Ep. 22 – Unlocking Hidden Value in Energy with AI.

In this episode of The Elusive Consumer, host Ellie Tehrani sits down with Akash Sharma, Director of Product Innovation at Enverus, to explore how AI is transforming the traditionally conservative energy sector. From implementing generative AI in high-stakes environments to uncovering unstated customer needs, Akash shares insights on product development strategies that balance innovation with human expertise, revealing how energy companies are leveraging vast amounts of untapped data to turn hindsight into foresight.

Transcript

00:11
Ellie Tehrani
Welcome to the Elusive Consumer. Today we’re joined by Akash Sharma, Director of product innovation and management at Enverus, a leading SaaS company serving over 85% of the North American energy market. With over 8 years of experience in energy tech, Aakash leads Andera’s in-house Innovation division, a company with a mission to turn hindsight into foresight for the energy sector and help organisations discover unseen opportunities. At the Elusive Consumer. We’re all about data driven decision making to create better products and services. So we’re very interested to hear what you believe this means for the future of energy tech. So without further ado, Akash, you have transitioned from petroleum engineering to product innovation. How do you translate technical expertise into commercially viable products that resonate with both engineers and executives?

01:24
Akash Sharma
Absolutely. First of all, thank you Ellie for having me on the show. It’s a pleasure talking to you from a standpoint of bringing technical expertise into product management. So when I first started my career, product management wasn’t necessarily like a goal in mind. I started my career as analyst and I was very focused on sort of solving the particular analytical problem that I was working with at any given point. And then over the years, throughout my career, I was able to solve certain problems and challenges and the size and scope of the problems that I started taking on got bigger and it just got to a point where the scope goes beyond what analyst can do essentially. Right. So that was my organic foray into sort of product management from a standpoint of leveraging my technical expertise into creating solutions for our customers.

02:21
Akash Sharma
I think it definitely helps by having that industry context. Right. Because a lot of the times when a consumer or a user provides you feedback on what their use cases and what they’re looking for, it is always good practice to ask the why and understand the real use case, not what the user is just communicating. And being from the similar or same industry, it just makes it easier for me to get to that why? There’s a lot of jargon and subtext that you just understand being part of the same technical background. And I think that has been helpful initially at least when I started just, you know, starting out the second step instead of the first one in the product management journey.

03:01
Ellie Tehrani
That makes sense and we’re going to return to that. The unstated needs of the consumers which is sometimes quite difficult and challenge challenging to get to. But before that I want to continue on specifically within the energy sector. The industry is known for long cycles. Right. And regulatory shifts. How do you approach decision Making and product development when some of the external variables are outside of your control and highly unpredictable.

03:34
Akash Sharma
No, absolutely. I think like. So that to me is also one of the most exciting things about the energy industry. There’s so many different variables and perspectives that go into it. The energy industry fundamentally changed in North America in the 2005-2010 range. We went from traditional offshore and onshore development assets, which take years and years to develop, to the shale revolution and unconventional development. And so what that basically did was it shortened the cycle time of development quite significantly. Like companies were able to turn around projects in weeks, not months. And that also changed the entire decision framework to go from a lot of different planning and stages of decision making into a lot of smaller decisions strung together. Right.

04:25
Akash Sharma
So, as you think about being a product professional working in an industry like that, creating your solutions in a way that helps customers approach their problem-solving at these different stages is very important. And then from a standpoint of geopolitical and just overall uncertainty, given that our industry goes through cycles, I think it goes back to the root of developing valuable products is if you find the right use case, then that product will always flourish. Right. A need to have product is much more resilient to market conditions than a nice to have product. So we really emphasise what is helping our customer do their job more effectively. It becomes a requirement for them to be valuable and successful at their job.

05:14
Akash Sharma
And that further aligns with the fact that as organisations go through these periods of ups and downs, they want to be a lot more lean and resilient. And so our products also help us help them sort of achieve that end goal.

05:27
Ellie Tehrani
You talk about your customers needs and helping them achieve that, their missions, but you’re working with many different industry segments within energy. How has your experience been influencing your product development philosophy as in working across these segments?

05:47
Akash Sharma
I think I’m going to answer something that is going to be the contrary to my first answer. But I think you have to start with customer empathy. Right. It’s very common for you to approach any new project or any new solution with a lot of biases and perspectives. And it’s great. Having experience and things like that is great. But you need to understand that each customer has a unique problem, a unique challenge. And there is definitely a balancing act between every customer believing their problem is just so unique to themselves. And when you talk to enough of them, you realise there are some commonalities. So you have to find sort of the right degree of granularity when you’re trying to tackle these challenges.

06:33
Akash Sharma
So when we work with a particular, within the energy industry, operators are fundamentally different than transportation, and midstream companies are fundamentally different than oilfield services companies, who are in a completely different industry than renewables and power markets and financial firms that power these different developments. So when we look at a particular product and we are designing products, oftentimes we will design them with common Personas in mind. So what is a particular Persona doing in these different segments? Does a common Persona exist across the board, or do we look at it from a specific application where these different organisations are interacting with each other? So as an example of one of our product solutions is our business automation division, which is a source-to-pay solution.

07:26
Akash Sharma
So a big part of that whole product portfolio is not just making the invoicing and, you know, ticketing application for customers easier, but also realising that within our ecosystem of products, we’ve got customers using different products, interacting with each other, so that sort of community effect becomes very important. So a long story short, I’d say, you know, we look at, even though we work in the energy space, there’s a strategic alignment, but we look at every use case and every product as its own specific case study and try to solve it for itself, just keeping the overall strategic framework in mind.

08:03
Ellie Tehrani
And you mentioned an interesting aspect there which is critical when forming your product innovation and how to better service your customers. You mentioned Personas and you mentioned Personas in terms of the different sectors as sort of segments that you serve. But what about the Personas in the sense of departments? So, so you’re serving every type of stakeholder from field engineers to C suite execs. How do you ensure that their needs align across the board?

08:38
Akash Sharma
Yeah, that’s a really big challenge. And I think, like, for. I think one of the biggest ways that we sort of handle that solution is having frameworks that sort of interact with each other, like the fabric of information that you’re providing and support that you’re providing. We interact with each other, but not try to create a one size fits all solution. Because the C suite member of a Fortune 500 energy company and a operations manager and an asset manager and a field engineer may all be leveraging information from the same data set, let’s say, but the kind of information they need is vastly different. Right. So trying to create a solution that kind of works for everyone is not really beneficial.

09:31
Akash Sharma
Having said that, because our customer base is also very engineering geology, like, very scientifically focused, we make sure that our solutions are mutually compatible at different levels. So as an example, if an executive is using our a powered system or our intelligence research to just get high level perspectives on things, the information that we are providing in the research should be able to back calculated all the way down to the raw data so that if the engineer wants to validate some specific numbers, he can very easily do that. Right. So the overall paradigm has to be mutually agreeable with each other. But we don’t necessarily want to design systems that or solutions or products. That is a one size fits all right. Because it’s just not going to provide that kind of value for everyone.

10:20
Ellie Tehrani
And can you share some examples of how some of the feedback that you’ve gained over the years from your customers have helped shape a product? Do you have any real life examples of that?

10:33
Akash Sharma
Oh, absolutely, yeah. There’s so many different things that we have leveraged and taken advantage of customer feedback on. So I’ll give you a recent one as an example. So about six months ago we launched this product called InstantAnalyst. Instant Analyst is our generative AI integration into our existing solution called the Invarius Intelligence Hub. So Inverse Intelligence Hub, simply put, is a repository where our company has our own SEC exempt research shop and they write research and investment advice for the energy sector. They’ve been doing it for 25 years. And we started getting a lot of customer feedback on that product which was, hey, this is great, really valuable, but you guys put out like 30 articles a day. Nobody, I mean I have no bandwidth to read this. Right.

11:26
Akash Sharma
And the other part is at 20, 30 articles and research reports a day, that information accumulates significantly over time. So in some way of speaking, the more information you put out, the harder it gets to find some information you read a while ago. And so as we were trying to figure out different ways of, you know, I think the team had come up with 5, 6 different solutions on how to solve that problem and try to figure that out. But we wanted to again, going back to that whole customer-centricity mindset, went back to our users, and we found that the ability to find and extract the relevant pieces of information was consistently challenging across a lot of our Personas. I won’t say all of them, but a lot of our Personas. Right.

12:10
Akash Sharma
And with that in mind, I think like a lot of the functionality that were also building required a lot of user experience, training, things like that. And people reading our research usually sit within the upper tier of decision-making. So as it is, are short on time. You have this sort of customer problem coming in. And then on the other end of the spectrum, this is around the same time as Generative AI was becoming a bigger and bigger deal. Right. So we looked at the potential application of that, look at this customer feedback, and we identified that this use case actually makes perfect sense for each other. Right. Unstructured data extraction, summarisation, and evaluation is what Genai models. It’s their bread and butter.

12:54
Akash Sharma
So were able to bring those two worlds together and be able to create a solution that came purely out of customer feedback. Right. This is not something that we had sort of on our roadmap, even like six months before that. There’s just something, a confluence of the right technology and the right customer feedback. And were able to put those things together. And you know, in the last few months since its launch, we’ve seen almost 40, 50,000 prompts that have been processed through that system. So it’s been one of the most actively used additions to that framework.

13:24
Ellie Tehrani
Right. And that covers the question about feedback, direct feedback. But going back to our earlier point in terms of unstated needs, sometimes customers tell you upfront what they need or they ask for incremental improvements. But how do you uncover the unstated needs that sometimes lead to these breakthrough products and services?

13:47
Akash Sharma
Yeah, that’s a really good point. And that’s not necessarily as easy to do because a lot of the times different people have different communication ways and so the way they communicate may be misunderstood by your product person. So we definitely do a lot of the standard, like go do regular customer interviews, just check in. A lot of times we’ll create a customer facing version of our roadmap and communicate that with them even before we start building things into it. Sort of bringing your customer into the product planning roundtable, I think is a big part of it. The other part is also just getting a better understanding of what the customer’s job is. Right.

14:30
Akash Sharma
Like there is this idea, like when you look at even within our product portfolio, I think you mentioned, like we have field engineers, the supply chain people do executives, they all work things differently, they all approach things differently. The mindset to look at them is different. So a lot of the hassle come from like building a good equation with customers truly understanding what their job is, what their success criteria is, and then trying to solve, you know, those specific challenges that they mention. The other part that I think also really helps in is like in those customer interviews, continue to ask the why, like why a customer wants a certain feature and you know why it’s helping them until you actually get down to the very base level, you know, because it helps me with procurement, let’s say, right.

15:22
Akash Sharma
The base level use case that they’re trying to build, I think those things definitely help. And then I think the third thing of understanding unsaid customer needs is having good practice of prototypes and wireframes. We have our entire internal incubation division. A big part of their job is to take a majority of customer ideas and bring them to life at the fastest and lowest cost option possible. Because the amount of feedback, a lot of people are very visual. So the amount of feedback you can get from when you’re like, oh, is this what you were trying to ask for? You get a lot more feedback than like, oh no. Actually what I was trying to do was this. Now I understand how I may have communicated a different way. Right. So I think we’ve got a.

16:09
Akash Sharma
When we implement products and from that mindset I’ve always had a very sort of rapid prototyping and rapid wireframing philosophy. And then not getting too attached with product ideas when you come up with them, right. Like you share something, you may think it’s awesome, but remember, you’re not the user, you’re not the customer. So having that sort of apathy towards your ideas and getting them validated and tested rapidly really helps uncover a lot of these needs.

16:35
Ellie Tehrani
Right. And how do you then, you know, in this race to rapidly come up with new innovative services, how do you equally maintain those human-centric, customer-centric, strong relationships with the end user?

16:54
Akash Sharma
I think it’s the challenge with that becomes is really distributing the load of that relationship, right? So as an example, with every customer that you work with, you’ll have some promoters, you’ll have some people who raise their hands say like, oh yeah, I’ll participate in this beta. But if that same user participates in 11 betas in a year, that’s the last year they’ll ever reply to your email, right? So you have to maintain some balance with that and build some those relationships sort of over time. The other thing I also say is like I never conduct a beta with a customer if using the beta negatively impacts their time. So even if it’s a beta product, it is something that should at the very least be not adding additional work for them, right.

17:41
Akash Sharma
So it should bring some value for them, even if that value is going to them for free. That is the sort of value exchange we are getting for feedback for them. So I think sort of continuing to build that Sort of bigger ecosystem and then making sure that you have relationships and partnerships with different Personas. So when there is a product about a certain use case, I will go and find people that specifically can benefit from it. So I think that makes sure that, you know, you’re not dipping into the same well over and over again. And that the second thing would be making sure that even your beta programs are adding some value. Right. These are not your employees. They’re not going to just do testing for the sake of testing. So it has to add value to their day job as well.

18:20
Ellie Tehrani
Right. And as you gather this insight and as you run these tests, how do you ultimately decide which of the new features or technologies you’re going to invest in?

18:32
Akash Sharma
Yeah, so that becomes like, once we get feedback from customers, things like that become a fun prioritisation exercise. As I’m sure you know, a lot of product managers love prioritisation. But what we try to do is we try to create equitable KPIs before we launch a beta program. So we identify we are not doing a beta program because, you know, Akash thought this was a cool idea. So let’s release it and see what customers think. When we do a market test or a beta program, we are very clear that there has to be a reason. Right. There has to be a hypothesis you’re testing. And there is degrees of that hypothesis testing, response that evaluate how good or bad your idea is. Right. So a lot of that is going to be usage driven, feedback driven and things of that.

19:19
Akash Sharma
And then once we have that sort of a rating on market test performance, then where it sits in the central product roadmap is a decision of customer feedback plus roi. Right. If there is a product that can be built out in a certain amount of time that has this potential addressable market attached to it, then we sort of look at both of those factors and make a decision where this sits within the pipeline.

19:43
Ellie Tehrani
Final question on the innovation piece, during product development, typically the energy sector overall has been seen as a little bit slower to adapt new technologies. If you compare it to healthcare, finance, et cetera, how do you balance that traditional nature of the overall sector with ensuring that you innovate?

20:09
Akash Sharma
Yeah, no, absolutely. I think you have to meet your customers where they are. Right. So if on a hypothetical innovation scale, my customers are at a six out of 10, I’m not going to bring them a 10 solution, I’m going to bring them like a seven solution. Right. Again, going back to the value, your customers, if you don’t meet your Customers where they are, you’re not going to get adoption. It has to be a product that fits with their current mental model of how this problem is solved while at the same time making their lives easier. Right. And you can design your roadmap such that you get them to a 10 solution, but it has to start a lot closer than that.

20:47
Akash Sharma
The other thing that is also sort of just from context standpoint is the energy industry is also very capital and risk intensive. There is hesitancy to take on projects and technologies that maybe a team does not have in house expertise on because we have a much more higher impact of errors outside of let’s say financial losses as well. Right. Like a model that is, let’s say an AI model that’s deployed to adjust, well, pressures and things of that nature. That thing failing is not just going to cost us lost revenue in oil, it could lead to an explosion. So the safety protocols are a lot more rigid and we have to be a lot more careful with it.

21:32
Akash Sharma
The other thing about the energy industry though, if you think about the industry as a whole, it is probably the only other industry outside of technology itself that has grown at the pace that it has, like crude was from the discovery of crude to standard production to renewable markets to this energy mix. This is not a thousands of year story. Right. The world has become dependent on energy markets, including fossil fuel as well as renewables, much faster than these systems have had a chance to innovate and develop and keep up with it. So I think in a combination of things like that, we are very cognizant of that aspect and then we deploy solutions and technologies that help customers in again executing their day job more effectively rather than complete fundamentally different mental models to solve those same problems.

22:32
Ellie Tehrani
Right. And as you’re implementing some of these new technologies, including generative AI, what are some of the unique challenges that you’ve come across?

22:42
Akash Sharma
I think with generative AI implementing that challenges? Let me rephrase that. So I think as we think about implementing generative AI and energy, I think the biggest challenges that we faced were challenges that generative AI application and production are faced by almost every industry. This technology is effectively, if that two and a half years old. So subject matter experts are also relatively new in this space. And when we started building our first solution, there wasn’t a standard go-to-market on how to take Gen products to market. Right. So there’s a lot of very steep learning curve in deploying this. The very steep learning curve in getting, you know, getting all the technology frameworks put Together, but at the same time, you know, working in a customer base where our consumers are effectively mostly engineers, geologists and people with that mindset.

23:36
Akash Sharma
Not a big fan of black box AI models. Right. So building an explainable framework where they understand the cause and effect relationship with how they are interacting with the system and what response they’re getting I think has been a pretty big challenge. Working with that. I think the other part of it that has been really important in communicating is how data is being handled by these language models. There’s a lot of concern about privacy and data security and language models in the current state of development can handle your information one of two ways. One is the in context learning and the other is fine tuning. Whereas previously most AIML models were trained and fine tuned essentially, LLMs don’t necessarily require that. Now this may seem like a very specific and nuanced data science point, but it fundamentally changes how customers believe in this product.

24:31
Akash Sharma
Right. It’s hard to communicate that if you put your data in an in context model, your data is still secure. And I think that has, that’s definitely, you know, that’s definitely a challenge to effectively communicate because generative AI and you know, complex AI models as such are very jargon-heavy. And so for an end user it becomes really hard to understand things. So I think explainability, transparency and just data security concerns have been some of the biggest challenges from a customer standpoint. And then for us, honestly just getting our own team upskilled and able to build products at the scale and with.

25:08
Ellie Tehrani
This technology now, how do you get that buy in from customers who need to see or typically like to sort of dig beneath the surface and get more raw data like the engineers, how do you convince them? You mentioned transparency and sharing, more knowledge sharing, but how do you overcome that challenge?

25:32
Akash Sharma
That’s a really good point. I think like it’s not something that’s solved overnight. Right. I can tell you this, that we have so many customers who’ve come in there and have had trouble understanding how the model works. And we work through repetitive education frameworks. You’ve got LND frameworks and things like that to help them train and get up to speed on it. But on the other end, on the transparency side, we provide full visibility into what our platform is doing at any given time. So I’ll give you an example. I’m sure you’ve seen the deep seq or the 0103, all of these reasoning models. The way chain of thought works is as you provide feedback, the model goes through a chain of thought process. These organisations showcase most of the chain of thought. So you can actually see how the model is thinking.

26:22
Akash Sharma
That makes you have more confidence, right? Because now if you’re so, if you’re following a logic track that makes sense, then you will believe the model’s answer as well, right? Because if you’re thinking in the right way, or in another example that I’m sure a lot of your listeners may be familiar with, is many times when you’re doing interviews for companies, especially earlier in your career, they’ll ask you a, you know, a logic problem and they’d want you to solve it, but they’ll also explain what you’re thinking, how you’re trying to do it. So that helps the listener get more insight into your thinking process. So we follow the same thing. We’ve got a product that we’re working on which is called our Prism Explorer.

27:04
Akash Sharma
And essentially it is a generative AI system that integrates with our visualisation and energy database of thousands of tables in it. So if I ask a question and if I say, let’s say how many rigs is Exxon running in the Permiand I just get answer, as an engineer myself, I’ll be like, I don’t know if I trust that number, right? Because I’m like, all right, this just came up with a random number. I want to see evidence. So the way we do that, it’s also hard to give evidence when you have a database that you source the answer from the evidence is like, yeah, sure, it’s in that table. So one of the ways we do that is we actually have a chain of thought model.

27:41
Akash Sharma
And not only do we have the chain of thought model, we have a multi-agent system, and everything the agent is doing at every step of the process is accessible and exposed to the user. So if a user wants to actually see how they came down to, you know, the answer, let’s say if it’s 25 rigs, like how did the model come up with 25 rigs? You could check step by step what it did, why it did, how it ran the equation and gain a lot more confidence into it, right? And what we’ve seen is users, when they first start using that product are always looking through the details, right? Because they don’t trust the model to do that.

28:16
Akash Sharma
But once you do it a few times, you gain some more confidence in the model, then you only look at it when it sort of doesn’t pass your sniff test, right? So just having more transparency, making sure there’s ample training available creating solutions that is closer to their customers, existing workflows. I think a combination of these three doesn’t necessarily solve the problem, but helps working us towards the place where it can start to be solved.

28:41
Ellie Tehrani
Absolutely. Especially in an industry like you said, where the wrong decision could have a catastrophic impact.

28:48
Akash Sharma
Absolutely.

28:49
Ellie Tehrani
So a question that I typically ask product innovators and developers across different industries is the, you know, what’s on everyone’s mind at the moment? How do you balance automation with maintaining human expertise? How would you answer that for the energy tech sector?

29:08
Akash Sharma
I think managing there is no framework that you can build without human expertise. Right. I think as you think about it, like AGI, which is like the best version or the ultimate singularity moment of this is for these models to get to human intelligence. Right. I think that’s an important frame of reference to keep in mind. These are subhuman intelligence. So as we think about automation and implementing these within our workflows, automation is a place for me where I can replace human tasks by automated tasks that are repeatable, have low cognitive decision load, and because of which are actually prone to human error. Right. So if those type of tasks can be automated and put together, it makes a lot of sense to actually automate those processes. Right.

29:59
Akash Sharma
So even if we build a solution that connects, let’s say a data extractor to a chart builder to whatever. I think what you have to keep in mind is that where in this framework and where in this workflow are the critical decision points, and those critical decision points have to remain human. Right. Because you can automate the rest of the process. But that last decision where that automation has a true impact on something physical, I think you need to have sort of a human in the loop aspect to it. And the advantage that AI has given on top of existing RPA automation systems is that I think one of the big ones is that you don’t have to create a case statement or an if statement for every possible scenario.

30:43
Akash Sharma
You don’t have to deterministically create 500 scenarios that could happen at a decision point. AI can handle some more uncertainty with that. Right. So there are definitely pros and cons with it. But without human in the loop, the models are prone to error and hallucinations. And it’s a very dangerous gambit handing over all those decisions, a cognitive load to an AI model. I think the blend is where it works.

31:10
Ellie Tehrani
And as we look ahead in the future, and specifically to the energy industry, what other emerging trends do you see shaping the Future of this sector.

31:22
Akash Sharma
Yeah, it’s a little bit of a harder thing to project because we truly believe this is the tip of the iceberg. Right. So scoping what this means overall is really difficult. Having said that, I think generative AI is going to fundamentally transform especially the knowledge management side of our industry. We have so much data. The energy industry collects petabytes of data on a daily basis. That doesn’t even include all the unstructured enterprise data that sits into people’s folders or network shared drives. And the challenge that is thousands and thousands of PDFs sitting there. There’s so much information that’s out there that is either just left to its own devices or some people know a little bit of it and then a very small portion of it. You have been capitalised and utilised for data-driven decisions.

32:18
Akash Sharma
So I think the energy industry specifically is poised incredibly well to take advantage of transformers and generative AI because I think it offers an ability with pattern recognition models to truly unlock and make data driven decisions and that can help us de risk a lot of these decisions moving forward. A lot of the easily available energy resources have been taken as the, and then the structurally more complex and engineering more complex ones require more cognitive load, let’s put it that way. Right. So I think AI is positioned to help us through that transition and make sure that we remain efficient even through that quite well.

33:02
Akash Sharma
And with all of these things coming and all of this data being driven forward, especially with the energy industry, I think it is a very interesting power dynamic, but the electricity power demand side of this whole equation is completely independent of what we’ve talked about so far. It’s very interesting. Right, because this is the first time in decades that the US power forecast is changing because of data centre demand. So there is no generative AI without energy. And with more generative AI, you could get more efficient energy. So it’s like a positive feedback loop that hopefully can sort of help us support and extract more energy more efficiently and more sustainably and more effectively across the board. So I think there’s a lot of different vectors, especially in the energy and air relationship, and we are just getting started on that.

33:57
Akash Sharma
There’s a lot more that’s left to be seen.

33:59
Ellie Tehrani
You mentioned data centres, and it made me think of the Stargate project that was announced. I know. What are your thoughts on that in terms of how that will impact the overall energy tech sector?

34:15
Akash Sharma
Yeah, I think more investment in that space is definitely going to help. There are a lot of use cases that Sort of fall below the line right now. Right. What I mean by that is I don’t know if this is valuable enough. Should I allocate time and resources and money into it? Because if were to build that, I would need GPU capacity and clusters are not easy to come by. But with investment like this, the target project and you know, not just that, but so many like Microsoft, Meta, AWS, they’re all building data centre after data centre. All that does for us as users of this technology and implementers of this technology is make it much more readily and easily available. And the ease of availability of technology is only going to help in accelerating the innovation that this thing can bring.

35:04
Akash Sharma
So I’m really excited about that.

35:06
Ellie Tehrani
I guess that’s the beauty of technology as it grows at record speed is that accessibility, like you said, it provides so many more people, consumers and organizations access that they would not have had access to in previous years.

35:25
Akash Sharma
And affordability, you know what I mean? You don’t have to make a dent into your corporate budget if you want to explore this technology anymore. Right. Three years ago, inference costs were like 98% more than they are today. It is fundamentally changing how people can use and test this technology.

35:43
Ellie Tehrani
And how is this transition affecting your product development strategy?

35:48
Akash Sharma
So my team is responsible for the sort of our generative AI strategic roadmap within our company and so we’re leading a lot of those initiatives. I think a lot of the ways we are thinking about this is we are rethinking things on our roadmap. Right. So we are looking at it from a lens of is this a solution that can be enhanced with a generative AI add on or a generative AI solution. Right. And so I think re looking at those problems from that perspective is definitely helpful because we also believe that Genai for the sake of gen AI does not make sense. You’ll end up with a solution that’s, you know, flashy but doesn’t really bring customer value.

36:29
Akash Sharma
So I think we’ve been doing, and we’ve done almost like a review of all the things on our roadmap, see how they can be improved and enhanced. And it’s really exciting because not only does it help us get some of these things out faster and more effectively, in some cases it is requiring us to fundamentally change how we think about a problem statement. There are things that these models and technologies can do that were rocket science. Right. Like a while ago. Like, it’s just we didn’t even think some of those things would be feasible. So I think that has really changed our product roadmap. And then given the fact that when ChatGPT released this thing in September 2022, almost everybody downloaded it became so prevalent.

37:16
Akash Sharma
The amount of ideas and inbounds we get from customers about like, hey, have you tried doing this or have you tried doing that? That has also been extremely helpful. Our industry and our customer base is really excited and we’ve been partnering with them for decades. So they’ve brought us a lot of different ideas that have, you know, given us a chance to really review our product roadmap and rethink prioritisation when it comes to integrating this technology into that process.

37:43
Ellie Tehrani
I want to talk a little bit about some of the things that you mentioned earlier in the sense of, you know, getting through the skills gap that we’re facing both internally as well as externally. You know, went from a phase where data was the new oil, and now with the endless amount of data, it’s the ability to connect the dots and tell the story. Right? Especially as you mentioned, the higher up you get C suite, et cetera, they just want to be able to get to the summary really fast. So, as the energy tech sector becomes more data-driven, what skills do you see as critical for future industry professionals? And if we are facing a talent gap, how can organisations address it?

38:32
Akash Sharma
If you think about it from a standpoint of how individuals can impact their sort of learning with this technology, I would honestly say just get familiarised with the key terms. If you understand the core behaviour of how an LLM works and why it works, you don’t necessarily have to know how to code up an entire language model. Right? Those are two different skill sets. So from a business leader standpoint, understand what language models are, why they work, what rag is, what are its pros and cons, what agents are and what hallucinations are, why they happen. So these key terminologies and truly understand the sort of logic framework around them I think would be very important because without that knowledge, sometimes you look at a problem in your industry or in your enterprise.

39:25
Akash Sharma
Without this understanding of AI, you won’t be able to connect those dots. Right? Because if you don’t know what the strengths and weaknesses of this technology are, then you don’t know if it works for your use case or not. So I think that would be one big thing. And from a standpoint of solving your internal talent gap, honestly, it’s going to be an aspect of training and investment. Companies have to train and invest their internal resources. My biggest recommendation with that always is that instead of doing a organizational wide training program for weeks where they come in and learn something about it, those have their own places, they have benefits from an introductory level.

40:07
Akash Sharma
But there’s nothing more valuable than actually doing a project from A to Z, like create a small team within your company, come up with a very simple but valuable use case. A low risk, low reward, internal use case. Right. Have a few people in your organisation dedicate some time to building it. So once you build your first product from A to Z, you’ll learn a lot more about. Actually you learn a lot more about what you don’t know. And then you can design your training program around that because there’s no better way of learning than actually trying to doing it.

40:39
Ellie Tehrani
So sort of like an incubation team within your company.

40:42
Akash Sharma
Yeah, I think I may be biased when I give that recommendation.

40:46
Ellie Tehrani
No, but it makes a lot of sense. And final question from my end is, if we look ahead, which can be difficult as things are developing so fast, but if we look ahead 5 to 10 years, what major changes do you anticipate in how energy companies consume and use technology?

41:07
Akash Sharma
The answer that I’m going to give, I’ll give you the caveat that this technology has been around for 18 months and it’s already changed things so much that five to 10 years is a timescale that is getting harder to imagine. Having said that, I think increasing number of companies have already started coming to us and talking to us about how to get their systems and their data sets, AI ready. Right. Their perspective of getting and storing and capturing information for the purposes of use for AI has become a much more strategic priority for a lot of CTOs and technology leaders out there than maybe even a few years ago. So I think that’s going to be a fundamental shift because that will allow a lot more application of this product.

41:49
Akash Sharma
I think the other thing that’s becoming really powerful is given with the cost effectiveness of new models as well as with the emergence of small language models. And as you were mentioning, with the compute capacity becoming cheaper, we are at this very interesting confluence of technology and ideas such that it is ripe for niche custom development. So targeted innovation I think is going to pay huge dividends in that period. So what I mean by that is you have large quantities of compute capacity available for cheap and you have access to small language models in a secure open source environment. You can create a vertical agent for every use case. Right. You can create a vertical use agent that can be responsible for tracking production data. So if it sees anomalies, it automatically does that.

42:40
Akash Sharma
It can be responsible for looking at tracking, you know, solar power outage or like weather tracking in a certain areas. Right. So very specific niche use cases with decent amount of training data packed together in a vertical agent. I think that’s going to change how a lot of tasks are done in the energy space because with a combination of these three things, you can really, as we were talking about earlier, automate some very specific initiatives. So I think the combination of this cheaper compute, small language models, things of that nature, I think is going to be huge for our industry. I think that’s kind of generally true for a lot of very specific industries.

43:23
Akash Sharma
And then with the multimodal and those technologies are becoming more and more reliable and accurate, it’s also going to unlock so much more data than we thought was even not possible now. Right. So went from tabular numbers to text to now we are at an accuracy rate with multimodal models that you can just drop images into it. Right. And what that means, whether it’s, you know, security tracking, safety tracking, processing, image information, video information, the huge applications in the energy space. Right. Like a few ideas come to mind as I talk about it is, you know, methane emissions monitoring, methane plume monitoring, being able to identify those patterns. So I think the industry gets more and more comfortable with this idea.

44:15
Akash Sharma
We are going to see people getting their AI, their data sets, more AI ready, building smaller niche models that fit their perspectives and then just seeing a much more integrated framework where AI is part of existing workflows.

44:31
Ellie Tehrani
And with all this benefit for the organization, the vaster amount of data, the more niche and targeted, do you think those benefits would be passed on to the consumers?

44:42
Akash Sharma
I believe so, absolutely. Because if you require a lot less efficient. Sorry, if you require a lot more efficient framework of people and resources to get specific tasks done, it makes companies a lot more efficient. Right. And you can do a lot more tasks in that manner. I think there’s definitely. If the, if this technology does get integrated in that manner, I think it is obvious to assume that it’s going to fundamentally change labor dynamics across the industry as well. But I think any transformative technology has that impact. Right. Like there is the way we did research before. The Internet is fundamentally different than the way we do research right now. I think it’s going to lead to a lot more new use cases, a lot more new opportunities for people in that regard.

45:36
Akash Sharma
But it is going to fundamentally change how jobs are done.

45:41
Ellie Tehrani
Given enough time is there anything that you would like to mention to our listeners that we haven’t covered today?

45:49
Akash Sharma
No, I think we’ve covered a lot of different things. I think the only thing I can say is if any of your listeners want to have any follow up discussions, want to talk more about this subject, they can find me on social media. And yeah, happy to have this conversation. I live and breathe AI Product implementation. So happy to have a chat whenever I can.

46:13
Ellie Tehrani
Thank you so much Akash. It’s been a pleasure having.

About Our Guest