Private GPT: Stop Sharing Company Data

Book an AI Workshop Today - https://www.fortunadata.com/ai-workshop/

In this episode of Smarter Strategic Thinking, Ray Quattromini is joined by Nipa Soni, AI and HPC Enablement Lead at FSAS Technologies, to explore their fully on-premises private GPT solution designed from the ground up for organisations that cannot afford to compromise on data sovereignty, security, or regulatory compliance.

FSAS Technologies is a Fujitsu spinoff operating across the UK and European market, aligned with the EU AI Regulatory Act and its emphasis on data control, governance, and privacy. Their private GPT runs entirely on-premises your data never leaves your environment, there's no internet exposure, and there are no per-query token costs. Licensing is user-based: unlimited usage for a fixed cost.

What you'll hear in this episode:

Nipa explains why public cloud AI models however capable introduce data sovereignty risk that regulated organisations and IP-sensitive businesses simply can't accept. They walks through the architecture that prevents data leakage, the department-level segregation model that keeps finance, HR, sales, and R&D siloed from one another, and how the system integrates with existing platforms including IBM Maximo via API and MCP.

The conversation covers the end-to-end customer journey: from an initial enquiry through to the £2,500 discovery workshop which includes AI consultants, data scientists, and mathematicians working alongside the customer's own stakeholders to identify one or two specific, high-impact use cases. The workshop takes roughly half a day to a day; a detailed findings report follows within a week and a half to two weeks. From there, POC development proceeds at no additional cost until the customer is satisfied before any further commercial commitment is required.

A UK manufacturer use case is discussed in detail: they deployed private GPT to address high error rates in part dispatch, training the model on their own proprietary data and seeing a significant reduction in errors and therefore returns within approximately three months.

The episode also addresses AI ethics and governance, including FSAS's programme delivered to first-year college students exploring the societal implications of AI decision-making, and how those same ethics principles are embedded at the platform level making it well-suited to education environments and regulated industries.

This episode is relevant to: CIOs and Heads of Infrastructure evaluating on-premises AI, MSPs and resellers considering private GPT for clients, organisations in regulated sectors (legal, public sector, manufacturing, education), and any business concerned about what happens to their data when staff use public AI tools.

Smarter Strategic Thinking is produced by Fortuna Data a UK-based IT storage reseller and consultancy specialising in data sovereignty, LTO tape archiving, ransomware resilience, cloud strategy, and AI-ready infrastructure.

Transcript

00:00:00:00 - 00:00:03:04
Unknown
I spoke to have a what I'm doing at the weekend. So I

00:00:03:08 - 00:00:09:00
Unknown
Let me show some attention. Dig it out. Oh, lovely. Sounds amazing. I have to do all this.

00:00:09:03 - 00:00:10:14
Unknown
right. We ready then, Millie

00:00:10:26 - 00:00:12:17
Unknown
Action.

00:00:12:21 - 00:00:22:09
Unknown
Hi, I'm Ray Quattromini from Fortuna Data. Welcome to another episode of Smarter Strategic Thinking. Today we have Nipa Soni from FSas Technologies.

00:00:22:13 - 00:00:23:38
Unknown
Hi. Good to see you. Thank you.

00:00:23:42 - 00:00:26:35
Unknown
Nipa, what is your role within FSas Technologies?

00:00:26:35 - 00:00:31:00
Unknown
I'm responsible for AI and HPC enablement across FSas Technologies in the UK.

00:00:31:00 - 00:00:31:25
Unknown
working with

00:00:31:31 - 00:00:32:14
Unknown
also staff,

00:00:32:14 - 00:00:41:18
Unknown
but also more importantly being in front of customers, understanding their needs, understanding their business challenges and bringing that to life, whether that's within AI, whether that's within HPC,

00:00:41:22 - 00:00:45:31
Unknown
trying to understand and really feel customer pain points

00:00:45:36 - 00:00:49:18
Unknown
and bringing solutions to the table that's going to help them

00:00:49:18 - 00:00:52:40
Unknown
work better, smarter, more efficiently and help deliver their ROI.

00:00:52:42 - 00:00:53:24
Unknown
That's

00:00:53:28 - 00:00:54:40
Unknown
in a nutshell, what I do.

00:00:54:44 - 00:00:57:49
Unknown
how long have you been in it and where did you start?

00:00:58:02 - 00:01:13:11
Unknown
worked my way through the industry. I worked at several distributors. Manufacturers always worked for the large majority of my career. I worked with reseller partners and then customers, and then of recent to the last 7 or 8 years. It's more with end customers directly.

00:01:13:16 - 00:01:15:06
Unknown
private GPT.

00:01:15:10 - 00:01:20:22
Unknown
FSas built an on premise private GPT solution. Why why did they do that

00:01:20:31 - 00:01:42:34
Unknown
So number of different factors came into this. So one there was definite need in the market for something that was private secure. And that is in line with government regulations and rules. There was a lot of organizations out there which like a GPT solution, which are in the cloud, accessible to everybody.

00:01:42:38 - 00:01:54:24
Unknown
But the question is, how secure really is that your data, any organization's data, is absolutely key to what they do without the data. If you effectively don't have a business.

00:01:54:28 - 00:01:59:26
Unknown
why would you not want to protect that every single layer? And especially with AI

00:01:59:26 - 00:02:00:46
Unknown
protecting your data,

00:02:00:46 - 00:02:03:38
Unknown
having the control and having that security element

00:02:03:38 - 00:02:05:44
Unknown
with your data and with your AI solutions

00:02:05:44 - 00:02:06:36
Unknown
is really, really key.

00:02:06:36 - 00:02:15:01
Unknown
how does it differ from an on prem solution to a Claude and open AI, a deep sea? What what makes it different?

00:02:15:01 - 00:02:21:38
Unknown
so a number of different things. So you know, we can talk about all the tokens and versus no tokens. But outside of that

00:02:21:43 - 00:02:24:27
Unknown
this is a fully on prem solution.

00:02:24:31 - 00:02:31:11
Unknown
All the data, all the promise, all the information that's provided stays within that customer's environment,

00:02:31:16 - 00:02:33:47
Unknown
giving you that full control of security

00:02:33:47 - 00:02:35:01
Unknown
and flexibility

00:02:35:01 - 00:02:36:11
Unknown
not having to

00:02:36:21 - 00:02:37:37
Unknown
for things that you don't need.

00:02:37:37 - 00:02:41:16
Unknown
Customizing a solution that is tailored to your business needs.

00:02:41:20 - 00:02:42:15
Unknown
mentioned tokens.

00:02:42:15 - 00:02:42:35
Unknown
Yeah.

00:02:42:39 - 00:02:48:44
Unknown
every AI command that goes to a cloud, for example, you pay for a bunch of tokens. But

00:02:48:44 - 00:02:52:44
Unknown
are you saying that your private GPT solution, you don't need that

00:02:52:47 - 00:02:55:09
Unknown
I'm absolutely saying no, we do not need tokens.

00:02:55:09 - 00:03:18:35
Unknown
We don't go down the token. We. You have a license for a number of users and that is it. There's no tokens. All you can use it to your heart's content as much or as little as you like. And we appreciate that certain departments within an organization will use a private GPT solution a lot more than others, so giving them the opportunity to use it freely without having to be restricted with tokens and the cost implication with that

00:03:18:41 - 00:03:35:34
Unknown
you explain what it provides to a customer, what a typical customer would look like. And in particular, you've mentioned about privacy. How is that secured? Is it that they put it in a room and put a Faraday cage around

00:03:35:38 - 00:03:48:49
Unknown
security comes at the heart and the foundation of everything we do. Everything on prem data sovereignty. It's no longer just a buzzword. It is is the foundation of what customers need when they're looking at AI solutions.

00:03:49:11 - 00:03:56:39
Unknown
And that's what we build from scratch. So everything we do has got data sovereignty, security, governance and protection in mind. Everything we do is on prem.

00:03:56:45 - 00:04:04:33
Unknown
What it gives customers and what it can do is imagine you have very document heavy processes

00:04:04:43 - 00:04:14:29
Unknown
those can be automated with private GPT, imagine being able to talk to your own data, not having access to the outside world where hallucinations may ramp up because

00:04:14:33 - 00:04:15:15
Unknown
potentially

00:04:15:18 - 00:04:18:14
Unknown
wrong or false or incorrect information or outdated information.

00:04:18:14 - 00:04:22:10
Unknown
This private GPT will talk to your data that you implement into the system,

00:04:22:10 - 00:04:24:44
Unknown
and you're able to have a full on conversation with your data.

00:04:24:48 - 00:04:28:23
Unknown
automating that process, making things a lot easier, streamlining

00:04:28:27 - 00:04:29:29
Unknown
document heavy

00:04:29:33 - 00:04:34:28
Unknown
mundane tasks as well. And we all have them, you know, whatever job throw your way. And we all have mundane tasks.

00:04:34:28 - 00:04:37:49
Unknown
Take that element away. Let's bring some fun back into into work.

00:04:37:49 - 00:04:39:31
Unknown
We can actually absolutely help with that.

00:04:39:39 - 00:04:41:12
Unknown
it's your own

00:04:41:19 - 00:04:44:33
Unknown
proprietary information that you feed in.

00:04:44:37 - 00:04:45:09
Unknown
Correct.

00:04:45:09 - 00:04:51:17
Unknown
So you know you're not getting any erroneous sources of data, an injection into your code or whatever you're trying to do.

00:04:51:17 - 00:04:51:49
Unknown
Exactly.

00:04:52:08 - 00:04:54:28
Unknown
So this solution is completely customizable.

00:04:54:35 - 00:04:55:38
Unknown
It's not off the shelf.

00:04:55:43 - 00:05:01:02
Unknown
what does a typical build and delivery time look like? And who owns the IP at

00:05:01:02 - 00:05:03:01
Unknown
100% is customizable. So we start

00:05:03:10 - 00:05:03:49
Unknown
off with

00:05:04:06 - 00:05:07:03
Unknown
almost out of the box type solution

00:05:07:03 - 00:05:09:33
Unknown
where you've got the basics of a private GPT.

00:05:09:37 - 00:05:10:37
Unknown
from there we

00:05:10:40 - 00:05:14:29
Unknown
tweak tailor customize to whatever the customer needs.

00:05:14:33 - 00:05:18:20
Unknown
start off with a workshop. And the reason for this workshop

00:05:18:28 - 00:05:25:32
Unknown
is we are then able to define and figure out a specific use case for those customers. So how it works is

00:05:25:36 - 00:05:27:41
Unknown
there are stakeholders from within the business

00:05:28:02 - 00:05:32:26
Unknown
that we set up with, with representations from FSas Technologies. So whether that's a

00:05:32:26 - 00:05:38:39
Unknown
consultant, an AI expert, a mathematician, a scientist, scientist, whatever it might be that you need on that day.

00:05:38:44 - 00:05:41:41
Unknown
From there, we let customers

00:05:41:45 - 00:05:48:15
Unknown
effectively talk to their own data. We give them access to a private GPT model where we've already pre uploaded their own data.

00:05:48:25 - 00:05:49:43
Unknown
And they get to see it hands on.

00:05:49:48 - 00:05:56:38
Unknown
And that's when we bring it to life in this workshop. Call it an ideation. Whatever we need to. That's where we bring this to life.

00:05:56:38 - 00:06:00:27
Unknown
And they can say actually this would help me in so many different areas.

00:06:00:31 - 00:06:18:01
Unknown
then from there we talk about, okay, well, what are the specific challenges within each of these departments? It typically if you have 15 stakeholders within a room, you'll probably get about 30 to 60 different challenges that they come up with. But I need it for this.

00:06:18:01 - 00:06:39:45
Unknown
I need it for this. And of course, realistically, you know, you're not going to do 50 or 60 challenges. So we figure out 1 or 2, 1 or 2 that are very, very specific. And that may actually benefit several different departments. And then we go from there. And that that process typically takes a week or two. So the ideation is a day workshop.

00:06:39:49 - 00:06:49:23
Unknown
And then the follow up after that is about a week, which gives us the opportunity to review and come back to you with, effectively a report to say, this is what we found out, this is where we're going to go next.

00:06:49:34 - 00:07:00:39
Unknown
the customer owns the instance in their environment. We own the original IP but the customer does with it whatever they wish to do.

00:07:00:43 - 00:07:02:09
Unknown
the reason for that is

00:07:02:13 - 00:07:06:13
Unknown
one, we want to make sure that the customer has full control over their data, over their environment.

00:07:06:13 - 00:07:12:08
Unknown
Everything stays on prem, stays with them. We don't we don't access it. The reason we own the IP is because of the,

00:07:12:08 - 00:07:21:22
Unknown
backend side of things. And what I mean by that is the support pack of maintenance and via the support pack is how the customers get the additional features. So

00:07:21:22 - 00:07:23:32
Unknown
as we develop our private GPT,

00:07:23:32 - 00:07:26:47
Unknown
the customer that's already got what in place in situ

00:07:26:47 - 00:07:31:33
Unknown
will then automatically receive those additional feature updates and upgrades.

00:07:31:40 - 00:07:32:46
Unknown
And that's that's the reason,

00:07:33:02 - 00:07:36:14
Unknown
a workshop starting at 2500 pounds,

00:07:36:18 - 00:07:38:09
Unknown
for an issue workshop.

00:07:38:13 - 00:07:39:47
Unknown
What does that deliver as an app?

00:07:40:01 - 00:07:56:39
Unknown
And what does it what how is it a controlled environment? And what what does the customer need from a commercial engagement process? And finally, if somebody says, I want to go on this, how quickly can that service be delivered or that workshop start?

00:07:56:43 - 00:07:59:02
Unknown
initial workshop is 2500 pounds.

00:07:59:02 - 00:08:14:26
Unknown
It's usually either half a day or a day, depending on the number of stakeholders you bring to that workshop. From the customer's point of view, we will then engage, either data scientists, mathematicians, AI consultants and experts to that workshop. Typically

00:08:14:30 - 00:08:24:38
Unknown
within a week and a half to two weeks. The customer then receive a full on report as to our findings and our suggestions, and then potential next steps.

00:08:24:46 - 00:08:28:31
Unknown
That's all included in that 2500 pounds initial upfront cost.

00:08:28:37 - 00:08:34:45
Unknown
And then from there, if the customer wanted to go down a policy route and investigate a particular use case further,

00:08:35:07 - 00:08:43:42
Unknown
there's no additional cost to that, right the way up until we get to policy. And it's just after the POC. If the customer is absolutely satisfied that this POC does exactly what we've

00:08:43:42 - 00:08:44:34
Unknown
planned out for its Do

00:08:44:34 - 00:08:45:09
Unknown
meets all their

00:08:45:14 - 00:08:46:21
Unknown
objectives and goals,

00:08:46:26 - 00:08:48:02
Unknown
then we move forward with a quotation.

00:08:48:02 - 00:08:48:34
Unknown
I think with

00:08:48:40 - 00:08:49:24
Unknown
AI

00:08:49:24 - 00:08:51:02
Unknown
there's a lot of options out there

00:08:51:02 - 00:08:52:10
Unknown
doing it this way.

00:08:52:10 - 00:08:54:39
Unknown
One, it's commitment from both organizations,

00:08:54:47 - 00:08:55:31
Unknown
but also

00:08:55:31 - 00:08:57:26
Unknown
it's almost like proof in the pudding. So, you know,

00:08:57:26 - 00:08:59:36
Unknown
we have the expertise. We have the knowledge.

00:08:59:48 - 00:09:04:25
Unknown
Let us show you first our capabilities. And then you decide if you want to go forward.

00:09:04:28 - 00:09:06:24
Unknown
And it's just a small cost at the beginning.

00:09:06:34 - 00:09:11:45
Unknown
AI models consultation can ramp up very, very quickly. That's not what we were trying to achieve here.

00:09:11:49 - 00:09:19:24
Unknown
So if somebody is interested in the workshop, what is the time frame to deliver an inquiry to actually engaging with that workshop?

00:09:19:33 - 00:09:22:35
Unknown
So initially we'd have a conversation with the customer.

00:09:22:39 - 00:09:45:07
Unknown
We understand what they're trying to do and if and it's okay if they don't know because a lot of customers want I know they need a I'm not quite sure what to do with it. That's where we come in and we can help. We have initial conversation to gauge your basic ideas, your basic requirements. From there, I probably need about a fortnight to gather all our experts together, make sure we have the right people in the room with with the customer.

00:09:45:11 - 00:09:51:05
Unknown
So roughly about two weeks, we think, from phone call to workshop is probably about right.

00:09:51:13 - 00:09:52:46
Unknown
our concern is data leakage.

00:09:53:03 - 00:09:59:17
Unknown
how do you ensure customer data doesn't leave that environment and isn't fit into a public cloud model?

00:09:59:25 - 00:10:04:38
Unknown
that's a really good question. And data leakage is a concern to many organizations that I speak to.

00:10:04:42 - 00:10:06:13
Unknown
The architecture

00:10:06:13 - 00:10:09:32
Unknown
is what prevents the data leak. We have right from the very beginning.

00:10:09:36 - 00:10:11:03
Unknown
Everything we design

00:10:11:03 - 00:10:11:28
Unknown
has got

00:10:11:28 - 00:10:13:43
Unknown
data sovereignty in mind.

00:10:13:46 - 00:10:15:23
Unknown
That's the fundamental thing that we do.

00:10:15:23 - 00:10:16:14
Unknown
The data

00:10:16:19 - 00:10:18:39
Unknown
that is generated, uploaded

00:10:18:46 - 00:10:20:10
Unknown
into our private GPT model

00:10:20:13 - 00:10:21:28
Unknown
stays with the customer.

00:10:21:34 - 00:10:22:29
Unknown
It has no

00:10:22:32 - 00:10:24:15
Unknown
outside access to the internet

00:10:24:19 - 00:10:25:16
Unknown
and also as well,

00:10:25:16 - 00:10:26:47
Unknown
when we look further around,

00:10:27:05 - 00:10:30:28
Unknown
let's say a customer purchase a solution, they're happy with it

00:10:30:38 - 00:10:35:17
Unknown
when we have our updates because I changes very, very rapidly.

00:10:35:26 - 00:10:39:25
Unknown
let's say a customer purchased a private GPT 1.5 version

00:10:39:33 - 00:10:43:34
Unknown
and within three months of implementation, there is a 1.6 version.

00:10:43:43 - 00:10:59:48
Unknown
The customer will receive all of those features at no additional cost and no additional access from us is just done via a link they can download and then they upload it. So again, even with the new features, there are no data leakages.

00:11:00:02 - 00:11:01:34
Unknown
The customer has full control.

00:11:01:44 - 00:11:11:49
Unknown
if a customer, for example, is scanning PDFs and they're indexing PDFs and doing everything else, and then six months down the line, somebody comes along and they want to do x, y, z.

00:11:12:03 - 00:11:13:34
Unknown
What what's the process

00:11:13:37 - 00:11:14:24
Unknown
for that?

00:11:14:28 - 00:11:19:44
Unknown
Please let us know. Is the is the first instance. Let's have that conversation. You would be appointed

00:11:19:44 - 00:11:23:05
Unknown
certain members of our AI team. Anyway to your project.

00:11:23:05 - 00:11:25:10
Unknown
you'd already have the contact. So get in touch with us.

00:11:25:14 - 00:11:33:20
Unknown
Let us know what you're looking for. It may be that it's already on the existing roadmap, and we can absolutely share that with you, whether it's 12, 18 or 24 months roadmap.

00:11:33:20 - 00:11:36:18
Unknown
And then she can appreciate these things move very, very quickly. And I

00:11:36:22 - 00:11:42:18
Unknown
however, if it's not on our roadmap, if it's something that is specific to a customer that they absolutely want to have,

00:11:42:22 - 00:11:49:38
Unknown
again, it's not off the table. Let's have a conversation. If we if we can build and tailor something to you, we absolutely will. Look at that

00:11:49:42 - 00:12:05:11
Unknown
Nipa, how does your solution integrate with existing platforms specifically, is there a link with Maximo, IBM's asset management platform, or or is it a completely standalone solution, or has it got any plugins or APIs that a user can use?

00:12:05:11 - 00:12:09:45
Unknown
So it definitely has API options and also NCP options as well.

00:12:10:05 - 00:12:31:19
Unknown
So if customer wants it to plug into certain applications, we can look at that in terms of Maximo specifically. This is where the workshop comes in. This is where we would work with the customer to investigate exactly what tools they use, and then work around a solution for that and integrate that at POC level to make sure the testing is done.

00:12:31:19 - 00:12:33:44
Unknown
And so working smoothly and efficiently.

00:12:33:48 - 00:12:45:06
Unknown
Okay. We have customers that aren't technical. How does the interface and on guide administration work for non-technical users? What is their experience?

00:12:45:06 - 00:12:47:16
Unknown
What would they say?

00:12:47:20 - 00:12:53:23
Unknown
the interface that we use is super easy, I love it, I find it so simple to use,

00:12:53:23 - 00:12:55:10
Unknown
imagine talking to a chat bot.

00:12:55:14 - 00:13:03:48
Unknown
Yeah, it's exactly the same type of feel to it. It's really easy. And also with this administration portal, which liaise obviously with the technical side of the business,

00:13:04:05 - 00:13:20:21
Unknown
they have the opportunity to pre-load certain prompts for certain departments. So, you know, if you're in H.R. You might use certain prompts versus somebody in finance or sales or marketing. Those prompts can be pre uploaded to make life easier for specific departments.

00:13:20:25 - 00:13:32:13
Unknown
Or they can create their own. So it's it's very very intuitive. It's very simple to use. You can do a lot of the pre-work ahead of time if needed. And also there's training involved okay. So

00:13:32:17 - 00:13:34:38
Unknown
partnering with the likes of FSas Technologies,

00:13:34:43 - 00:13:44:46
Unknown
we do offer training to our customers. So whether that's technical training or whether that's actually the use of the private GPT itself to non-technical users, that training and support is available.

00:13:45:07 - 00:13:51:26
Unknown
So is it a browser window that they would get or is it or can it be built into an application.

00:13:51:29 - 00:13:57:16
Unknown
you see the link you logging save into your favorites. You log into your private GPT and it's just like talking to the internet

00:13:57:16 - 00:13:58:24
Unknown
but with your own data.

00:13:58:28 - 00:14:06:37
Unknown
Nipa, what does the API and development layer look like from a customer wants to integrate and develop their own applications moving forward.

00:14:06:37 - 00:14:13:31
Unknown
private GPT is customizable to a certain level. It has the ability to add APIs and also

00:14:13:31 - 00:14:14:18
Unknown
MCP.

00:14:14:18 - 00:14:17:47
Unknown
That said, if the customer wants to further develop

00:14:18:07 - 00:14:26:26
Unknown
it then almost moves into a slightly different solution. That's when we start looking at the enterprise. Well, that's when we start looking into the likes of using

00:14:26:26 - 00:14:30:10
Unknown
to connect or cohere, as opposed to something like Mistral.

00:14:30:10 - 00:14:36:21
Unknown
So it is available to do. It just depends on what the customer is trying to achieve and what that kind of

00:14:36:21 - 00:14:38:43
Unknown
1 to 3 year plan looks like for the customer.

00:14:38:48 - 00:14:49:16
Unknown
And I'm going to bring it back to that workshop again. This is why we have that workshop is to really understand the customer's goals, their business ideas, their desired outcomes.

00:14:49:20 - 00:15:10:20
Unknown
And the ROI is going to deliver today and also in the future, so we can start working with the customer on that journey to where private GPT may well be the answer today for certain use cases. We can also, alongside that, work with customers on further developing additional AI solutions for them

00:15:10:31 - 00:15:20:05
Unknown
what are the ethics and governance controls built into the platform especially relevant if we're developing this in an education or a regulated industry?

00:15:20:05 - 00:15:28:46
Unknown
So private GPT is built with ethics and governance from the ground up. It's as simple as that. Because it runs fully on prem.

00:15:29:00 - 00:15:59:05
Unknown
It remains under the control of the customer, ensuring privacy, security compliant, unregulated regulatory standards, especially in education environments and something else that we do. And this this one's quite close to me because I as I mentioned, I work in it in enablement, bringing things to life. And making things happen for customers. Ethics within education was such a strong topic.

00:15:59:09 - 00:16:06:44
Unknown
We had to do something with that, and we did so as an organization, we believe in giving back into the community. We believe in social values

00:16:07:03 - 00:16:21:13
Unknown
and that social impact that we as a business, as a technology company can have to our local and wider communities. And one of the things that I looked at was delivering the ethics of AI into education.

00:16:21:17 - 00:16:33:30
Unknown
So yes, we can do that from a technical standpoint. And private GPT absolutely takes those boxes, but also from a student teacher professor standpoint as well.

00:16:33:41 - 00:16:39:00
Unknown
I could do whatever we want it to do as human beings. The question is, should we let it?

00:16:39:04 - 00:16:46:42
Unknown
What are the ramifications and the impact of that? And what are the societal impacts of of saying yes to an AI doing certain things?

00:16:46:46 - 00:16:51:39
Unknown
So we created what we done. We created a program which is aimed at first year college students,

00:16:51:45 - 00:17:03:36
Unknown
we take them on a journey around the ethics of AI. What's good AI, what's bad? I should we do something? Should we not do something? For example, we looked at okay, should I be used to mock exam papers?

00:17:03:39 - 00:17:05:21
Unknown
Imagine the uproar.

00:17:05:25 - 00:17:10:15
Unknown
And there was good points and bad points to it. And in some instances it can be a very, very good tool to use.

00:17:10:21 - 00:17:12:22
Unknown
thing is, it always needs that human in the loop.

00:17:12:22 - 00:17:28:35
Unknown
but it's educating people the future of tomorrow about the ethics of AI. So not only do we embed it and build it at the core from within our platforms, we actually deliver that message in different ways to give back to that societal impact.

00:17:28:39 - 00:17:42:13
Unknown
how does private GPT silo of different departments high charge finance, finance, the marketing to sales, how does it actually create that segregation so those other departments can't tread on other people's toes.

00:17:42:17 - 00:17:47:48
Unknown
Important right. Yeah. So that's all done at the

00:17:47:48 - 00:17:48:43
Unknown
technical standpoint.

00:17:48:43 - 00:17:52:28
Unknown
So when when we set up a project that's in the customer's environment,

00:17:52:36 - 00:18:03:32
Unknown
can set the right parameters, the right rules for every single department. And we exclude the all. So finance can't see what is doing. Sales can't see what was doing.

00:18:03:32 - 00:18:09:32
Unknown
You can segregate each department. You can take it even further to segregate groups of people.

00:18:09:37 - 00:18:10:22
Unknown
Management,

00:18:10:26 - 00:18:22:26
Unknown
upper management, high management, CIOs, etc. you can segway to who get access to what information so you can. Absolutely. And that's all customizable. All of that is customizable at the customer level. So whatever they want

00:18:22:38 - 00:18:29:21
Unknown
and however they want to I guess top that data and who the access to that data is absolutely doable.

00:18:29:25 - 00:18:31:25
Unknown
Yeah. Because some of them might be in R&D,

00:18:31:31 - 00:18:32:17
Unknown
correct.

00:18:32:17 - 00:18:32:34
Unknown
Yeah.

00:18:32:41 - 00:18:43:27
Unknown
intellectual property, you know, all these sorts of things. And it's for me that's quite important that, you know, whilst we've got a private chat bot, we, we saw a lot of different departments.

00:18:43:27 - 00:18:45:17
Unknown
100%. Yeah. It's important.

00:18:45:21 - 00:18:48:13
Unknown
Nipa we've mentioned a lot about private GPT.

00:18:48:13 - 00:18:52:20
Unknown
What are the use cases for the products from a customer's point of view?

00:18:52:20 - 00:18:57:17
Unknown
recent use case, which was a manufacturing organization in the UK.

00:18:57:21 - 00:19:09:16
Unknown
They wanted to do something very, very specific. And that was they wanted to reduce the number of errors when sending out parts. So they wanted to make sure that they send in the right part

00:19:09:24 - 00:19:17:27
Unknown
for the right product. Now this manufacturer supplies various parts of various other third party manufacturers.

00:19:17:34 - 00:19:18:35
Unknown
what they were seeing was a

00:19:18:35 - 00:19:20:13
Unknown
high error rate

00:19:20:22 - 00:19:23:08
Unknown
in the number of parts being sent out for send product,

00:19:23:08 - 00:19:24:14
Unknown
so they wanted to reduce the

00:19:24:18 - 00:19:26:33
Unknown
they put their data in a private GPT model.

00:19:26:42 - 00:19:28:47
Unknown
They trained it, they set the parameters around it.

00:19:29:00 - 00:19:34:41
Unknown
And then within about three months they saw a huge reduction in the number of errors.

00:19:34:46 - 00:19:44:03
Unknown
So not just a number of errors therefore then resulting in the number of less returns coming back. So that was a really, really good strong use case for profitability.

00:19:44:03 - 00:19:47:45
Unknown
In Europe, however, waste in a lot of organizations,

00:19:48:03 - 00:20:00:47
Unknown
especially around public sector, around law firms in particular, where they've got a lot of document heavy processes, they are utilizing private GPT within those spaces.

00:20:01:01 - 00:20:08:26
Unknown
And I think a lot of the time, if you look at Europe, they are much more in line with the AI, EU Regulatory Act,

00:20:08:33 - 00:20:25:42
Unknown
which is all about data sovereignty. And yeah, that's that's mainly the reason that we're pushing the data. That's the reason that data sovereignty is key to us as well. You know, we are Japanese company, but the UK work is part of the European team,

00:20:26:03 - 00:20:26:28
Unknown
Yes.

00:20:26:32 - 00:20:30:18
Unknown
and we are much more aligned with European versus say,

00:20:30:22 - 00:20:31:20
Unknown
Yeah, like U.S acts.

00:20:31:30 - 00:20:37:28
Unknown
a customer wants to go on a journey and, and look at private GPT, what are the next steps?

00:20:37:28 - 00:20:45:29
Unknown
If you've heard something today that sparked your interest, let's have a conversation. Get in touch is have a phone call. Let's have a teams call, whatever it might be.

00:20:45:29 - 00:20:58:24
Unknown
Let's meet up. Let's understand your concerns. Let's and let's talk about what you're trying to achieve. And we take it from there. It's really, really simple. We're trying to keep this as simple as possible. AI's very, very complicated and can get very complex.

00:20:58:24 - 00:21:01:33
Unknown
Talk to us, pick up the phone, that it's that simple,

00:21:01:37 - 00:21:03:27
Unknown
now. It's been a pleasure. Thank you very much.

00:21:03:37 - 00:21:11:49
Unknown
If you liked the video today, give us a thumbs up. If you're interested in talking about the FSas private GPT solution, get in touch.

00:21:12:03 - 00:21:18:46
Unknown
Or you got any questions for our next podcast or you want to ask Nipa any questions then? Thank you for your time. And give us a thumbs up.