Bhavin Turakhia is making one of the clearest founder-led statements yet about where enterprise software may be headed next: not toward another layer of AI features bolted onto old products, but toward a full rewrite.
Turakhia, the Indian entrepreneur behind companies including Directi, Radix, Titan, and banking software firm Zeta, is personally putting $30 million into Neo, a new enterprise work platform he says was designed from the ground up for the AI era. Neo launched internally in April as an integrated workspace that combines project management, documents, file storage, and AI in a single product.
That matters because most of the large productivity suites now being marketed as “AI-powered” still carry the logic of the pre-AI software era. The underlying architecture was built for discrete apps, bounded documents, and human-driven workflows. Neo’s thesis is that this model is now too constrained. Instead of adding assistants on top, the company is trying to rearchitect the workspace itself around AI-native interactions and shared data flows.
Turakhia framed the decision in analogies that point directly at the technical challenge. If the product is going to behave like an AI-first system, he argues, it cannot simply inherit the assumptions of legacy office software and expect those assumptions to disappear.
A unified workspace, not a point tool
Neo’s product shape is important. According to the launch details, it is not being positioned as a standalone chat app or a narrow copilot. It blends project management, documents, and storage with AI inside one environment.
That design suggests a different data model from the one common in incumbent office stacks. In a conventional setup, a project tool, a document editor, and a file store often live in separate services with separate permissions, metadata, and search layers. AI features then have to bridge those boundaries after the fact.
Neo is trying to reverse that sequence. By building the workspace as a single integrated system, it can expose project context, document content, and files to AI workflows in a more native way. In theory, that makes it easier to move from simple text generation to actions that are aware of task state, stored assets, and shared organizational context.
The practical appeal for enterprise buyers is obvious: fewer disconnected systems to stitch together, fewer handoffs between apps, and a tighter loop between content creation and operational work. But the engineering burden rises as well. A unified stack has to reconcile permissions, indexing, storage, collaboration, and inference in one product surface rather than spreading those responsibilities across multiple vendors.
Model-agnostic by design
Neo is also taking a model-agnostic approach, which may be the most strategically important part of the launch. Rather than tying the product to a single AI provider, the company says the platform is designed to allow switching between models as the market changes.
That is a meaningful hedge in enterprise AI, where model quality, cost, latency, and policy constraints can shift quickly. A model-agnostic architecture gives a product team more room to adapt to new foundation models, route workloads by task type, or swap providers if pricing, compliance, or availability changes.
But model-agnosticism is not just a procurement story. It has architectural implications. The abstraction layer has to preserve enough consistency for workflows and permissions to remain stable while still letting the system call different models under the hood. That means careful handling of prompts, retrieval, context assembly, response normalization, and output validation.
In other words, the real technical work is not simply “support multiple models.” It is making model choice an implementation detail rather than a product fracture point. If Neo can do that well, it could offer something incumbents have struggled to deliver: AI capability that can evolve without forcing the rest of the product to be rebuilt each time a provider changes.
The bootstrap signal
Turakhia is not starting this as a venture-backed moonshot from day one. He is funding Neo with his own capital, the same pattern he has used before. That changes the story from a typical startup launch into a founder-controlled infrastructure bet.
A $30 million personal investment is not enough to prove market adoption, but it is enough to signal conviction and buy time. It also means Neo is entering the market without the immediate pressure to optimize for a rapid fundraising narrative. For an enterprise product that needs to work through issues of integration, trust, and deployment, that can matter.
The flip side is that a bootstrap model concentrates risk. If Neo is trying to compete with office software incumbents, it will need to do more than ship polished features. It has to build an operationally credible platform with security reviews, admin tooling, reliability guarantees, and support processes that enterprise customers expect before they standardize around a new system.
Governance will decide whether the architecture is usable
The hardest questions around Neo are not about whether AI can be embedded into workspace software. They are about whether the stack can be made enterprise-ready at scale.
Data governance will be central. A system that blends project data, documents, and storage with AI needs clear control over where information lives, which content is indexed, which users can access which assets, and how AI requests are scoped. Enterprises will want to know how Neo handles isolation between tenants, how permissions propagate through AI workflows, and how it prevents sensitive material from leaking into the wrong context.
Security is equally important. A unified system increases the blast radius if something goes wrong. That makes auditability, access logging, encryption, and policy enforcement non-negotiable. Enterprise buyers will also ask where data is processed, how long it is retained, and what parts of the system depend on third-party AI providers.
Then there is performance. Model-agnostic systems can be more flexible, but they can also introduce latency if every request has to move through routing logic, retrieval layers, and provider-specific abstractions. In a workspace product, that overhead affects the user experience immediately. If AI features slow down document work or task management, the advantage of a unified platform quickly erodes.
The internal launch in April suggests Neo is broad enough to cover multiple work functions, but breadth is a double-edged sword. The wider the integration surface, the more complex governance becomes. For Neo, proving that the product is not only ambitious but also administratively and operationally safe will likely matter as much as the AI itself.
What would count as success
Neo’s competitive claim is not that Microsoft Office or other incumbents cannot add AI. It is that their core architectures may limit how far AI can go if it remains an overlay.
If Neo can show that an AI-first, model-agnostic workspace can handle documents, files, and project management in a coherent enterprise environment, it would strengthen the case for a different product philosophy: build around AI from the beginning, and let the workflow structure follow from that design.
That would put pressure on incumbents in a more fundamental way than another feature launch. Instead of asking buyers whether they want an assistant inside a familiar suite, it would ask whether the suite itself should be rebuilt around AI-native workflows.
For now, Neo is still a thesis more than a market outcome. But Turakhia’s willingness to fund the company personally gives the project unusual clarity. He is not testing whether AI can be appended to office software. He is testing whether the office suite should be redesigned as software that assumes AI is already part of the operating model.
That is a bigger bet, and a harder one. It is also the kind of bet that could matter if enterprises decide the next productivity platform is not the one with the most AI features, but the one with the cleanest architecture for using them.
