ZohoDay 2026 – How Newcross Healthcare builds AI apps on a governed platform
Mo Umergi and Jon Reed at ZohoDay 2026
In my ZohoDay 2026 podcast recap with Thomas Wieberneit, I touched on the persistent gap between where customers are at with AI – and where vendors are.
This is (somewhat) the case with Zoho as well. Zoho has articulated a bold vision of AI that includes model right-sizing, a context layer of governed data, a full AI stack, and data sovereignty.
Zoho argues that out-of-the-box LLMs are becoming commoditized, and that they’ve anticipated the market’s de-valuation of software. I can understand why customers might be in a different place. Customers will look to Zoho to build that AI future, while they contend with industry challenges of their own.
On the other hand, we can learn plenty from customer realities. The customers that spoke at ZohoDay were all knee deep in some kind of data quality, harmonization, or analytics project – what I think of as ‘AI readiness’ initiatives. However, one customer at ZohoDay was able to speak to Zoho’s agentic AI pursuits firsthand – and in a vivid way.
Newcross Healthcare on AI: “a governed architecture for operational intelligence, not dashboard theater”
That customer was Newcross Healthcare. During my on-stage interview with Mo Umerji, Head of Enterprise Architecture at Newcross, he described their thoughtful-but-bold AI moves in the healthcare field. After ZohoDay, Umergi took it further, via a provocative LinkedIn post:
AI is not the strategy. Architecture is. Models are rapidly commoditising. Access is equalising. Feature parity will follow.
The real advantage will not be who “has AI.” It will be who has:
- Governed, structured data
- Unified platform architecture
- Operational workflows designed for intelligence
- Strategic control of their stack
AI layered onto fragmented systems magnifies complexity.
AI embedded into disciplined architecture compounds enterprise value.
That’s the real divide.
We’ll see much more of this in 2026: not all AI implementations are create equal. How did ZohoDay stack up with Umergi’s take? He added:
Own the infrastructure.
Govern the data.
Design for context.
Simplify deliberately.
At Newcross Healthcare Solutions, this is exactly the direction we’ve taken – unifying CRM, workforce systems, finance, analytics, and bespoke care management into a governed architecture built for operational intelligence, not dashboard theatre.
That sets up our on-stage discussion. But we didn’t start with AI – we started with the pressures/opportunities of the UK healthcare economy. Newcross is a leading UK healthcare agency, employing 165,000 registered healthcare professionals at over 4000 client sites. I asked Umergi: the US healthcare system is a hot mess. How is Newcross navigating on the UK side? As Umergi told us:
We have almost in excess of 120,000 healthcare workers providing care services. One of the biggest challenges was: how do you get all of this data into one unified platform, providing the relevant information at the right time to the care workers, to be able to provide a quality service to our clients?
The Newcross data strategy: a unified data platform across applications
Umergi has fifteen years of experience working on Zoho’s platform – including the last five years at Newcross. How has Zoho impacted the data/service quality challenge?
Where Zoho has really helped us is not only to drive cost down, but more importantly, how we amalgamate all the data from different data sources – and bring it all in one unified application… The advantage with Zoho is: because it has such a breadth and depth of applications, almost 45 applications in Zoho One, that means we can transfer data from one application to another application, all in real-time. In turn – a reduction in manual efforts, reduction in deduplicated information, and an increase for us in terms of how much we can actually, actually deliver.
The secondary point was obviously the cost benefits of licensing in the UK, to pay around 35 pounds payments per user license. Compare that with some of the other competitors, there is no clear comparison.
The real payoff of a modern application platform? Build and deploy (well-governed) data applications quickly. AI can be a factor here, but it’s not the only driver. So it went for Umergi’s team: they built a mission-critical care management application with Zoho Creator.
Building a data-driven application for care management – Newcross on Zoho Creator
Why? What’s the problem with third party care management systems? If you want to track care delivery to people in their own homes, you’re looking at clunky/expensive solutions. Umergi:
We were paying in excess of a million pounds in licensing fees. The clunkiness, or the disadvantage we had in using a third party care management application, was that it wouldn’t integrate into our internal systems – and the data resided elsewhere.
Back to that AI readiness theme: Newscross now has an data imperative with competitive implications. But that doesn’t change the data privacy rigor. As Umergi told us:
When you talk about patient medical records, it’s a lot of sensitive personal information, which meant having access to that data, which belong to a third party application, trying to get that data into our own data lakes was becoming very time consuming – andsometimes nearly impossible.
So the business made a decision to create our own care management application using Zoho Creator. In an ideal scenario, you would have six months to plan a project like this. Unfortunately, because we were coming to the end of the agreement with the third party application, we had six months to develop it.
Six months on the clock – but the application was built. Result:
The Zoho Creator app we have now developed in-house is used by all other healthcare workers providing frontline services. It also integrates data in real-time with all of our Zoho ecosystem.
Getting AI results with data apps
Three months after that go-live push, Umergi’s team is fine tuning, sorting out the glitches you find after any new software go-live. This provides the chance to pull in user feedback:
What we’ve now done is: we’ve taken feedback from the frontline staff, seen where the good the bad, and the ugly is, and we’ve recognized where the pitfalls are in the application. Now we’ve agreed with the Zoho Creator team to create a version two of this. So we’ve not only had cost-saving benefits, but it’s also meant we can transfer data from one application to another application in real-time, which is fundamentally really important to us.
In health care data scenarios, the service stakes are high:
If you’ve got a care worker in somebody’s home delivering care services, and that could be people with disabilities, or somebody at the end of life, it’s really crucial that they have the information at their fingertips when they really need it. So this Creator application, integrated with Zoho CRM and Analytics, gives us that full 360 view.
But today’s executive teams want to move fast with AI. They are likely to ask, “can’t we build this with ____ (LLM du jour)?” Umergi was ready for this issue: he experiments with all these tools as well.
You can use tools like Replit, Claude, and Lovable to create proof of concept apps. Where they really don’t work is when you try and create enterprise-level applications that talk to one another. I’ve run many tests just using Replit. You can give it an engineering prompt, and it gives you a certain output.
The next day, you give the same engineering prompt – and the output is totally different.
The other limitation you have with these off-the-shelf AI tool generators is the incompatibility for one custom application to talk to another application.
On AI trust, risk management, and governance safety nets
Moreso than any workplace tech we’ve seen, effective AI is about trust. If Umergi can’t solve the data trust problem with a particular tool, he won’t move forward:
When the conversation started about AI, it didn’t begin with AI. It actually began with trust: where is the trust layer in this? Could we trust the AI tools to keep our our data [private]? Some of these healthcare records also end up in open source, or something like ChatGPT, teaching somebody else’s Large Language Model.
What about Zia? Umergi says that Zoho’s AI architecture makes the trust difference, with the “safety net” of a governed platform.
Zia is drawing from existing data that sits in CRM, in Zoho recruit, in Zor creator, and it brings it all together. So that gives us the confidence our data remains within a controlled, governed ecosystem. And that’s been the biggest kind of safety net for us.
Now you have a trusted AI platform – but you still have to build. What is Umergi’s team working on?
We’ve identified two solid use cases. One is having an augmented reality version of a 3D body, where the healthcare worker and interact with the body, and say yes to the applied medication or treatment.
Patient care plan summaries are on deck also:
When looking at a care plan for a patient, the care plan could be 30/40 pages long. Now, when you have a care worker delivering the care service to multiple patients, they don’t really have that time to understand the entire care plan.
For this scenario, Newscross is also looking into speech-to-text conversion. The Creator application will transcribe sessions into text, and convert it into a care plan. Another Zia use case? Organizing a noisy stream of to-dos:
The other use case we have, which is fundamentally crucial for us in the near future, is Zia prompting the healthcare worker what the next task is they need to do for the client they’re caring for – as opposed to having a to-do list of about 30 activities they need to do in the morning.
But as Umergi points out, none of these use cases would be viable without built-in compliance:
Zoho has done all of the groundwork in terms of ISO accreditations. All of the governance and the cybersecurity. So when we wanted to connect our Zoho creator appliation with the NHS database, the health care worker can go to a patient record in the Creator application at the click of a button, retrieve the medical record for that patient in real-time, and then also run it back to the NHS database to say what care services they’ve delivered.
Umergi’s team is also working on an API integration to the DMD (Direct Medication Database), to allow real-time updates to medication releases and changes into Zoho Creator. That’s why comparing Zoho to out-of-the-box LLMs doesn’t make sense:
You can’t compare Zia to other AI tools because the use cases are so different. We’ve drawn a concrete line in the sand about things we cannot consider with other AI tools. AI tools have a tendency to hallucinate which will reduce the confidence in the business in these tools.
But even within Zoho’s platform, Umergi is more aggressive with AI in some areas, and more cautious about deployment in others. What is the difference?
How we measure success with AI, especially with Zia, is first: we identify where the friction in the business is. You know, where is all of the time going with manual efforts.
That’s where the risk filter comes in:
When it comes to medical records or sensitive information, we don’t want Zia to make recommendations in terms of clinical governance. We want to aid that clinical decision with the information in Zia. I’m of a strong opinion that you can’t use AI tools to replace professional opinions of GPs, doctors or clinical nurses, but you can certainly aid them to get information quicker and faster.
AI projects with Newcross will roll out over time, but real-time visibility is now:
The real benefit has been providing the business with real-time information at the right time. Whether that’s senior leaders in the business who want to see financial performance, or the healthcare worker who needs real- time information in terms of the care they delivering to the patients.
My take – customers should own their AI narratives, and keep their domain experts!
The best customer AI stories I’ve documented involve a genuine ownership of the internal AI narrative – and a culture that embraces an iterative process of data improvement and AI app/agent building. But for most enterprises, a trusted vendor factors in heavily.
Reviewing the Newcross Healthcare story, you can see why: a compliant, well-governed AI apps platform is not something easily conjured up internally. But if you get it right, that platform will give you a much more sturdy and defensible runway than you can get from freelancing with frontier models. That’s why Zoho has built its own models, in a variety of sizes, to fit for individual customers and use cases. Of course, Zoho also provides customer access to frontier models, but within a governed framework, where the customer’s data does not become part of the LLM’s training fodder.
I have to address one chip on my shoulder: Umergi’s insistence on expert human review when it comes to clinical decisions. There is a persistent fiction that as LLMs become more sophisticated, the need for human domain experts is eliminated. I see contrary evidence again and again. LLMs do not understand causality, though they can, at times, assist in the generation (and utilization) of causal graphs.
But again, a domain expert is almost always required to iron out the flaws in the causal illustrations, including hidden confounders. Why does this matter? Because healthcare is all about determining causal factors for health conditions, and successfully treating them. Getting those decisions right is high stakes indeed. That goes for care management at scale also. It is encouraging to hear from a healthcare AI leader who is pursuing the strengths of this technology, but with a careful eye towards AI overreach.
That frees up Newcross Healthcare to focus on the opportunities/pain points they need to address – and utilize AI where it best fits. As I said to Umergi as we closed out our interview:
My big takeaway from talking with you today: it seems like AI stems from an overall pursuit of excellence as an organization – and that ‘trust’ seems to be the confounding word in the middle of all of this , that we need to somehow conquer.
Let’s see how this unfolds. Umergi has indicated he wants to take this excellent conversation further – something else to look forward to.
