Why AI Should Make a University’s Own Faculty More Valuable: Cloudforce CEO Husein Sharaf on Governed AI Beyond the Chatbot

Key Highlights

NebulaONE is positioned as foundational AI infrastructure for entire universities rather than a niche tool. Why did you start with higher education, and what did you see on campuses that made them the right place to prove the model?

Higher Ed is one of the few industries where you can get access to tens of thousands of new software users in a single implementation, and those users are not afraid to try a new product and generate a ton of useful insights very rapidly. Combine this with the varied nature of the audiences (students, faculty, staff, researchers, etc.) and the constant influx of new users each semester, and you have a software testing ground unlike any other. If you can survive as a new product in this environment, and you can gain organic user adoption through unique value creation and peer-to-peer sharing, you have something quite special. Not to mention that the people likely to be most impacted either positively or negatively by AI are new grads, and as an organization we want to be a force for good in contributing to a positive outcome for the millions of students who have access to our platform.

You've said the latest AI models shouldn't be a luxury reserved for the few who can afford high subscription costs or trade away their private data. What does equitable AI actually look like in practice, and how does your approach differ from institutions simply buying per-seat licenses?

From a practical perspective, buying thousands of per-seat licenses for a single frontier model provider’s product, like ChatGPT or Claude, makes very little sense for large, distributed environments like a university or a state government. The reasons are obvious: very few can afford to license every user at any set dollar amount monthly, with such large audiences the users’ needs and preferences will require different models, and the nature of AI inference cost means these budget-constrained orgs are penalized both for underused licenses (which our data implies is over 50% of AI licenses in our customer base) and for users that exceed a certain tokens-per-month limit (and are subsequently charged overages). Our model is simple: pay a flat platform fee for the entire organization per year (typically pennies per user per month when calculated at scale) and then pay only for the models you use on a consumption basis. No model lock-in, no wasted per-seat licenses, no unexpected overages because every user has an administrator-defined monthly budget that is transparent to them and to organization leadership on day 1.

The platform lets customers bring their own model, from providers such as OpenAI, Anthropic, and Meta, inside their own private cloud tenants. With new models arriving constantly, how do you help an institution choose among them while keeping control over data sovereignty, compliance with FERPA, HIPAA, and GDPR, and cost?

Keeping up with the barrage of new AI models released weekly is a bit like playing wack-a-mole these days, but our team has well-defined processes to handle this work for our customers and to ensure we always have the latest models and capabilities available to all our platform users with every release. We do the heavy lifting of vetting every new model for safety, privacy, compliance, capability, performance, cost, and environmental impact centrally, and all our customers benefit collectively as a result. There’s no reason every institution should have to do this work alone, and because they each can weigh in on what they feel is important, and we listen to their needs intently, we’ve been able to cultivate a collective that is stronger than any one institution on its own. We host monthly customer advisory board and office hours meetings (attended by over 50 institutions regularly) where we create space for peer-to-peer sharing of best practices, lessons learned, and input into our platform roadmap.

NebulaONE is built exclusively on Microsoft Azure; you were named Microsoft's global Education Partner of the Year in 2025, and Microsoft's venture fund M12 backed your Series A. What are the advantages of being that tightly tied to one cloud, and how do you think about the risk of dependency, or of competing with Microsoft's own AI offerings?

We’ve built a strong and mutually beneficial partnership with Microsoft over a decade of working together, and we’ve gained an incredible distribution channel for nebulaONE as a result. We chose Microsoft as our hyper-scaler partner very intentionally, due to a few key factors that matter to us and to our customers:

1. Every large institution in the world buys from Microsoft today. Regardless of which cloud they choose to operate in, we have yet to find a single one that doesn’t have a relationship with Microsoft.
2. Microsoft has invested far more heavily in data and AI than others, arguably well ahead of the curve, and that investment was proven by their early backing of OpenAI and their rapid integration of the GPT models on Azure.
3. From a security, identity, data sovereignty, and compliance perspective, no other cloud provider has the depth and breadth of services and capabilities that Microsoft offers out of the box, and at truly global scale.

What that means for Cloudforce is we can focus on what we do best; spending time with the actual humans who are solving real-world problems with AI, and building a platform our users and administrators love as a result. Microsoft takes on the burden of running safe, reliable, highly resilient infrastructure in every region of the world where our customers rely on nebulaONE to deliver.

Cloudforce earned HIPAA certification this year and signed its first healthcare customers. How different is healthcare from higher education as a market for governed AI, and what has to be true before a hospital or health system lets AI near sensitive workflows?

Large user audiences, sensitive data sets, and strict compliance requirements are our typical operating constraints in both Higher Ed and Healthcare. That said, the margin for error in a clinical setting is certainly tighter given the nature of the potential outcomes, and we take that responsibility incredibly seriously at Cloudforce. The work we’ve done thus far with our customers in healthcare centers primarily around operational and research use cases, as we collectively gain greater confidence in this technology’s ability to deliver accurate and reliable results in every scenario. That said, saving a clinician from spending an extra hour a day filling out insurance claims means they can spend that hour with their patients, and that space to focus should result in better patient outcomes even when our AI isn’t directly involved in the clinical interaction. Working closely with Microsoft’s Healthcare Data Services team, we’ve brought capabilities like the Healthcare Agent Orchestrator for tumor board evaluations into nebulaONE as a turn-key solution, and we’re excited to continue pushing the boundaries of the Frontier for clinical workflows.

Part of the Series A is going toward vertical-specific AI agents for education and healthcare, and Cloudforce pairs deployments with forward-deployed engineers and change management. In your experience, what is the real barrier to adoption: the technology, the policies, or the people, and where do agents fit?

The biggest barriers to adoption that we’ve identified firsthand are a combination of all of the above, but tend to start with people, process, and use case identification. AI is a general-purpose technology, meaning it can be applied to just about any business problem, but identifying the right use cases that drive both ROI and end-user delight is a far more nuanced exercise than most organizations realize. Every large enterprise has hundreds or thousands of use cases for AI, but less than 20% of them are fit-for-purpose and worth the investment of resources in this moment. That’s where our customers arguably benefit the most from our experience having seen what works and what doesn’t across well over a hundred enterprise environments. Building repeatable patterns, agents, data integrations, and the change management, training, and strategic communications around them has been a heavy focus for our team in the last year. The net result is we can deliver high-value agents in the first 30 days of a new nebulaONE implementation, along with very targeted OCM, that drive rapid user adoption and measurable ROI.

At the time of your funding announcement, nebulaONE served more than 90 institutions and over 3 million users with 100% subscriber retention in 2025, and you've since announced plans for 250 new jobs and a doubled headquarters in National Harbor. How do you scale delivery without losing the hands-on, forward-deployed model, and why did you choose to double down in Maryland?

One unique benefit of building an AI platform is we’re also a heavy user of our own technology, and we’re embedding AI agents and workflows in every aspect of how we deliver to our customers. What that means practically is we can grow our customer base and our revenue at a rapid pace, while growing headcount responsibly in the areas where we believe a human touch has the greatest impact. Humans want to buy from other humans, they want to build relationships through personal connection, and their needs are most deeply understood by people with experience and empathy. These are the areas we’re choosing carefully to invest in, while AI augments tasks that don’t create the same level of leverage when done by a human. Choosing to stay and grow in Maryland was an obvious choice, as we’ve received incredible support from both the State and our County to continue to expand our footprint here, and we believe wholeheartedly in investing our profits back into the community that enabled our success.

Looking two to three years ahead, what is one prediction about how institutions will use AI that you're confident about, and that most of your peers would currently disagree with?

Within three years, I believe AI will make a university's own faculty more valuable to its students. That runs against the idea that increasingly capable models will make it matter less where you study or who teaches you. My prediction is that institutions will compete on how well their AI expresses what makes their education distinct.

Faculty spend careers developing expertise and learning how to help students acquire it. That includes the questions they ask, the boundaries they set, and the judgment they want students to develop. Institutions will embed those choices into AI that supports students throughout their courses. A student working through a problem at midnight should get help shaped by their professor's approach and intended learning outcomes. Faculty should be able to extend that experience without surrendering their work as training material for someone else's next model.

For students asking whether their degree is worth the cost, access to another chatbot is a thin answer. More access to the expertise they chose that university for is a much stronger one. That's what we're building toward with nebulaONE: faculty shaping an AI-supported learning experience only they could deliver. The more students learn with AI, the more they should benefit from what makes their university worth attending.