Refold AI

Last Thursday, we brought together a small group of product and engineering leaders for an off-the-record dinner in SF. No slides or agenda, just an honest conversation about what's actually slowing teams down. The same thread came up, again and again: the roadmap is ready. The customer is ready. But the integration work isn't — and that gap costs months. Custom ERP connections, bespoke CRM workflows, systems that were never designed to talk to each other... The last mile between a product that's built and a customer who's live is still, somehow, one of the hardest problems in enterprise software. Different companies, different stacks, different stages, yet all were facing some flavor of the same friction. There's clearly appetite for a better answer. And that's what we're building at Refold. Thanks to everyone who joined us, and stay tuned - this is just the beginning.

Akshay J., our co-founder and CTO, has been named to the Forbes 30 Under 30 Asia 2026 list in AI. When we founded Refold AI in 2023, the vision was clear: replace the manual service work still holding enterprise integrations together. Two years later, we’ve raised a $6.5M seed round led by Eniac Ventures and Tidal Ventures and are scaling hundreds if not thousands of enterprise integrations 90% faster than the industry norm. Akshay is a large part of why. He shipped the product that landed our first customer, and has played a critical role in our growth serving some of the fastest-growing brands in enterprise software, including Rillet, Coupa and more. Every milestone our company has hit has his fingerprints on it. Read more of the feature here: https://lnkd.in/eypZ98zU

AI agents are changing what's possible with building and maintaining enterprise integrations. But the value only lands if security and governance are treated as a first principle — not an afterthought. That requires three things to be true. 1️⃣ A shared context layer. Agents need more than connectivity — they need to know how your systems work, what's been built before, and why decisions were made, all while having enforced data privacy at the source. 2️⃣ Full auditability. When an agent touches ERP data, financial records, or customer workflows, every decision needs to be traceable. Audit trails shouldn't be generated after the fact — they should be a structural consequence of how the context layer stores decisions in the first place. 3️⃣ Human control, precisely scoped. Agents propose, but nothing reaches production without explicit approval. Access controls should be customizable by role — separating who can build, review, deploy, and diagnose. Speed is only an advantage when governance keeps up with it. We broke down best practices for establishing agentic AI guardrails, and how leading enterprises are ensuring integration delivery that doesn't sacrifice security or governance for speed. Read the blog: https://lnkd.in/d4HqVpyr

Based in SF? Dinner is on us next Thursday, May 28. We're getting a small group of product and engineering leaders together for an off-the-record conversation about enterprise deployment and why the last mile is still the hardest mile. Specifically: when your product is ready, but your customer's environment isn't. Fragmented ERPs, CRMs, bespoke systems — and months of integration work standing between deployment and value. We'll be talking about how teams are compressing deployment cycles, scaling delivery without growing headcount, and turning that complexity into an advantage. Best of all: no death by PowerPoint. Request to join: https://luma.com/0xno9lva

Every integration = decisions. Which fields map to what. How edge cases get handled. What the API actually does versus what the docs say. Those decisions live in someone's head, or they get rebuilt from zero the next time. That's why enterprise customization doesn't scale. It's not a tooling problem. It's a context problem. Without a layer that captures what was decided, why, and how, every new deployment starts from scratch, regardless of how many times you've done it before. The Integration Graph is Refold's answer to this. Every deployment writes what it learns — field mappings, API behaviors, exception handling, business logic — into a structured, confidence-weighted system of record. The next deployment starts with that context already loaded. Unlike RAG, it retrieves by structural match. Unlike fine-tuned models, it doesn't freeze at training time. Unlike agent frameworks, it doesn't forget between sessions, it compounds. Across production deployments of the same connector pairs, human hours dropped by 70%. Context is what makes an integration fit a customer, not just connect to their systems. The org that captures it at scale is the one that can actually deliver custom enterprise integrations without rebuilding the same work forever. Swipe to see how the Integration Graph works → Full breakdown on our blog: https://lnkd.in/eKzu2Npj

Refold has been featured in Z47 and OpenAI's latest report as a top startup for implementation automation in the enterprise AI landscape. Enterprise integrations have always been the bottleneck nobody talks about. Months of implementation, teams rebuilt for every deployment, systems that break the moment something upstream changes. AI doesn't fix that by default. You have to build for it. That's what Refold is: an AI integration delivery layer that learns, reuses, and scales so teams stop rebuilding the same work and start shipping outcomes. Learn more about how the AI landscape is evolving + how Indian enterprises are deploying and realizing value in the report: https://lnkd.in/ghmmarpz

I was wrong about the timeline. When I started Refold AI three years ago, my pitch deck said the $400B SI Tax would collapse over the next decade. Investors thought I was being aggressive, CTOs told me I was too early. But I thought I was being conservative. The collapse is happening in eighteen months, not ten years. - Accenture: down 32% this year.  - Cognizant: down 27%  - Infosys: down 28%  - HCL: down 27% - TCS: 24% - Wipro: down 25% These are not rotation losses. TCS's CEO called it "degrowth" on his own earnings call. HCL's CEO told investors to expect a 3 to 5% revenue decline next year because of "AI deflation." When the people running the SI business are admitting on earnings calls that AI is shrinking the work they sell, the structural shift is no longer a thesis. We had built our runway plan for 5 years. We hired with that timeline in mind. Every investor conversation ended with me explaining how much patience this market would require. The market moved faster than we planned for. The lesson is not that we got lucky. The lesson is that structural shifts do not follow the linear timeline founders used to plan. They sit dormant for years and then collapse in months. The math does not change gradually. The customer's perception of the math changes overnight when one earnings call admits the thing the analyst was already writing about. I bet on the structural shift before it had numbers. The numbers showed up faster than I planned. Now the challenge is keeping up with a market moving at the speed I once described as “eventual.” Abhishek Kumar, Akshay J., Tanmay Bhethanabhotla, Drishti Tulsi, Rishabh Nijhawan

Enterprise AI is officially moving past the “demo era.” What enterprises need now is not another AI model. They need infrastructure that can actually operationalize AI at scale. Had an incredible conversation with Jugal Anchalia on the latest episode of the SimplAI Podcast — and one thing became very clear: The biggest bottleneck in enterprise AI is no longer intelligence. It’s orchestration. Every enterprise today is running: • Multiple clouds • Legacy systems • APIs everywhere • Security & governance requirements • Human approvals • Complex workflows • Fragmented data environments This is exactly why the next generation of enterprise AI winners will not just be model companies. They will be the companies building the operating systems, orchestration layers, governance frameworks, and deployment infrastructure that allow AI to reliably run inside real enterprises. One line from the conversation stayed with me: “AI progress is exponential. Enterprise deployment is still linear.” That gap is now the biggest opportunity in AI. At SimplAI , this is precisely what we are building: The Enterprise Agentic AI Operating System — enabling enterprises to orchestrate agents, workflows, systems, governance, and enterprise data securely at production scale. The future of AI will not be won by who has the smartest demo. It will be won by who can deploy, govern, orchestrate, and scale AI reliably across the enterprise. Fantastic conversation with Refold AI on where this market is heading. Watch the full episode here: https://lnkd.in/g7mkGfyF #EnterpriseAI #AgenticAI #AIInfrastructure #AIAgents #EnterpriseSoftware #LLMOps #AIEngineering #WorkflowAutomation #AIInProduction #SimplAI

Most integration work is still done in one-offs. Teams build something for one customer, then rebuild similar logic for the next — accumulating duplication, maintenance burden, and slower delivery with every new deployment. Open data delivery platform Incorta found a better way. Today, their product team can design complex workflows without deep platform expertise or engineering support. Using a plain English prompt, they can describe exactly what a workflow needs to do, and have it built and ready to deploy. That same workflow can then be rolled out consistently across multiple client environments, without starting from scratch each time. What used to require significant technical resource and turnaround time is now something a non-technical stakeholder can drive independently. Incorta can take on more customer demands, move faster across deployments, and keep their engineering team focused on higher-value work (without adding headcount to keep pace). Listen to what Anurag Malik, VP of Product at Incorta, has to say about what their AI integration delivery layer unlocks in the clip below.