Bianca Lewis, Executive Director, OpenSearch Software Foundation, discusses unified AI data infrastructure, agentic AI, open-source governance, digital sovereignty, and why eliminating data silos is essential for scalable enterprise AI.
As enterprises race towards agentic AI, the conversation is rapidly shifting beyond LLMs to the infrastructure that powers them. At the centre of that transformation lies the challenge of managing massive volumes of enterprise data while balancing cost, governance, compliance, and AI performance. Bianca Lewis, Executive Director, OpenSearch Software Foundation believes that the traditional approach of operating separate platforms for search, observability, application monitoring, security analytics and AI workloads has become economically and operationally unsustainable.
In this interview with CIO&Leader, she explains why organisations must move from solving isolated technology problems to building unified data systems, how OpenSearch is positioning itself as the AI data infrastructure layer for enterprise AI, why NVIDIA’s adoption of OpenSearch signals a broader industry shift towards open AI infrastructure, and why open-source governance has become a strategic safeguard against vendor lock-in.
CIO&Leader: In reference to “data tax” that enterprises pay by operating multiple siloed platforms. With Agentic AI pushing organisations towards entirely new operating models, has this fragmented architecture become fundamentally unsustainable?
Bianca Lewis: The idea behind the data tax is actually very simple. The way enterprises have traditionally approached technology no longer works in the AI era.
Historically, organisations solved individual business problems. They bought one platform for observability, another for search, another for security monitoring, another for application performance monitoring, and so on. Each solved a specific problem exceptionally well.
You’re generating the same enterprise data and feeding it into observability, search, security, APM and AI separately. The question is: why pay millions of dollars to duplicate the same data across fragmented platforms when a unified data layer can serve every use case?
But Agentic AI fundamentally changes the equation. Today you are no longer solving isolated problems. You are building intelligent systems that continuously interact, reason and make decisions. Once you start operating at that scale, fragmented architectures become economically and technically unsustainable.
The first reason is economics. Every organisation generates enormous amounts of operational data. That same telemetry feeds observability platforms, AI applications, enterprise search, application monitoring, cybersecurity systems and analytics platforms.
The question becomes: why pay multiple vendors to store, process and analyse exactly the same data?
You’re essentially paying repeatedly for identical information simply because each platform maintains its own copy. That is the data tax. Instead of extracting multiple forms of value from one unified data layer, organisations duplicate infrastructure, duplicate storage, duplicate compute and duplicate operational costs.
The second issue is compliance and digital sovereignty. If your observability platform sits with one vendor, your search platform with another, your security monitoring with a third and your application monitoring somewhere else, you’ve effectively handed your enterprise data to numerous providers.
At that point, you’re relying on every individual vendor’s compliance posture, governance framework and security certifications rather than maintaining direct control over your own data estate.
With regulations around digital sovereignty becoming stricter globally, especially across Europe through frameworks such as the Cyber Resilience Act, organisations increasingly need direct ownership and visibility of the data they generate.
That combination of economics and compliance is why data silos are rapidly becoming unsustainable in the age of enterprise AI.
CIO&Leader: From a technology perspective, how does OpenSearch actually unify these fragmented environments into a single data layer?
Bianca Lewis: The objective is to give organisations a single AI-native data infrastructure capable of supporting multiple workloads simultaneously rather than forcing them to stitch together disconnected products.
OpenSearch was designed as an AI data infrastructure layer. Its foundation supports enterprise search, observability, security analytics, application monitoring and AI workloads on the same underlying platform.
For example, using OpenSearch’s Launchpad capabilities, organisations can rapidly build AI-native search applications. Those applications leverage hybrid search, combining traditional keyword search with semantic vector search depending on the use case.
That flexibility matters because different enterprise workloads require different retrieval techniques.
Once those applications are deployed, the same platform can monitor them through integrated observability, allowing organisations to manage AI agents and production systems within one unified architecture.
The goal isn’t simply consolidation. It’s enabling enterprises to move from isolated experiments towards safe, scalable and production-ready Agentic AI.
CIO&Leader: OpenSearch was recently integrated into NVIDIA’s NeMo Agentic AI platform. What does it mean when one of the world’s leading AI infrastructure companies chooses an open-source search engine as the storage backbone for autonomous AI agents?
Bianca Lewis: Naturally we’re delighted that NVIDIA selected OpenSearch. In reality, many Fortune 500 organisations already rely on OpenSearch, although not every company publicly announces those decisions the way NVIDIA did.
The significance extends beyond one partnership. It validates an industry shift towards open AI infrastructure. Organisations increasingly want AI platforms that are transparent, scalable and community-driven rather than proprietary black boxes.
OpenSearch benefits from an exceptionally large global community, with billions of downloads, thousands of active contributors and an enormous developer ecosystem continuously identifying issues, improving features and strengthening security.
That community model creates resilience. If vulnerabilities emerge, they are often discovered and addressed extremely quickly because thousands of engineers worldwide are actively contributing.
As a foundation, our responsibility is to ensure enterprises can confidently adopt OpenSearch as the underlying data infrastructure powering Agentic AI, whether they’re companies like NVIDIA or enterprises across regulated industries.
We’ve also introduced long-term support releases, structured vendor ecosystems and enterprise support programmes so organisations can adopt open infrastructure without sacrificing enterprise-grade operational support.
CIO&Leader: Why should enterprises trust conversational memory to an open-source search platform instead of specialised vector-only databases?
Bianca Lewis: First, I’d argue that virtually every serious database today has become a vector database in one form or another.
The real question isn’t whether vectors are available. The question is whether vectors alone are sufficient. Vector search is extremely powerful for semantic similarity. You search for one concept and retrieve related neighbours automatically.
However, enterprise workloads aren’t always semantic. Sometimes organisations search for precise part numbers, policy IDs, invoice numbers, years or sequential identifiers where exact keyword matching is critical.
If you rely exclusively on vector search in those scenarios, you introduce unnecessary hallucinations while dramatically increasing token consumption. Every inaccurate retrieval means additional AI reasoning, more token usage and ultimately higher operational costs.
OpenSearch solves this through hybrid search. It allows organisations to combine keyword search, semantic search and vector retrieval dynamically depending on the workload. That flexibility significantly improves retrieval accuracy while making the economics of Agentic AI substantially more sustainable.
Accuracy directly impacts AI ROI. If your retrieval layer isn’t accurate, your AI system consumes more compute, more tokens and delivers lower-quality outputs.
CIO&Leader: At a time when many vendors are abandoning open-source licensing in favour of commercial alternatives, how does the Linux Foundation governance model protect enterprise buyers?
Bianca Lewis: That’s actually one of the most misunderstood aspects of open source.
Many technologies begin life as open-source projects inside commercial companies. Initially they’re openly licensed, but because the company owns the intellectual property, it can later decide to change the licensing model.
We’ve seen that happen repeatedly across the industry. Projects such as Elasticsearch, Redis, MongoDB and others have all changed licensing structures over time.
The difference with OpenSearch is governance. OpenSearch belongs to the Linux Foundation through the OpenSearch Software Foundation. The foundation is a non-profit organisation. We do not exist to maximise shareholder returns. We exist to ensure OpenSearch continues innovating as an open ecosystem.
Once a project belongs to the Linux Foundation, no company can suddenly decide to close the licence. That’s the real insurance policy for enterprises—your rights, your data and your choices cannot simply be taken away.
That means no individual company can suddenly decide to close the licence because revenue expectations weren’t met. The funding we receive from organisations such as IBM, AWS, SAP, Uber, NetApp, ByteDance and many others isn’t distributed as profit. It’s reinvested directly into engineering, community development, innovation and ecosystem growth. That’s fundamentally different from open-core business models.
In open-core software, the basic functionality may be free, but the enterprise capabilities remain locked behind commercial licensing. Eventually customers become dependent on one vendor’s roadmap, support model and pricing decisions.
Under a foundation-led governance model, enterprises retain choice. They can select different support providers, different cloud providers, different consulting partners and different deployment models without losing ownership of their data.
That freedom is precisely what open source was originally intended to deliver.
CIO&Leader: India has emerged as one of the world’s fastest-growing AI startup ecosystems. How is the OpenSearch Software Foundation positioning its ecosystem to support India’s innovation journey?
Bianca Lewis: India is one of the most strategically important markets for the OpenSearch Software Foundation. The numbers alone tell the story.
India has thousands of AI startups, ranks among the world’s leading startup investment destinations and continues attracting massive infrastructure investments from hyperscalers building data centres across the country.
India isn’t just another market for OpenSearch. With thousands of AI startups, one of the world’s largest developer communities and massive infrastructure investments, India is becoming one of the global centres of innovation for open AI infrastructure.
That makes India not only an important market but also a major innovation hub for OpenSearch.
Our first objective is lowering the barrier to innovation. Because OpenSearch is fully open source, startups can begin building sophisticated AI infrastructure without immediately worrying about expensive licensing models or enterprise feature restrictions.
They gain access to the complete platform from day one. As those companies scale, they can leverage a growing ecosystem of vendors, support providers and cloud deployment options without having to redesign their underlying architecture.
The second advantage is talent development. India produces one of the world’s largest pools of software developers every year.
OpenSearch allows those developers to contribute directly to globally recognised open-source projects, participate in GitHub development, join community discussions, earn certifications and build internationally recognised technical expertise.
For startups, that’s incredibly valuable. They aren’t simply consuming technology. They’re actively contributing to the technologies that power next-generation enterprise AI.
That combination of open innovation, community participation and technical skill development creates a very strong foundation for India’s deep-tech ecosystem going forward.