Nimbus: How a UK-Based Start-up Is Building Enterprise AI for the Next Generation of Business

There is a disconnect in organizations now that are deploying AI. That is meaning that people are moving in a lot of different directions at quite a lot, you know, accelerated speeds, but not in a coherent direction because they're all using different tools, different contexts and what is needed, which is what we're aiming to achieve with Nimbus is fleet intelligence. So you can think like we want to ultimately turn organization into a hive mind where every solution that is or every part of AI deployment within an organization knows what the other part is doing.

Hello and welcome to the marketing and AI podcast, the Map. I'm your host, Hal Kimber. I'm thrilled to see Nick as always. Nick, how are you?

Very well, thank you. Yeah, we've had the, we've had time together this week, haven't we? So it's, we've had quite a week. Nick and I were in Switzerland earlier this week running the second of two AI workshops with the E commerce, marketing, HR teams of Helen of Troy.

A great session, wasn't it? It was. Lots of practical ideas started to germinate at the end, which was, which is great to see. Lovely people.

We'll see, we'll see if we can do any good for them. Absolutely. And you know, if we could be of interest for your organization, do get in touch with us via LinkedIn now, two episodes ago we spoke to Tim Flagg, CEO of UK AI on how Britain can lead. I think it's very fair to say that most British companies are now doing AI to some extent.

It often means they started with tools like Copilot, Very sensible and often valuable first step. But today's question is how do UK businesses move from these first steps with AI to true enterprise level adoption? And what does enterprise level adoption even look like and mean? Which means I'm thrilled to welcome our guests to help us explore this topic.

James Campbell and Zach Radbone from Nimbus, a UK founded enterprise AI platform with international backing that is focused very much on making enterprise level AI adoption achievable. But complex organizations are organizations large and small. James and Zach, very warm welcome. Thank you.

Yeah, thanks for having us guys. Oh, it's great to have you here. Tell us a bit about your journeys to Nimbus to date and your backgrounds before we really dig into the solution. James Campbell, CEO and co founder of Nimbus I come from an extensive engineering in the kind of consumer and product development space background.

So I studied engineering at university and then I went from straight from there to work for McLaren Automotive and had the Luxury of spending most of my time working on the McLaren P1, which was a marvelous opportunity at the age of 22 and really started to flesh out my integration into the various different engineering disciplines from there. So looking right from aerodynamics through to powertrain efficiency and getting really into the kind of simulation of the powertrain from there I went to work for LEDC at the time was London Taxi company and helped them to work on the legacy black cab vehicles, the London Black Cab, and then looked at helping to develop the concept of the electric version of the black car which is now rolling around the streets of London.

From there I went and worked on the Dyson Automotive project, worked there for four years, helped build out a team of anything from thermophysiologists through to thermal management experts in kind of thermal system adoption. And when unfortunately that project was cancelled, I then headed up the engineering team, actually scoped out the whole engineering team for a local to Gloucestershire company called Ruroc that make motorcycle helmets, extreme sports helmets and embedded electronics and helped them to launch another eight products to market, take them from what was already quite a good annual revenue up to almost tripling that annual revenue and really kind of fleshed out my understanding of the broader product development space, how to then integrate that with the marketing teams and the executive teams from there.

I then founded my first tech business which was looking at teleoperation technology for ride hail. So this is really remote driving vehicles as a taxi service. We raised, we raised kind of the seed funding we needed to develop the technology and actually get a small kind of controlled demo trial running. And unfortunately through kind of a number of different circumstances we couldn't raise the following rounds.

So I then raised, I then raised some small funds to start a company called grow, which is software company that develops visualization tools for E commerce and interior designers. So 3D visualizations, augmented reality and virtual reality. And then through a relatively serendipitous set of circumstances, I was in the same co working space as Zachary. We got introduced to Jeff, who's the CTO of Nimbus and really, and I'll let Zach kind of come into a bit more of the details of that conversation, but we started speaking, seeing how Nimbus platform could really apply to some of the problems I had personally seen through my corporate and startup experience.

And we began to flesh out the idea from there. That's a brilliant story, whole bunch of really smart stuff. But you were good enough to kind of pivot when you realized that there was new opportunities. And also, I know, I know Zach, a little bit that you recognize that you need other skills within, within that skill set, which, which I think is very, very important part of this story.

So Zach, I will, yeah, I am, I'm not an engineer by background, so I actually started in finance before moving over to more of a GTM role in Australia, then moved over to, got an opportunity in London where I was club head of digital for big professional services organization. Then spent some time in America, two years over in Dallas or California working on a very large digital transformation project before coming back over to the UK and was managing director for a tech business called HubSync.

They later exited to Tomo Bravo. From there went out multiple agency, been doing a heap of different fractional work and yeah, got introduced to Jeff. Jeff reached out to me through LinkedIn. James, we probably ought to give Jeff a bit of an introduction onto the stage call as well.

I'll pass it over to you in a second. But he can't be on this call unfortunately. He's based over in California and he, yeah, he reached out to me. We were.

And I saw what he was building and his idea for Nimbus and I was like, oh my God, this thing could be massive. And I knew James's background to a lesser extent, as I do now as I've got to know him more. And I just went over to him, I'm like, James, like you ought to have a look at this thing. And when he saw it, he was like, this could have saved us millions and millions of dollars at, you know, the big product development and FMCG companies that he worked at.

So yeah, really excited to, to get stuck into Nimbus today with, with you, HAL and Nick. But James maybe just worthwhile just giving Jeff, who's currently absent from this call, just bit of a quick introduction so it gives more context to the listeners. I think, I think, I think it's really important actually. Yeah, to, to, to give some background to him because he is obviously the pivotal part of how the Nimbus platform has come together, but also the reason in which we are able to address what we believe are like the real frontiers of AI at the moment and where we're really starting to push the boundaries.

He's quite a remarkable, in fact categorically a remarkable engineer. Software engineer. He comes from a very extensive enterprise level software background. So he's worked for Disney, Dreamworks, Nike, Nordstrom, Oracle and Amazon.

Wow. And in very, in, in various kind of senior positions. He was the, the lead on a, a very large content management system project for Disney, where they were looking at like terabytes and terabytes of data being ingested into this platform and how to fundamentally architect a platform that could ingest that level of data. And then, you know, early AI work with them looking at image recognition algorithms to ultimately auto meta tag images of things like Mickey and Minnie and various different, you know, various different Disney imagery content that they had.

So he was, he was right at the front. This was pre LLMs. He was working on that stuff. And so that really sparked his engagement in the, in the AI space.

And this, the engagement that for him, from a technical perspective started when he was 8 years old. Starting to kind of design his own websites at a very, very young age. He actually entered into the kind of local. California, California State Science fair.

Yeah. At 13 years old with the, with the California Science Fair's first ever website. And he won it. Yeah.

Oh, wow. Yeah, yeah, absolutely, absolutely. So he, he, you know, it comes from a very strong pedigree in that respect. And then he went off to found his own tech business in the logistics space and looking at kind of deploying AI and ML learnings to the very complex kind of shipping container triangulation aspects of what they try and achieve.

Has scaled that up to, you know, very decent ARR. And then was looking for the next kind of challenge in the AI space. And it was this, this space that we're entering into that really is like, you know, and we'll come on to this in a lot more detail. This is the evolution of where AI needs to go for it to be widely adopted within business and enterprise.

Indeed. Nick, you had a question? Yeah. And this is a common recurring theme.

The marketing teams, the brand teams, the businesses that we speak with are smart enough to know what solutions that they need and they keep up with the trends in the marketplace and they're looking for ideal solutions. And then on the other side, there's brilliant smart people. You know, you three at Nimbus are creating these solutions. And what it seems to me in, in a lot of cases is that bringing those two communities together is either a fluke because you've sat together in an office and you've overheard and you've kind of got on, or.

Yeah, there's no kind of structure to, to how we are joining up the, the business needs with the solution providers and we, we partner with, with Gray Hair Works, which is a company that does that brilliantly, but it can only do it at a certain scale. Is, is there something, maybe this is a conversation for, for later on in, in this podcast. But, but is there some way that you would recommend a more structured way that either the, the, the marketing communities can find you, the business communities can find you, or that you can go out and you need some support in going out and selling your solution into the marketplace and kind of raising the flag on Omnimbus?

Do you have any advice for people on either side of that fence where there's clearly a fence that is getting in the way of innovation and progress? I think I'll just take a very like kind of early start or early kind of input on that and then I think I'll hand over to Zach who's really been leading that kind of the gtm, the conversational aspect of that. But I think just to start with, I think what we are up against at the moment is a couple of things. One, the market's been flooded with thousands and hundreds of thousands of AI tools that are point solutions that do one thing.

And we saw from a recent report from MIT that actually 95% of pilots, AI pilots don't go through to full blown contracts because they're not delivering against what people actually expect and require for business. So there's, there's this kind of because of the flooding, because of the, the, the floodgates being open and, and people kind of all approaching everyone's inbox with thousands of messages saying I can do this for you, I can do this for you. But there's a reticence to trust new solutions coming through the door that it actually can do.

What they've previously found has failed. But then there's this education piece of actually what is required to successfully deploy AI into businesses needs to be, they need to be educated into, into these organizations on actually what is required to do that and how then it can be achieved. And that piece has got to be in place first before the conversations in the GTM and the kind of exposure to kind of bring those two people together can work because otherwise they're going to be brought together with a misunderstanding of what they actually need to get this thing working.

And you know, one of the things that I think is a big problem in the industry which, which feeds into this is that because of all these point solutions there is a disconnect in organizations now that are deploying AI that is meaning that people are moving in a lot of different directions at quite a lot, you know, accelerated speeds, but not in a coherent, a coherent direction because they're all using different tools, different contexts and what is needed, which is what we're Aiming to achieve with Nimbus is fleet intelligence.

So you can think like we want to ultimately turn organization into a hive mind where every solution that is or every part of AI deployment within an organization knows what the other part is doing. And I think that offering that as a solution to businesses and being able to communicate that in a way that they understand, ah, we're not getting that now. And now we, and we can get it. That will help to bridge those conversations.

Zach, I'll hand over to you, like kind of how you actually facilitate that. Like, I think, you know, where we're at at the moment is there's a couple different problems. There's businesses that are adopting like random different point solutions. And I wouldn't say that that means that your business is really doing AI, to be honest with you.

And I think what one of the biggest problems that we see is when we're talking to customers is they're giving access to 100 ChatGPT licenses, or Claude license or Gemini licenses to their employees. And they're all running off in 100 different directions and asking 100 different questions, using 100 different pieces of context, getting a hundred different answers at ten times the speed that we're able to previously. And that is a real problem for businesses, which means that actually they end up being more confused with their strategy because all of a sudden you've got some person that's been inputting data into a chatgpt and getting results come back and they all sudden like an overnight expert in a certain area that they, that they're not, they don't specialize in.

So like the way that I see it, and this actually came from one of our advisors we're speaking to yesterday is a good analogy, is like right now we're sort of at the moment in the 1980s when, you know, people were using computers, but none of that was connected to the Internet where we're moving to and where AI is going to have the most profound impact for organizations is once everything's connected and they can actually start talking to each other and all the different tools, all the different data points, all the different external signals can actually start talking to each other and inform the business of the decisions they need to make.

When we haven't seen, you know, short of companies that adopt, like Palantir, the very top tier companies, we haven't really seen anyone really do that successfully. There's a lot of people just adopting, you know, you know, an AI notetaker, for example, or something to help on legal contracts over here. And Then you've got the finance team running their own thing as well. And that's actually quite disjointed and not very efficient for the organization.

Great. Which gives. Well, actually, Nick, I'm just going to leap in here and I'm going to ask the question, the stupid question, which is my role on this podcast, which is what an enterprise AI solution actually looks like. What it does.

I mean, you know, you. We talked about many businesses now moving forward with tools like Copilot or so on, but there will come a point where they may think we need to do something bigger at an enterprise level. What is your vision for what that is and what does it actually entail? What does it encompass?

What will Nimbus give them that these other solutions we've just touched on? What is this next level? I think to circle back on that statement, fleet intelligence. It cannot be a singular integration that isn't connected to every other team.

One of the biggest thing that hurts these consumer brands that we're dealing with, but really it's business as a whole, is what we call decision lag. And that is the time that's taken between either taking external information and linking it to your internal decisions, or passing the internal information around businesses to then get a collective decision and then move forward with an aligned decision. And that lag is now at a length that is longer than the time it takes the external market to decide and move on to something.

So what we're seeing is that actually businesses are kind of consecutively and continuously getting behind the market. So we look to try and remove that decision lag. We read a report that 80% of business or enterprise effort is wasted essentially on processing data and not making decisions. And so we want to flip that on its head and say, okay, we're going to now change the dynamic within organizations such that 20% of the time or effort is spent on processing data and getting it together, cleansed and unified.

And then 80% of that time is spent using that data to make a decision. Now, to do that, what do enterprises need? Well, they need three key things. They need context and causality built up for the AI agents to be able to understand.

You can deploy an agent, but in order for it to be this kind of web of agents that understand what the others are doing, whether that's one for marketing, one for finance, one for supply, one for for planning, they need to be able to understand how their actions affect the others and how anything else that's happening is going to affect them. Now, to do that, you need the Two graphs that are critical, a context graph which tells them what the structure is and what the pieces of information are.

And then the causality graph which allows them to understand how they link together, what variables are linking that change the various different pieces of information. Now for enterprise, those two graphs are enormous. And the biggest challenge for enterprise enterprise adoption for these kind of platform. You know, the idea of this kind of platform is how to build that graph in a way that doesn't require huge amounts of forward deployment or immense amount of data analysts within the organization.

It's a huge undertaking for them. So having some way of building up that graph in an automated and much more manageable way is a critical part of getting into enterprise. Second to that is being able to then deploy these orchestrated agents that can use that information and then deploy against processes and speak to each other about what they're doing. And finally to that is the ability to then take those external market signals that effectively will impact any aspect of the business and be able to interpret that information in real time and bring that in and fuse it together.

And that I believe to get to what we like to call organizational sentience, is what organizations, enterprises need to be able to then really unlock the true like calorific impact of AI. Yeah, I think a really good way to think of Nimbus and what James is talking about there is like effectively it's your digital chief strategy officer where it knows every single data point internally, every single external data point it can ask of the AI and run workflows of a thousand things that you've thought of and a thousand things you haven't thought of to give you the right answers to move the business forward.

And that currently is very unique in the market. We're not seeing many others really do this at the moment. And that and going back to what James says, you really have to think about how to adopt it by just plugging in a point solution over here and a point solution there. It's not really adopting AI.

There is a, there is like, has to be. It's a joint effort across the organization to want to actually move forward in this direction. Right there's. So I agree with the kind of central hypothesis here that we see a lot of organizations where individually everyone's using different tools, AI tools and they're doing their job 10%, you know, 10 times quicker, but they're running individually and there's a hundred people learning how to do that, you know, a net they're all learning the same lessons and the, the impact on the business is that they are not that they're actually slowing down.

They're using tools individually to go a lot quicker, but they're not joining it up across the enterprise and, and coordinating those. And we see that a lot and we try and put our arms around it. Can you give some kind of a tangible. Here is a couple of examples of where Nimbus would absolutely intervene and join up, stitch together the silos of the organization just to make it real and tangible.

So Jay, let's talk about his demand planning workflow that we're doing for that customer. Absolutely. And the standard connected workflows as well. Yeah, absolutely.

I mean, yeah, this is something that we're speaking to a lot of our early customers about because it's an area that is fairly consistent across all of those kind of consumer brands. So what we're looking at is an automated forecast and demand planning workflow that is effectively taking data from historic sales data from whether that's data purchased from retailers, SAP data marketing calendars or strategic calendars that, that they would look to use to conduct their forecasting and demand planning exercises.

And it's taking that information and then synthesis, cleansing that information and unifying it. And that's kind of actually that it's at that point there that you're taking spreadsheets and just pulling them in. That's where then their current use of AI that we're seeing with a lot of these businesses stops. You can take a single spreadsheet and chuck it into copilot and it will tell you about some estimated trends from that one.

There's no auditability and traceability in that, but it's an individual spreadsheet. Taking multiple pieces of different disparate data and unifying that into a centralized understanding is already stepping Nimbus ahead of the curve. Now what we do from there is we then creates baseline forecasts from that using the kind of standard plus AI statistical analysis. But on top of that, what Nimbus can then do is take the external signals that it's extracting from large language models, from social listening to inform that statistical baseline as influences to forecast so it can push and pull the line to ultimately make it as real time relevant as possible.

And you end up with this enriched demand plan. Now, in an isolated workflow situation that is already very compelling for businesses, it's like taking, you know, the need to have multiple FPA analysts to look at this. It is enhancing what the existing analysts can go and do and it's centralizing that all within a platform that you can See all the audit trail and the history. However, what Nimbus offers on top of that is the ability to then have multiple workflows that can work in conjunction with that.

So you might, you what, from what I've just described, you've got an unconstrained forecasting workflow, but you then couple that with a supply chain workflow that understands exactly what capacity they have and manages that aspect from the supply base. You can then link the two to give a constrained forecast and demand plan that says here's what the market wants and here's what we can actually produce. So here's the limit and the constraint. But beyond that you could have a financial reconciliation workflow that is able to speak to the both of them and say, well the market really wants this.

Like there's a significant over demand here from the market and we can only meet this capacity which is constraining our forecast. So let's release some more capex and have that discussion to say let's, let's enhance the capacity. So this is, this is this like web of information that's happening in real time. But in this instance the humans in the loop are the ones just looking at the forecast and saying yeah, that makes sense.

And looking at the supply chain saying okay, I'm seeing red flags here, I'll go and address those, et cetera. And you've got someone that going okay, well actually I do approve that CapEx because that does make sense, let's do that. And it becomes more about driving the business forward as opposed to then processing that data and seeing how it kind of links together. Right.

That this is fascinating. I think it's probably useful if I try and play back my layman's understanding of, of what we're talking about here. We have really is, I think it's an ecosystem of agents that are centralized under I guess an over overarching one agent to rule them all at the center that provides that. And then that ecosystem is able not only to use your own business data that huge sway.

The business data, which I think you called, you referred to as the graph with the retrieval augmented generation. The external sources, whether it's, you know, it's the live web, it's external data sources to help inform your business strategy, whether that's on manufacturing and demand side or it could be any area. Because I think that leads into my next question, which is what sort of business is this for? Is it for complex manufacturing businesses or do you see this as being applicable for pretty much any business out there?

Sorry, yeah, so one other use case we didn't really touch on as well, that's probably quite relevant to your audience. James spoke about, like how this could apply to the financial side of an FMCG company. But obviously your audience is quite marketing focused. So say, for example, you've got a FMCG brand that's looking to launch into the US market.

And they want to, they want to, they want to know what the US market is saying about their sort of product category, what the marketing messaging is. You know, how, how much money should they put behind this? They can use Nimbus to actually simulate that launch into that market by using these different workflows. So we.

One thing that we haven't mentioned as well is, is when we run these workflows, particularly the simulation one, we are going out and we're sending out thousands of different prompts to all the different large language models and getting all that information back in and distilling it very much in a similar process to like distilling sitting whiskey. You've got this, whereas you're sort of like getting a drip feed of all the distilled whiskey coming out. Because what happens is, and James will explain this much more eloquently than I can, is these, as everyone knows, these AI large language models, they hallucinate and they give back answers that don't make any sense, but we play them off against each other when they do come back in and hallucinate.

So when you are getting these answers come back in from these thousands of prompts that we send out to these large language models, you have quite a high confidence score in the results that you have there, I think. So as a marketer, marketer, you talked about demand forecasting, demand planning, and marketers are kind of, we are the demand end of a business and that's the supply end of the business. And everyone talks about demand forecasting and how technology will allow us to get into demand sensing.

And essentially it is always reactive. With all of these models, you're just getting slightly better, but at the end of the day, all you're doing is becoming slightly better at being reactive. And I think what is interesting for a marketer, and in fact the way that organizations should now be thinking about this, is to link up the supply part of their business with the demand part, which is what your solution can do. You understand the velocity of sales or the whatever it is that's going out or the supply chain feeding into your warehouse and how many things are going out on trucks.

At the back end of it, you can also influence that by the marketing side, which is we've got 20 of these things left in the warehouse and it's, you know, if we don't sell them in the next week, so you can, by joining up the enterprise, you actually break some of these accepted models that are reactive and become proactive. And only an enterprise solution, only an enterprise way of thinking about what is it we're trying to do. We're trying to maximize the profit of selling these, whatever it is, the services, the products that we're selling.

And until now, everyone's focused on either silos or point solutions. What Nimbus is now bringing is a very easy way of joining up organizationally across the whole piece and making decisions, making it transparent, but also allowing some autonomy to make those decisions that optimizes the activities across the whole organization. Absolutely, Nick, absolutely. And I think just to kind of touch upon one point you made there, which is something that I ought to just embellish on a tiny bit, is that what Nimbus is doing is building up for an individual business or brand.

It builds up, essentially, you can think of it as a small language model within the Nimbus platform that is learning more and more about their business and learning about how they operate. And it's, it's, and it's, it's being tuned to ultimately be the brain of the business now that has multiple uses, whether it's to protect the business from people leaving the organization and the knowledge drain from an organization because it's all in there already and a new employee comes in and knows exactly how to do that.

And that's such a key point. I mean, just really want to leap in because in the entire experience of my career has been how much pain, that inability to maintain institutional knowledge, either because the person you need to speak to left last week, or it was on someone's desktop and they left it on a train. You know, that kind of thing is very common, I imagine, for absolute. Absolutely.

And you know, and it's, it's becoming worse because, you know, people, the time at which people expect to stay within a company is shrinking and shrinking and shrinking. And you know, you could look at someone's LinkedIn profile and see like, that they moved company every 18 months and that looks normal. You know, so actually if it took six months to train them, you've actually only got 12 months of usage out of them before they've gone again. So, you know, that, that in itself is, is, is a powerful tool.

But what this, what this learning is also doing is Building up a massive piece of information about your demographics, your markets, and starts to give you the opportunity through these workflows to simulate the market. Now this isn't, this isn't looking at projecting or giving you like, you know, I think this is what sparked this, this point. Nick was not, was, was not being reactive and actually starting to get into the position of being proactive and so being able to simulate the market and say, right, okay, now it knows this stuff and I'm doing my forecasting, what would happen?

And you can say this conversation links to the platform. What would happen if I were to change our plan to do a 30% Black Friday discount to a 20% discount? What's that going to do? And it could like go well actually, because of this, this and this, your competitors are likely to do this.

So this would make you uncompetitive and, and perhaps impact the sales in this particular retailer and yada, yada, yada, start to war game what might happen ahead of committing any capital to it. And that in itself is an enormous risk mitigation for all of these businesses. That's fascinating. So Zach, you spoke about Nimbus being your AI enhanced strategy officer.

To what extent does Nimbus operate fully agentically and therefore make decisions itself? Or to what extent do you see humans as being essential within this ecosystem, within the loop, to have final sign off to make that final decision? Okay, so every single workflow that we have within the platform is very consistent. It consists of five different blocks.

It's got analysis phase, synthesis phase, it's got. And then there's a human in the loop phase. What are the other three phases, James? The input and the, and the action.

Yeah, yeah. So analysis, synthesis, input, and there's a human in the loop that is consistent through every single workflow. So before, so what happens is, so say we're going back to this demand forecasting workflow or even the marketing market launch simulation. You'll run this workflow, you have all your data inputs that comes from your digital twin, the business, which is your context, ontology, where it knows everything about it.

It then goes out and makes sense of that data and then it brings you to the canvas, which is effectively the human in the loop. At that point there the workflow stops and that's the opportunity for everyone that's a part of that workflow, whether that's the cmo, the VP of Marketing, and Marketing Manager, marketing associate, to come in and review that data. Because you don't want these to be running fully agentically and running Wild and taking actions on your behalf without having a chance to review that information.

So every single workflow has this human in the loop step. They come in there, they review all the data that's there. And if they see things that don't quite look right, they agentically talk back to the platform, be like, this doesn't quite look right, we need to run this again. Goes back and runs it again.

And then they get to a point where they're happy with that output. And then from there there's, you know, like they can have different action points. It's like, okay, I approve this. And then that triggers an action to send a slack message to the team or it can even trigger something else to run a separate workflow.

Or it can send out, you know, an email to the executives with a PDF attached to it with that report. So in every we, we see this as, you know, helping businesses really like 10x the investment they can get out of their employees. Because now they're not so much as James says, they're not spending 80% of their time inferring all the different data points and then spending 20% of their time, you know, reviewing and making decisions. They're spending 80% of their time reviewing all that data, making decisions, and then to move the business forward.

So it's crucial across all of our workflows. And we recognize this because that every single workflow needs human intervention. Otherwise that's when hallucinations happen. And you don't want AI running wild across your business, just doing things willy nilly.

So, yeah, I think there's a, there's a very relevant related point and this seems to be coming up repeatedly. And we're going to kind of, we're going to copyright something syndrome, probably the Darby syndrome or the Nick syndrome, but. There'S medical books already written on that. Yeah, yeah, yeah, yeah, yeah.

We'll have to think. But the number of people who are in it and say, well, I'm using AI tools now and I can do this as well. I can do SEO as well as a marketer because, you know, I could type something. They just don't know what they don't know.

And we had it this week. You know, I could run the HR department because I can type in and get the policies and personalize them. And you know, people are starting to imagine that these tools equip them with the human expertise. And what you're saying from the human in the loop and the network is you are actually amplifying the expertise within, you know, kind of marketing or within hr, within operations, all of that, that good stuff.

And your solution is just freeing up time to allow people to apply their actual experience, their relationships, their there and it will capture some of it. But what I, I'm hearing so often, and I just kind of, I roll my eyes almost that they appear looking out of my ears at the end of it is that is a ludicrous statement to say, well, I've got an AI tool and I've spent 20 years in it, so now I could, you know, I could do the job of a marketer because they've just never done it and, and they imagine how it is, you know, I can write.

I wouldn't put in my passport that I'm an author. Yeah. It's just that it's that leap and that gap that people are doing. So I really like what you're saying, which is it amplifies the human expertise, it joins it up, makes it transparent so people can apply their individual expertise to it.

I think that's the way it should be going. There's been a couple of horror stories that your audience may be very familiar with. One very famous one out of Australia actually, where Deloitte handed a $400,000 report to the government. They handed a four.

They charged the government $400,000 for this report. And it was just done through a Gemini Deep Research or GPT Deep Research. And you can argue what their human in the loop was, whether it was much review there, but if there was actually something within that workflow, quote unquote, to like stop and review and actually in detail review that information, then that may have never happens. So it's, it's, it, it's absolutely crucial that with any workflows that you're running with any business, whether it's through diminutive or any other workflow platform that you do have that human in the loop step.

And you're not solely just relying on the outputs from AI. Yeah. But I think what's also interesting, I'm gonna, I'm hopping back to a conversation we had recently on the podcast with Dave Martin, a hugely experienced Chief Product Officer. And he'd been in some AI enhanced workshops from his work and he was very interested in how the AI had removed the emotion from the decision making so that it was no longer a question of ego or we've all been in those meetings where the highest paid person's opinion, the hippo, as it were, tends, you know, tends to override.

And in fact it just then became, you know, a very fact based Data based decision. But then as a marketer I would then say, well yes, but actually a lot of depends a little bit on the product. But still most people, it is an emotional decision as much as a practical data driven decision which leads people to actually purchase. That's why creativity is still hugely important.

And I'm just interested to understand how far you feel Nimbus is able, you know, where, where, where does, where does it have its limits in terms of. Yes, we can absolutely synthesize and give you the absolutely most insightful marketing plan and consumer behavior is all considered. But to what extent do you still have to say, well there still needs to be some of that secret sauce from our creatives, our humans to really make that fly? I think, I mean, well, to touch on like the, the hippo side of things.

I mean I've been in a number of organizations, I shall not. I mean you guys can try and work it out, but you know where we have put all of the data absolutely kind of crystal clear in front of people and still they still there's like someone in there that just says well now I want to do it this way. So. And I don't think you're ever going to bypass that in any organization.

But what Nimbus unlocks is, is that data is. It unlocks the ability to see the wood for the trees. With every decision that's made, not just the like the super critical ones or the ones you've got time for, every decision becomes one based on factual information that's relevant at the time and no more gut feel kind of decisions that are ultimately putting the business at risk. And then whatever business decision is taken is still down to the humans in the organization.

The secret source of what humans still add into the loop. With regards to that, that is critical. Let's not let, I don't want to, I don't want to even slightly sugarcoat that one that is critical to AI deployment. What humans need to be working with AI to ensure that not only is it steered in the right direction, but that it is still doing some of the things that, you know, humans do best.

The creative thinking, the outside of the box thinking, sometimes the, the analysis using things that aren't completely rational and therefore can be considered, which would otherwise be hard for something that's kind of reasoning continuously which is what AI is doing. I think that is super important and I see that being a massive part, if not a continual part of how AI is growing into businesses. I think as James says, it's like the real creator side We've got this workflow where we're using it as, you know, the pre funnel intelligence where we can go out as a consumer goods brand and query the market on your behalf and understand, okay, where what, what is likely to be the next product that people are asking for.

And we do that by what I've said, by going out and prompting thousands of different large language models, doing social listening and that kind of stuff. And that will, you know, bring you back a heap of different data points and you know, you can even spin up a different render of that, that product of what it could look like. That is when that, you know, someone like a, you know, experienced product designer or experienced marketer needs to come in and be like, actually hang on, this doesn't look quite right because, you know, there needs to be, the laces need to be a little bit different on this kind of shoe.

Or you know, we can't have like a, we can't have the football studs be in this certain position or whatever. Like that is when that human comes back in. And that's where we see the Nimvis platform being like, like pouring gasoline on their potential because we can help circumvent a lot of the market research side of things, but we still need that people, those humans to come in and actually add their input and their creativity. Yeah, Nick, so I'm coming back to my original question and I'm increasingly interested in how marketing people identify the right solutions and how the solution providers.

So you have a brilliant solution here. And as a marketer sitting in a marketing team, I'd be thinking, okay, so first of all, why would I spend any of my budget benefiting anything outside marketing? So, so I'm, I'm just constantly thinking what's the setup involved here in, in terms of the reality of, of that setup? And then also when with kind of a framework tool like you're applying what's, what's use case number one for me.

What's use case number two for me. How are gonna, how am I going to save my career? Because I've invested in, in bringing you in and paying you the consultancy fees to set it up and plug it, plumb it into kind of all of the platforms. What is it tangibly that I'm going to start with that's going to demonstrate something for marketing because that's what I'm KPI'd on.

Or is it that the buyer has to be the join at the top of the organization, which is the CEO or the cfo. I'm going Back to how do I sit in a company and find a solution or how am I a solution provider feeding into a company. That's the central question that lots of people are wrestling with. So I'll, I'll just kind of take the first part of that.

There's, there's kind of three, three elements to that. Isn't that there's why, why would I do it if it's, if it's impacting other people around the business? I just care about my, my sector is, you know, that's not an uncommon way of thinking. You know, how, how does it then, how do I use it?

What's the first use case? And then, you know, whether or not there's someone above that should be the, the key buyer. I think to, to answer that one. No, I think there's, there's, there's various different ways in which we could, we could address that.

If I was addressing yourself as a marketeer in, in an organization, why would you care about something that can impact the other teams around you? Well, as a market, from my experience within these product development companies, you know, the marketeers need to know what the product development teams are doing so that they know, you know, how far away are they? Do I really want to kick off some marketing launch now? If, if I've been told they're six months off, but actually they're 12 months, you know, it's understanding exactly what.

Well, okay, if we are six months off, what features are going to be at that drop? What can I start talking about? Is there anything that I should really be focusing on? Because that feature is really resonating with the market.

How do you find out what's really resonating? Well, that, that comes from the product definition team. So there's like, there's the broader spectrum of information that you can get from the platform that is going to inform what your, how you market. You know, again, circling back to the financial, the forecasting, demand planning that would have access to all the historical sales data.

You might be thinking, yeah, well, right. We've got a bit of a boilerplate for how we market here. This is work before, this is what we're going to proceed ahead with. And then the Nimbus platform tells you.

Hold on a second. Actually, that didn't perform last year very well because of X, Y and Z. And actually the sales are dropping here. So I wouldn't touch upon this way of marketing.

So I think there's, and that information, you don't need to go have a review of them. It's all there at your fingertips. So it's giving you a lot of external peripheral information that you can only get if the other teams are engaging with this platform. So that unlocks that for you.

For Nimbus to be the most impactful within an organization, unquestionably you need to deploy it across the entire Org while saying that though we can spin up like very specific context for different areas of the business. So, you know, the context being the digital twin. So say if you're a marketing team and you're okay, well, there's three key workloads that I really need help with. That's you know, running up campaign messaging.

That's getting understanding of like how where we should be spending our money on paid ads and you know, whatever, whatever. You could use Nimbus and build up a semi like a subcontext within just the marketing team, all the different data points, and then have those workflows that run over the top of it and then that you can use it as a proof point to then potentially roll this out across the broader organization. So there is, there's, there's different ways that we could, we could definitely cut it with a team that just wants to use it within the marketing team.

And you know, we're doing, doing, doing that right now with a couple of our other customers where we're dealing with, you know, a certain segment. But there's roles, there's, there's a plan with those customers to roll it out beyond those segments into a broader pilot or broader engagement. I should say. Yeah.

If we look at how even kind of sales and marketing teams, there's digital and there's field sales and there's B2B. There's B2C in the organizations we work with. Even that, that small part of it. I'm not small part of it.

It's a contained part of it. Sales and marketing is still enormously siloed either by brand, by market, by by function within sales. Is it online? Is it field sales?

Is it just. It's enormously. And I think if you, if you can be pragmatic and it sounds like you can, depending on the size of the organization, you can go in and say, let's put this umbrella in over the top of the organization we benefit from because it's a smaller organization, we break down the silos, bigger organization. Let's not try to eat the whole elephant in one go.

Let's a coherent either a market or a brand or a sales and marketing within the organization and prove that breaking down those barriers and giving that visibility, the transparency and the decision support is of benefit and then we can elevate it up. And as a marketer trying to make my career, I would absolutely be all over that because then I've got visibility to the CEO and the CFO saying, you know, I got, I got proof that this works in my small area and I've had the vision to bring this in.

Let's do it on an organizational basis. And I think that is a huge benefit to people. Just, just, just one extra thing onto the kind of, how would you deploy it and why would it, why would it matter to you? You know, Nick, as a marketer, the, one of the biggest challenges for anyone looking to market a product is, is, is how do you understand in real time what the exact sentiment of the market is at that point?

Almost prior to large language models, that was absolutely impossible because if you scraped the website, if you scraped the web, you're still really only getting stuff that's being published by someone at a particular point. It's, it's, you know, it, it has a delay cadence to it. If you went to the traditional channels of doing, you know, focus groups, interviews, speaking to a big consultancy, there is at least weeks in that, if not longer, prior to actually getting that information in.

And huge bias. Exactly. Huge cost. Right.

So, but what, you know, there's, in any given day, there's millions of conversations happening in the large language models. Across the whole model range, there's tens of millions. And those conversations are people typically asking a question of something they want to know a bit more about whether that's like going to go out and do a trail run at the weekend. And I want to know what the best shoe is because it's kind of slippy at the minute, yada, yada, right, all of that.

And it will answer back and then they query it a bit more and the answer back will be based on other people asking similar questions and perhaps like giving their input, saying, actually this shoe really worked well for me, but I'd like something else. So they know then that that shoe worked well. So that information is up in the ether at the moment in the large language models. And if you can extract that information, which is what Nimbus is doing, what you're giving marketers is the opportunity to get real time understanding of what the market cares about relative to their context.

Now that in itself is very, very powerful, but that, and that will allow you to inform what the product needs to look like, how you might go and market which markets to look at first because they've got the strongest sentiment and therefore the most clear ROI on any, any marketing spend. But it also allows you to test against the market. So if you come up with some marketing copy, test against the market, will that copy and it doesn't have to release the exact copy to then expose what you're thinking, but it can start to use pieces of that copy to see whether the market would resonate with that.

So your A B testing almost becomes free through the Nimbus platform where you can then start to see that and it's at a very, very big scale indeed. And your competitor analysis will be probably much enhanced from what traditionally you've been able to do. Exactly. So you touched on this.

But I mean enterprise solutions do invariably require a lot of stakeholder buy in beyond marketing teams, they require Chief Technology Officers, etc. How and you've talked about actually the Zach the most effective way is to use Nimbus in its entirety. But often organizations will require a pilot scheme to as proof of concept. How long does it take to get Nimbus working in your organization up and running to some capacity to start to make those decisions and to start to see the benefit.

At the moment we're working with a handful of selected clients and those deployments take anywhere between four to six weeks for a certain set of workflows. Wow. Where we're getting to though is a point where you'll be able to go to the website and tell the Nimbus platform, okay, these are the workflows that I want to run. And it will build up your own custom workflow package for you and then you'll sign up and get going.

So it'll end along the whole journey you'll be handholded through this onboarding process. So connecting your, the relevant integrations that you need, connecting your data, uploading different pieces of information like everything like that. And then it'll have the workflows there for you to be able to run. So eventually in the like late Q1, Q2 this year, it's going to get to the point where you go to the website, you tell the platform agentically what workflows you need, what ones are really causing massive pain points and then it will build up your own custom package for you.

And that, that onboarding will be very seamless and only take a few hours to a few days. AI being used to configure AI is a. But this exactly that. It's, it's, it's, it's robots building robots in a.

In essence you know, it's exactly that, that that onboarding process is, is AI. If you aren't sure how to configure something, you speak to it and it will speak back to you or it will go and do it and you just tell it. It is in this agentic era that we're starting to enter and, and I mean true agentic, not just the kind of the agents that we've heard about in 2025. Yeah.

In this era, you know, it is becoming about these things being able to just do for themselves and you be there to just like say so really, you know, in a really dystopian sort of outlook on what Nimbus could be. You've just got to be there. As long as you've got a credit card you tap and then it does the rest. And that's how we want this kind of deployment to be.

And yes, the larger the organization, the slightly more in depth that will be. And there will be some instances, don't get me wrong, where we could deploy to support. But in majority of cases we see this as being a very hands off experience. Yeah.

And we see this being a True unlock for SMBs and mid market enterprises where they can just be incredibly more efficient with their existing resources, existing headcount and really maximize their investment into those areas. And then obviously as James says, like you know, the enterprise level of clients that will on board that will require some sort of like hand holding to get them in because there's, they're more complex. So. Yeah, can I, can I get your views on a question that again comes up quite a lot and there's a range of opinions on this and the market is so nascent that nobody really seems to have a good answer.

But the span of expertise that an individual agent should have, some people are saying you have a marketing agent and the same thing you'll have an SEO agent for Google and you have a geo agent for you know, each of the individual LLM. What's the, what's the granularity that you see it will, will settle out at will that depend on different factors or should it be very small niche expertise or should it be broad expertise and that supports. Where's the granularity of that expertise going.

To say there's a, there's a couple of schools of thought in that at the minute and I mean well to speak to Nimbus at the minute, there's 31 agents that we've deployed within the platform and they all have their own different functions. It's not so Much necessarily about the capabilities of an agent, rather the guardrails that you put in place with an agent so they can't go off and start impersonating. That's another thing. So models hallucinate AI agents impersonate, and that's a new thing.

Yeah, yeah, absolutely. So they actually ensuring that agents are constrained to their function or constrained to their particular remit is essential. And we are kind of developing agents continuously. And as we get more complex situations, some agents will actually be developed by the centralized orchestration agent itself.

There is this other school of thought around models, the large models, but a model trained specifically to be able to train agents and developing models that ultimately are in the background on our server end that are able to spool up new agents given a set of requirements or constraints. So we would see us continuously developing almost like this conveyor belt of develop or create, evolve and then perhaps destroy agents on this conveyor belt of churn as the Nimbus platform is used.

And I kind of like, I, I, I, I see that being very much like, you know, the, I think it was the Maravingian in, in the Matrix. He was that program that was there to coordinate the kind of creation and destroy, destruction of, of bits of software within the Matrix. And Matrix is here. That's how I'm gonna, that's it.

So we, we, I hope we aren't to blame when we're all stuck in capsules as batteries. It'll be Elon Musk, won't it? Let's be honest. But yeah, that's, that's how I see that happening.

Is this like kind of this agent at the top that can help to create and destroy agents as and when required? Yeah. The best way at the moment just to quickly round up here is like this. There's three layers to the Nimbus platform, four layers, I should say.

You've got the digital twin, the context ontology that sits at the very bottom, which is all your data. You've got the workflows that sit in the middle. And then at the top you've got the command center, perception console. And now the, there's like a spectrum of users that use this.

So at the very top level, the command center, that's like your C suite executive asking questions and triggering workflows from that perception console in the same way that you would in Perplexity or chatgpt and ask questions of your data. Beneath that you've got like the workflows which would be like your analysts, marketing managers, stuff like that that are working in there and then your developers are in, in the context ontology that are connecting all the data all across.

Every single of those three layers are the agents that work in different capacities. Some have longer term memory than others and others, you know, others have shorter term memory because they don't need that longer term memory. So the platform is more efficient and you use less tokens. Whereas, you know, there's, there's others in there that are designed specifically to have very long memory and to, you know, always maintain and continue to, to understand what your data is doing.

So hopefully that helps add some more confidence. I think that's the best and clearest explanation of where it's going to level out and how to think about it. Because everyone's thinking about what's the expertise of each of these agents. Actually you're saying it's going to vary across different strata according to the purpose of them.

And actually the thing that defines them is the guardrails. And you will probably have something eventually that is an agent creating agents or an agent training agent. So, so, so we'll work out the right level of granularity, the right level of the memory, the right level of the training and that's going to be enormously useful to everybody who's asked us that question that, you know, that that's a great explanation. Thank you.

Great. James and Zach, it's been a really exciting and illuminating conversation. It really sounds like your solution is going to enable businesses to really scale at velocity in a way that, you know, historically has never been possible and to maintain that, that level of legacy knowledge to again further iterate and evolve and accelerate way beyond what we've been capable of doing to date. So just, you know, I hope this has piqued a lot of interest from our listeners.

For those out there who want to find out more about Nimbus, they want to follow what you're doing and get in touch. What's the best way to do that? So, well, we, you can, you can find out more on our, on our website. Although that is continuously evolving, as they always are.

Yes, as you can imagine, that's it. Go Nimbus AI. So literally it's G O Nimbus AI because some people have gone to Nimbus AI instead of. They take the Go Nimbus.

Yeah, we'll put that in the, in the show notes. Yeah. When, when we can afford to, we'll buy Nimbus AI but at the moment we can't. But yeah, Gonimbus AI to find out more or you can reach us on email.

You can reach me at jamesimbus AI And Zach, go Nimbus AI Z. I say excellent. Yeah, well, I mean, we started looking at this from a British perspective. I guess you'll take phone calls or inbound from anywhere.

But I really hope this really helps accelerate a lot of British businesses in the first instance to really. To really power up and scale globally based on the great work you're doing. Thank you so much for your time. It's been a pleasure.

Thank you so much. Yeah, thanks very much for having us, guys. Yeah, best of luck. And, Nick, until the next one.

Thanks for listening.