent of the vendors on stage.
Key Points
- 1A Deloitte 2025 survey of 1,854 executives found only 10% of organizations report measurable ROI from agentic AI deployments.
- 2Skyflow's CPO said blocking agents from business context, not model capability, is the main self-inflicted cause of poor agentic AI ROI.
- 3Panelists from Razorpay, Wingify, Shipsy, and Avalara argued clear operational boundaries matter more for ROI than adopting the newest models.
Scoring Rationale
Fully verified against both the Inc42 report and the underlying Deloitte survey; the piece now includes genuine named executive quotes (Skyflow, Avalara, Shipsy) rather than an anonymous 'session,' and rests on a real, citable 10%-ROI statistic. Kept at the lower end of notable rather than higher because it is a summit panel recap with some vendor-interest framing (Skyflow and Avalara sell into the exact narrative they are promoting) rather than independent research.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
