← The AI in Business Podcast
The AI in Business Podcast
Why Predictive AI in Service Only Works on the Right Foundation - with Niken Patel of Neuron7.ai
00:28:14
Loading summary
Transcript29 lines
- [00:00]
A
Foreign. Welcome everyone to the Emerge AI in Business podcast. Today's guest is Nikin Patel, CEO and co founder of Neuron7AI. Nikin joins us on today's episode to discuss the necessity of building an intelligence layer that transforms fragmented, incomplete service data into a format ready for automated decision making. He explains how service organizations must move from basic productivity tools to a deterministic foundation that resolves complex technical issues across departments. By establishing this data foundation, companies can transition from reactive repairs to predicting equipment failures and reducing unnecessary technician visits. This shift allows leaders to achieve significant financial gains and maintain equipment uptime rather than settling for minor efficiency improvements. Today's episode is sponsored by Neuron7AI. Position your brand alongside the Fortune 500 leader is defining the Enterprise AI roadmap for the opportunity to showcase your solution to the executives currently funding and scaling global initiatives. Partner with Emerge to reach the decision makers holding that strategic mandate. Secure your partnership@go.emerge.com partner that's go.emerj.com P A R T N E R Now the conversation with Nikken Nikan. Welcome to Emerge AI in Business.
- [01:29]
B
Thank you. It's fantastic being here.
- [01:31]
A
I want to start with something that keeps coming up in every conversation on our show lately is that the service leaders come here and they walk in saying that we already have AI deployed somewhere in our organization, sometimes in the contact center, sometimes in dispatch, and they usually still have the same problems they had three years ago. Even with AI now integrated. They see repeat truck rolls, they see issues that just keep on coming back. Where do you actually think the service organizations are still falling short even with AI already in the room?
- [02:01]
B
Yeah, that's a great question. I think if I take a quick step back for companies and leaders who are saying it's super important to kind of figure out what is it that was their goal when they deployed the AI they deployed. Right there is the easy button of AI, as I call it, where you can do quick summarizations. You can call that productivity. If you take first time fix and first call resolution, a whole bunch of things go into influencing it. And I could still go in and I could take a couple of things, couple of minutes off my call and call it productivity. Call it, hey, I'm helping first, you know, faster resolutions. But I call it the easy button of AI. It's basically the 50k ROI. It's not the 5 million ROI. Right. And what we want to do, we want our customers and people that are working with us to be bolder, to be ahead of the innovation curve. If you will, because I think we are at that point now that enough has been done and innovation and impact are absolutely coming together. Right. And so if they benchmark against other industries and stalwarts in their own industry, they will realize that we are ready for the 5, 10, $20 million ROI of service resolutions. And from a neuron 7 perspective, if you think about it, we've always said this. We've been in business for a good six, six and a half years. If you think about what is the most important decision that gets made in the world of service, it's how do I resolve this complex issue. Right.
- [03:37]
A
Yeah.
- [03:38]
B
And we are going after the heart of resolutions. And so I think when either you've not aimed high enough or benchmarked high enough, or you don't have the foundation in place, the decision layer in place to make that impact, those are the two pieces, business and tech, where I think people are falling short.
- [03:56]
A
Absolutely. And before we get into resolutions, I think we have many listeners that might not even realize that they have an AI problem. They say, no, we've got AI integrated and deployed. And they see that 50k ROI and they think, well, that's enough, or that's as good as it's going to get. So when a service leader at a Fortune 1000 tells you we already have AI in our contact center and you dive into that and you look at what they are referring to as we already have AI, what is it that you actually see there at the moment?
- [04:26]
B
Yeah, I think we do a pretty awesome exercise of discovery first. Right. So I think there is multiple layers of the onion. We want to kind of take a look at what the executive has. Right. And most of the executives that we work with, our customers, our ICP is pretty clear. You're either spending a lot of money in service or you're making a lot of money in service. It's one of the two. And so the leaders are looking at, hey, cost to serve. And a lot of leaders are saying, I'm moving beyond operational efficiency. I now need to make more money from service than I used to before. So either drive my revenue or make my cost to serve. I want to do more with less. Right. And I think if we take the bigger picture of what they want, the next level, we kind of go in and say, okay, what is it that the senior directors and other folks and people who manage the call center and field service, what are they looking at? We try to get a good mental map of what this organization is really dealing with. And what is it that Their goals are right. And hopefully we land up with somebody who's in the 5, 20 million ROI bucket and not in the 50K bucket. And that's when we start with benchmark and education for our leaders who work with us saying, if you don't do this to the industry, the industry is going to do it to you. Right. If you are not at the forefront, your customers are going to get better outcome base service contracts from somebody else, they're going to get better uptime from somebody else and it's a matter of time. You're going to be forced, your hand is going to be forced. You as well may be proactive.
- [05:56]
A
It sounds like there is a lot of proactive requirements there and we'll get into what that means in the industry in a bit. I've heard you talk about it publicly about fragmented signals. Can you give me an example of a recurring failure that no single team in a service organization can see end to end at the moment?
- [06:17]
B
Yeah, I think, you know, we, we have looked at the data of 300 plus companies in the Fortune Thousand, right. We've been lucky enough to kind of look at a lot of, you know, what's, what's the reality of data today? Right. I may have a desire to go to X, I'm a Fortune thousand company, I got 10,000 employees in service. But I have a woefully low data foundation to work in. A good example would be the entire resolution path, which includes everything from understanding what the issue is to determining how to solve something to get to a step and order parts. But then the parts may not be available because inventory is low. The world is dealing with supply chain issues and you need to find like, for like parts. That is a cross departmental solution, right? This is not something that just service takes care of. It goes all the way from finance to operations to service to call center and including customer success and sales. So there is a data foundation or a foundational layer that needs to exist, what we call the decision layer for service that will actually touch all of these people. It will touch the data that exists in applications across these departments.
- [07:36]
A
So it sounds like the foundation is where it's going to be at before we get into that decision layer or predictive layer. So if the foundation is the gap, and this is where I want to take this next because right now every vendor on a trade show floor is using the word predictive. Right? And most of the conversations we're having on the show are circling the same uncomfortable question of does the foundation underneath actually support the word if we get into what predictive means, if a senior leader is hearing predictive in every pitch, on every call right now, what actually has to be true under the hood before that word means anything? So what is that? Foundation?
- [08:19]
B
Yeah, yeah, that's a fantastic question. I think today the fallacy that a lot of leaders are falling for is they're assuming that their underlying data that comes from CRM and KB articles and manuals is what we call AI ready. The truth is companies are not sitting on AI ready data. So putting in giving all of your service data to an AI vendor or yourselves, working on models and directly going and finding outcome based returns from it is not possible today. We don't think the world has got AI ready data. And that again goes back to the number of companies we've seen in the past, right? I'll give you another stat. You know, I think a lot of people think about agents and the world is going to go towards agents, right? Agents are now going to access your enterprise data. If you go to your website today, 20x30x more bots come to your website than humans do, right? Your websites may not be ready to talk to bots, right? Machines look at data very differently than humans. That same thing is going to come on your. Within your enterprise, within your firewall, agents are going to access data. Now how many companies have got their data, quote, unquote, AI ready for agents to access it, for algorithms to access it? That is woefully low. And I think that is the foundation, right? Creating a layer or a foundational intelligence layer, whether you call it knowledge graphs, context graph, I think the world uses ontology. There's a lot of technical terms being thrown around. But an intelligence layer that gets your data AI ready is the most important thing missing in the world today. And I think I cringe every time somebody says data is the new oil, because that is not true. I want to just like that is not true. Getting data ready for decision making is the new oil in a way that I've got the entire foundational layer in place. When I need to make a decision, I can dip into this AI ready intelligence layer and make a decision. And thousands of them with multiple different roles. Now that is the oil, right? And I don't think enough work is being put in to create that layer, right? We're going in in a siloed, fragmented, incomplete environment and trying to get AI to work. And that's where everybody is tripping.
- [10:41]
A
I think that lack of, I don't want to say lack of trying, but that the fact that we're not putting enough work into getting to that layer is because there's also not good understanding of the technical terms. So if we use the phrase deterministic base, so if we need to have something that, that actually uses the information in a way that makes sense for a non technical CEO listening to this, what does that mean in plain language? Why can't they just use an ordinary LLM on its own? Why is that not enough?
- [11:11]
B
Yeah, I think, you know, I think what, here's what people will understand because they use this in business all the time. Like you got to think about, you have to keep the end in mind, right? So I have a goal and I have a goal to get sentiment analysis done because I have a huge churn problem, right? Or I have a goal to go predict before something goes down. I want to predict with excellent confidence when an issue is going to happen or a component or a subcomponent is going to fail. Those are two completely separate goals. Both of them significantly, significantly impact your customer base and your customer experience. It can absolutely impact the next renewal that your top customers are going to are they renew service contracts with you or not? And everything falls under the customer experience bucket, right? But there is just too much jargon being thrown around. There is the same vendors talking about first time fix and customer experience in the languages. And I think the main thing the leader should say is all right, I want to work backwards with my goal, right? My goal is X versus Y. And I think people will understand that I need to have the domain specific data ready to go hit that particular goal, right? And so if I just take my existing data and just give it to an LLM, right? My existing data has got a lot of incompleteness, right? There's a lot of case resolutions that say application was fixed is working per design. What does the LLM understand and learn from? How does thing resolved? You're now just giving it fodder for hallucination. Same thing on the incoming cases. A lot of people would come in and say application is not working as designed or device did not function. What do I learn from that? Nothing. I think the reality is if 60, 70% of the data is either incomplete or inconsistent, then AI is going to work exactly as it's planned. It'll basically learn from that, from that fragmented, incomplete and inconsistent data. And that's why it's not as easy as just giving your data to an LLM using a rag or a search. And we like to say this is not a search problem. We like to say that a lot to our customers saying, hey, you're not going to just give data. This is not a search problem. The best AI search tool in the world is not going to help you solve this is not going to help with tribal knowledge capture the knowledge exists in your organization that does not translate to the knowledge exists in my data. Those are two completely separate things, right? And you need a purpose built solution intelligence layer to work towards the goal that you have decided for yourself.
- [13:59]
A
And that's going to be different for every organization, which just adds to the point that you cannot just use an ordinary LLM for that. So if this predictive layer is built on the data that's already there and that has obviously been structured and improved and makes sense. And we're also using the term readiness. So there's a world where these two terms or these two ideas need to meet. How do we know when we have enough validated patterns or information that we can actually start to trust the predictive layer?
- [14:30]
B
Yeah, that's a great question. I'd basically take a step back. I think that this is a sequencing issue and a foundation intelligence issue, right? So let's focus on two separate things and it'll come together, right? So first is, we talked enough about the foundation, right? We need to have a data foundation based on the aim that you want to deploy and the outcome that you want for your business, right? Let's talk about sequencing for a second. Right? I think, I think for predictive to work and I'm talking about it in complex environments, right? I'm talking about CT scanners and elevators and data center equipment and optical network switches and stuff like that, right? Where we depend on these things up and running for us to live our lives properly. And if you think about it, if you are a company that has got thousands of products and hundreds of thousands of issues for a second, imagine that you have 5 million cases and work orders. That's your universe. You are the VP of global service and you have 5 million cases. But those 5 million cases translate to 30,000 repeating issues or 5,000 repeating issues. What does it mean? And I asked this question to a lot of different VPs. People have P and L owners and like, what is your universe of recurring issues? And very few know the answer, right? That is foundational. If I don't even know exactly how many issues and how are they getting sold? In my current cases, my universe, like, hey, my thousand products have 40,000 issues. 40,000 is a lot, but at least I know exactly issue number one happens 8,000 times issue number two happens 6,000 times my universe. If I don't have that foundational understanding of how things are, what are my issues and how are they fixed? How in the world am I going to get to predicting, hey, after this particular issue, this is the nearest neighbor of the issue that is going to happen next. It needs to sit on the foundation of foundational understanding of resolutions or resolution platform that we call it. You can call it something different, but it's a sequencing thing. First, get to a fantastic understanding of your issues and the resolutions. At any given point in time, you should get break fix done. If you are at a point where all issues that come in go into fantastic resolutions. Now you have the right to go in and say, I'm going to go into the predictive preventative mode because that comes with its own data foundation. So let's say you got the resolutions foundation down right on the predictive side. Again, keeping the end in mind, what you want to go and do is am I trying to predict my next likely error so that I can now do preventative maintenance? And that's my aim. I'm trying to reduce my truck rolls, I'm trying to keep my equipment up and running at my customers. If that is the aim that is different from proactively monitoring all of your assets to say, hey, something is going to go down in the next 14 to 20 days and so on and so forth. So understanding that different. And here is where the second foundation comes into picture. Because the world is divided between connected devices and non connected devices. Things that you can monitor all the time and things that you cannot monitor all the time. The good news is both of them, in both scenarios, you know this particular problem has been solved, right? But even within companies, you will see I have these 10 products that are connected. These others are legacy, they are not connected. And I can't tell you at any given point in time what am I going to go fix, right? I'm starting today. I got five work orders, a couple of them are connected devices. Tomorrow I may be working on non connected devices. I still need to give predictive service to my customers across this entire universe. That is where the second foundation comes into the picture. You want to know and you want to have a very clean data set of what we call causal discovery. You should be able to go in and look at, here are my assets, here are the issues that happen. Here is how which factors impact these issues in my environment. And everything from configurations to environment issues and so on, so forth, all need to be put in place for this causal discovery to work. If you have these two foundations, you have the whole resolution platform and you have a very good understanding of issues and resolutions and you have in place the Predictive Data Foundation. Now you have what it takes, the air traffic controller view of your business and then in real time, I'm working on a particular asset based on this asset history and behavior. I can dip into the air traffic control view and not voila. I basically have predictions happening at scale that are very impactful.
- [19:26]
A
These sound like so many things that are super essential and you would think that VPs would just by default want to know this information. So it's scary that you still need to ask them about it. But I can already hear the conversation kind of deflecting when some, when a leader tells you no. But we don't have the luxury of time. We cannot wait until the foundation is right. The board wants AI ROI this year. So how do you respond to a leader that cannot afford to do this in the right order due to time constraints?
- [19:58]
B
Yeah, I think I'm saying that the foundation needs to be in place. I'm not saying that the foundation is going to take a lot of time. Those are two separate things, Right?
- [20:08]
A
Okay.
- [20:09]
B
We are in that world of AI and technology where this foundation can be created incredibly quickly, which is the fantastic news out there. Companies like Neuron7 have what we call pipelines in place. The pipeline is in place. It has got a whole bunch of steps. It uses a boatload of AI models and LLMs as and when needed to go, take RAW log files, take RAW data and the output in matter of days is things that you want to achieve. Now my pipeline is going to be very focused based on the deep domain use cases that I am trying to deliver for you. I may not have every single thing out there. I may not be for example, in the supply chain world which impacts parts and so there's somebody else as a vendor. And this is how the best of breed is going to win in the future is how I think, think about it, right? There's going to be agents and pipelines and products that are in a particular domain doing something like Neuron seven is focused incredibly on service and that too on industries that have complex service. Right. We don't belong in retail and fin services and stuff because that's not where our pipeline and our domain is fixed. But we have thought through every single point of this, right? Yolandi? So for us in terms of the issues and the use cases we serve, everything from how do I create AI readiness and quickly. How do I reduce the amount of time your SME spent? How quickly can I give you governance to approve something that came out of my pipeline? How can I leverage imperfect manuals? We have thought through every single friction point. How do I deploy my AI? Very easily in your Salesforce and ServiceNow environments into the system of record. Right. I think we have thought through everything there is to think through and of course, more our customers are giving us even more. But I think the speed to value is incredible compared to what we saw three to four years ago.
- [22:18]
A
Well, that's definitely something to think of. And as we dive into our last question, I'm going to just hammer on that a bit, but try and see it from the leader's perspective. So I know that the person listening to this is most likely sitting with a budget conversation happening soon. And like I mentioned, the board wants AI ROI in this financial year where we have the team that wants to do everything in the right order and somewhere in the middle the decision actually has to get made by someone that does not really know much about this. So how should a CEO listening to this think about the order of getting ready without skipping the foundation, but without standing still either?
- [22:56]
B
Yeah, that's a great question. I think knowing everything that I know, if I put myself in their shoes, I'll do two or three things in parallel asap, right. I'd first go in and ask the vendors you're working with or your industry peers. I'll do some benchmarking, like has this been sold in my industry or alike Industry, right. High tech devices and med devices very similar to each other. So I can still get, I can still get benchmarking done, right. So I would go do that second. I'd basically take my core team both on the IT and the business side, the SME, so to speak, that are going to drive my outcomes and get them some AI ready education. Right? From a data perspective, from what's possible out there, what are the big risks and failures? I do those two things which are like softer and then I'd go in and work on the data foundation. And with, with companies that can deliver this very, very quickly, right. There is no point in going and experimenting with like, I know there's a lot of POCs going on, but Enterprise World is reference based, right. If your vendor or whoever you're working with cannot showcase exactly what they've delivered somewhere else, I would not waste time with them. I'd absolutely go in and get to somebody who's got the data foundation that can get to the pipelines and stuff very quickly. But between these three, benchmarking against somebody in the industry, providing education to my team and then getting my data foundation very quickly so I can see it, do it for a couple of products, do it for the products that are the most valuable and impactful and something that is the most complex. Right. But most of our customers are now going for impact. They're not being risk averse, dabbling with the easy button. They want impact because they're under pressure from CEOs, are pressure from the board. CEOs put pressure on the service department. Right. And they want to deliver impact very quickly. And I think that is where working on these three in parallel is the advice I would give.
- [25:01]
A
I think that's good advice and it's practical, it's easy to do. It's not that we need to go and Google a bunch of new terms to try and figure out how to implement this. It's been great. I think my key takeaways from our conversation today, if we had to break it down into problem solution and action plan, is that first the problem is that the gap isn't understanding AI, it's whether the foundation underneath actually means something. That was our problem. And then we looked at solutions and we said, okay, being proactive and having that predictive layer, that is what we need to make sure that there's a deterministic base feeding our AI. And then what you just mentioned, our action plan, is first benchmarking, which is such a common thing for a leader to do. So you have to implement it in your AI strategies as well. Getting your staff AI ready with education, and then really building that foundation as quickly as possible. Is that right? Did I get it right?
- [25:59]
B
Yeah, you got it absolutely right. And I think the intelligence layer, start small, see the results, but then go very quickly. Once you see the first two results. Right. The deterministic foundation is huge. The companies that we work with, getting from what we call it 0 to 65% is very easy. The POCs will work left, right and center. It's very easy today, with today's technology to get at least to 65% accuracy. Getting from 65 to 95%, that is where the rubber meets the road. But for our customers, If I give 85% versus 95%, that 10% is millions of dollars, right?
- [26:35]
A
Absolutely.
- [26:35]
B
For an MRI machine, getting that 10% right could be the difference between machine uptime and downtime and contract renewals and stuff like that. And I think that is why the deterministic foundation is something that they should keep an eye out for.
- [26:50]
A
That is a great thought to leave our audience with today. Nikan, it's been a pleasure speaking to you. Thank you so much for your insights.
- [26:57]
B
Thank you so much.
- [27:00]
A
Wrapping up today's episode, I think the three key takeaways from our conversation with Nick and Patel first, executives should focus AI initiatives on resolving complex technical issues that drive substantial financial impact, rather than setting for minor productivity improvements. Second, the success of any implementation depends on building an intelligence layer that transforms fragmented and incomplete enterprise data into a structured format ready for automated decision making. And finally, organizations must establish a foundational understanding of their recurring issues and resolution paths before they can successfully transition to predictive or preventative service models. Position your brand alongside the Fortune 500 leaders defining the enterprise AI roadmap for the opportunity to showcase your solution to the executives currently funding and scaling global initiatives. Partner with Emerge to reach the decision makers holding the strategic strategic mandate. Secure your partnership@go emerge.com partner. That's go emj.com p a n e R for further executive level analysis and to join our network of leaders delivering workflow impact with AI, visit emerge.com on behalf of the team at Emerge. We'll see you on the next episode.