The power of AI fuels Simfoni Analytics’ award-winning technologies to help our clients make fast, well-informed procurement decisions. Simfoni Analytics’ Head of Product Sreeram Venkitakrishnan recently appeared on “The Procuretech Podcast,” hosted by procurement expert and consultant James Meads to discuss how Simfoni leverages Artificial Intelligence in its procurement technology. Learn how Simfoni’s AI works to make strategic sourcing significantly easier for our spend and procurement analysts.
Podcast Summary on AI-Powered Spend Intelligence
Welcome to the Procuretech podcast, bringing insights and inspiration into how digital technology is shaping our profession. I’m your host, James Meads. Tea drinker, ex-pat. Definitely not your typical consultant. Yes. Hello, greetings, wherever you are. And welcome to another edition of the Procuretech podcast, where every week we bring you news stories and insights into everything, digital procurement from service providers to thought leaders, right the way through to actual case studies of successful transformations. And on today’s show, we’re going to be talking all about spend analysis and anyone that follows me on LinkedIn, one of the things you’ll know is I’m all about understanding your spend. So without any further ado, I’m going to introduce you to today’s guest. He is Sreeram Venkitakrishnan from the company Simfoni who has developed a tool that uses Artificial Intelligence to leverage smart spend analysis. So I’m really looking forward to this conversation today, because I think I’m going to learn a lot more about what this tool can do and how some of the sort of insights into the future of how we may do strategic sourcing and how some of what was traditionally seen as being some of the more strategic end of the spectrum in terms of sourcing skills may be able to leverage AI going forward to make our tasks easier.
So Sreeram welcome to the podcast. First of all, maybe tell us a little bit about yourself and what Simfoni does, and then we’ll jump straight in.
Right, James, thank you so much for having me. So Simfoni my company has got two main products, we are becoming a technology company. So two main products, Spend Analytics, which everybody knows. The second one is something different that we are doing is called spend automation. We have a bunch of different solutions coming together on that side, primarily focusing on mid-market and large companies, helping them with their tail spend, indirect spend, sorting of categories, automating how they purchase those categories and stuff.
Great. Okay. So one of the reasons I was keen to get Simfoni to come on the show is because a lot of what the tool claims to do is what we’ve been accustomed to seeing as being sort of key procurement skills, first strategic sourcing and category management professionals. Whereas a lot of other software that’s out there tends to, to facilitate the process more for transactional and operational buying and easy of use or make that less administrative. So in essence, what you do is a spend management tool, but the signature line, your website is AI-powered spend intelligence. And I wanted to dive into that a little bit because that’s, that’s a really interesting concept. So maybe you can explain to anyone that doesn’t really understand it, the capabilities of AI, and what that can do in the spend management space.
Sure. So the Artificial Intelligence, which we talk about on our website is primarily to spend analysis that you would see. We have some elements AI capabilities on the spend automation side of the business with the major application that we have spend analysis software. So when you talk about Artificial Intelligence, typically AI is a larger umbrella of this intelligence and then under that, you have ML (Machine Learning) and different other areas. So if you take AI is fundamental for anybody, kind of a layman to understand the basics is that Artificial Intelligence helps us predict an outcome based on certain parameters, right? So the typical application of let’s say computer vision, converting an image into some sort of vectors and numbers, and then predicting whether that is a cat or a dog or an apple or an orange. So, that is an AI capability, which allows us to read the image using computer vision and can work.
Similarly, in our context, we are dealing with text data, especially when it comes to classification and normalization of suppliers, we take the data and convert that into input parameters, which then can predict some sort of outcome like categories or supply group names, clearing out the suppliers, similar differences, that sort of stuff. So, we use a number of open-source libraries that can allow us to convert these into input parameters, take on the description, convert those into input parameters, and then understand what categories you should follow through. One example is how do we take a description, which is called a document in the data science world. So how do you take one document, which is a description, and then understand what is living in that description, which we can use for classification or which can be used to predict the classification or category. So for example, if somebody says to purchase laptop bags, then we have a laptop, which could be an item. Laptop bag is a different item, it’s an accessory versus a laptop. So you use AI or natural language processing to understand this text and the context, and then to look under the right category. So, that is one way of using it. And then from an AI context or perspective, that is different applications from our perspective, from a spend analysis perspective, that is one way of using it.
So I’ve spoken about this at length before in terms of poor quality of data can often be a reason why digital transformations can, can fail or not be as successful, as they’re anticipated to be. And I see that one, one of the things that you just explained there is it a little bit around data classification and taxonomy, and, I know an accusation that tends to get thrown around and I sort of maintain a neutral view on this as, as a podcast host, but it’s, that data classification still requires a human touch because AI isn’t smart enough to identify certain anomalies yet. So are we there yet in terms of AI and machine learning, being able to accurately classify PO data into different material groups or, or do you still think that while a tool like Simfoni is very useful and can speed up the process that we still need sort of a hybrid approach and someone with human cognitive capabilities to do a sanity check and fine-tune the analysis at the end?
Yeah, I think that’s a great question. So, AI is no magic bullet or magic brand. It’s not going to kind of solve all the problems and it only assesses the accelerator, to be honest. So we are given, let’s say a one million set of data records of invoice line items, how quickly we can just do it manually versus how quickly we can do it using an AI-driven analytics solution. So in our case, we not only just rely on this parsing of nouns and key-value pairs, we also rely on a rules engine. Anything that we kind of classify this automatically, we capture into our master rules repository, and this master rules repository gets reapplied to the new data. And again, even with that. It’s never a hundred percent. You definitely…it’s a manual effort. So if you, if you can get a good percentage covered, 60%, 70% automated through this AI process, and then somebody will have to sign in to check and approve them, which gets converted into rules still.
There is about 20%, which then requires somebody to manually classify them. That’s the nature of the data, as you said, if the data is so complex, so contextually different from one customer to another customer, even within one business unit to another business unit, they call the same product in different names. It’s just not going to be able to kind of automate completely. So we have inbuilt tools within the system, which allows us to do this automation plus the kind of classification within the tool itself. So we have created an Excel sort of interface, which can allow us to quickly classify and capture. All that was fundamental. AI is an expensive solution for all the problems. So we try to capture all those rules. So the next data comes, we can use the rules first, which is a much quicker, faster, cheaper solution to apply. So it’s definitely hybrid, and I think a hundred percent automatically can only happen if the organizations can keep their data very clean, which is a huge challenge.
Yeah. And I think you, you alluded to it at the very beginning of that answer. There is no silver bullet, but I guess the advantages are that it can take a lot of manual work out for the sort of 60-70% that it’s able to classify. And then that makes the task for the data scientist or the data analyst, I guess much, much simpler than if they have to then go and do a manual classification and then organize everything into a final taxonomy. So just a quick interlude before we move on with the rest of the podcast, just to say that if you are a procurement leader or a finance leader in a manufacturing company and you are struggling to get to grips with your spend, or you just maybe need an extra pair of hands to resolve a specific issue and drive some bottom-line results, just drop me a connection request on LinkedIn, or just ping me an email to [email protected] or just follow the link in the show notes to book a free 30-minute initial call with me.
So as I can learn more about your business and what I can do to help you. So now let’s jump right back into the interview. Another thing that Simfoni claims to do is, is around risk management because especially this now is going to be big business. As we start to see the landscape emerge post Covid-19, to what extent can a machine or artificial intelligence project potential risk in your supply chain and how does it go about analyzing that data? So it does it from existing vendor records and supply patterns, or does it, or does it look on the marketplace and look at things like, you know, geopolitical risk and, and civil unrest and that type of thing.
Yeah. Again, it’s not automation as such, it’s, it’s more of an acceleration of decision making isn’t that? So the kind of news articles or not information, which is not otherwise readily available for the category managers or sourcing managers right in front of them. So great examples of supplier commodity and market news. So within each of these different categories, we have got sources in integrations with third-party providers and other news sources through which we bring in news articles and publish articles, which will provide the category and sourcing managers, some sort of insight with which they can take action. So we could address or try and give our customers are top 20, top 50 suppliers news on a daily basis, which is automated and improved, but the decision has to be done manually. And the category managers will have to read them what interests them and so on, so forth. So it’s not an AI capability. It’s more of providing them with insight or information readily available for them to make decisions.
Right. Okay. So similar to what we said around the data classification, it’s a tool that gives the knowledge to make it available to the procurement category manager, but it’s not something that does the work for them. They’re still going to have to have sort of a human thought process and analysis to be able, to make that call. It just puts all of that information and data that they need to be able to make an informed decision all in one place. And as such, it saves a lot of time and increases the awareness of the different factors that they should be considered when making that decision.
Yeah. And a great example is COVID is happening. We brought in some third-party data sources, we converted them into some sort of scoring, for example, which countries locked down, what sort of factory activities are there, what sort of policies they are putting in place. We brought in a bunch of key markets, key scores, or key parameters and scored them and kind of mapped that numbers against the suppliers and supplier spend for our customers. So then the customers will have to then look at, “Hey, Simfoni’s analytics tells me that these 10 suppliers in these regions are at risk. What do I do with it?” Then they can reach out to the suppliers and clarify it. They are still going to meet their delivery deadlines. Is there any kind of factory shutdowns, so on and so forth. So it’s about giving them a direction in terms of where there is risk and then they can validate it from that much easier.
Yeah. Giving them the visibility tools to ask the right questions and improve the supply chain. Right. So we’ve talked about data classification and taxonomy. We’ve talked about identifying risk management in terms of sort of securing our supply chain. Let’s get now to a topic that’s, you know, usually at the sharp end of what buyers are tasked to do and what we’re typically measured on, which is cost savings and expense reduction. How can AI and machine learning assist procurement departments in terms of being able to find the right costs or the, or the most lucrative cost savings opportunities from the spend analysis software that Simfoni can provide?
Yeah, this is more often automation of higher quality software than an AI capability, but it’s still very powerful because at the end of the day, having done all the analytics, all the customers want to know is where they can reduce costs. So one of the things we do, or one of the borders that we offer our customers is the opportunity assessment module. There’s a standalone dedicated module on that. In that module, customers are given a bunch of opportunities, which is like almost a kind of prescriptive saying, “Hey, these are the categories. These are the opportunities that you guys can potentially work on and reduce costs.” And these opportunities are calculated using a bunch of automated calculations. So we look at about 15 different parameters or 15 different savings levers. Each of them has got its own way of analysis, which otherwise takes a lot of time to manually calculate.
It is still possible, very much possible for somebody to sit down and do it manually, but doing it over a period of three, four weeks and over a period of three, four days. And allowing our customers to validate in an interface much easily and through a PowerPoint or Excel is what we try to do. And then the customers can turn those into projects and work on them. That’s about 15 different opportunity levers or savings levers that we run through the data, things like usual suspects, like supply consolidation, price, variance, payment on rationalizations, all those sort of things, every different angle from which they can reduce cost, improve compliance for the customer.
Does it also feed into things like commodity pricing for, for certain raw materials as well, kind of benchmark indices and look at where maybe you’re paying way above what the market price should be.
Yeah, we call it Cost Analysis. Uh, one of the recent additions in that 15 is this root cost analysis will be connected to different data sources for this. One is a data source, which provides the cost drivers for certain kinds of products or industries. So for example, uh, I buy a lot of plastic, plastic-based items and I can then figure out, “Okay, where am I buying from?” And for that industry, what does the cost look like? How does it look like what does the sales and marketing cost look like? And we kind of tap into a data source, which provides those directional cost breakdown. So he then says, this is a cost. This is the spend I have on this plastic product. and this is how the industry cost driver looks like. And then we also connect to commodity sources for plastic in that market or other commodities in particular markets, which then allows our customers to kind of create a simulation in terms of, Hey, can I reduce by a supplier or sales and marketing by 3%? What sort of profit can or savings I can make, how the commodity market has performed versus my unit price? Is there anything I can leverage making for a very powerful negotiation tool for our customers.
And, and that’s typically one thing that if you were to ask a bunch of different buyers, you know, why do you fail? Or why don’t you get the optimum outcome from your negotiation? I think probably a majority of them would say just lack of time to prepare and lack of availability of the necessary resources to go out and pull that information. A lot of companies have cut back their budgets. So, you know, things like market price analysis, they may have canceled subscriptions, which I personally think is a stupid decision, but big companies do make these very barbarian decisions to cut expenses left, right and center. And I think the other points on this is, and you hinted to it in your answer that even though it may be possible to do this type of analysis manually, and a lot of companies, especially in larger corporations, what you often find is that there is a lot of rotation between different positions.
You know, some companies have a policy of moving people around every three or four years, especially in sort of more senior roles. And if you’ve got of this information and data at your fingertips, then if you move into a new procurement category manager position in the category that you’ve not managed before, whereas, in the past, it may have taken you six to 12 months to really sort of know what you’re doing and be able to make an impact. You know, the demands of the business as such these days are that you need to be able to hit the ground running. And I guess this helps you to get a much faster induction into something that you’re not familiar with. So, you know, the core procurement skills go with you from category to category, but having that market intelligence and understanding of, of your spend your suppliers, your risks and the marketplace is, is really valuable to get that quickly all in one place.
Oh, totally. I think for me negotiation, so that two levels, so typically there is a fact-based negotiation we can look at how do we reduce supplier? How do we look at payment terms and reduce the number of payment terms with which we pay, etc. And the second level of negotiation is the optimization sort of negotiation. So look at the category itself and see beyond these pure cost-based negotiation waters. And we do so far that this is very powerful in terms of providing them the cost drivers and stuff, just to add onto that. It’s a very important point in terms of kind of maintaining knowledge, is that in opportunity assessment, when we provide the customers with categories and where they can attack and stuff, we also give our customers in terms of what sort of, kind of negotiation levers they have within each category. So certain categories have got certain negotiation levers, and it differs from product to product or categories to categories. So we try to provide that intelligence too, so we have been building and it grows over time, a library of negotiation, levers by category, which then the customers can use. So even if somebody does not have a lot of background, let’s say on IT, network equipment, they can’t, they can have some input from us, not everything, but some input from our technology or platform, which they can kind of build off.
With all of these different things that a tool like Simfoni can do. And with the procurement landscape changing. And, you know, you hear a lot on LinkedIn and in the procurement and supply chain press around more focus being on total value rather than just a rather cutthroat approach to price negotiations. And I definitely buy into and agree with all of that. Although I do think obviously we were still going to be measured on savings for a long while to come. So maybe if I can ask you a little bit of a philosophical question, as we start to round up out of these three key pillars that Simfoni offers to its clients, or perhaps you can maybe even think of something else, what do you think will have the most profound impact on procurement performance and strategy over the next sort of five years.
These are essential components. These are not the only ways of doing it. These are essential components of doing a complete or holistic analysis per se, but that is procurement organizations have got different maturity, and organizations with different maturity can pick different solutions. So I look at a platform for procurement in four different areas. Hierarchy, the fundamental, or the base that everybody should have should be the minimum capability of data management and dashboards. So everybody should have visibility ability to manage their own data, get the data that they want when they want all this sort of stuff. And once they have it, the next level is to kind of put any intelligence in terms of Artificial Intelligence or predictive algorithms. So we also do predictive modeling, which tells us when you can expect to buy an item in the next six months, that sort of thing, which again, accelerates the speed at which procurement can perform.
And if they have it, then if they are little bit more mature organization, then there is definitely much more things they can do around going deep-dive analysis around categories, opportunity assessment, procurement, operation, cycle time, that sort of thing for the top global big companies, if they have a great majority and they’ve got nailed the other three, then they should try and leverage a strategic network opportunity. When I say network, it’s about bringing into 20 data sources using the collective intelligence of the platform like benchmarks, so on and so forth. So I see this in our kind of three, four different hierarchy in terms of leveraging analysis, and in terms of where the current maturity of the organization stands, they should be able to identify and go with that. But at the bare minimum, everybody should have a great data management dashboard for analysis process capability or dashboard tool for them to have the visibility. Current ERP, Source-to-Pay, or Procure-to-Pay solutions such as SAP Ariba just don’t compete with a best-of-breed analysis solution. CPO’s know this.
And I completely agree with you. You can’t run before you can walk, and if you don’t even have that basic in place, then there’s no point trying to do something at the very high end with AI because you’ve not got the fundamentals in place. So that’s a, it’s a really interesting answer that you gave. Thank you for that. It’s highly dependent on organizational maturity and data quality in terms of how quickly you can get up to speed and leverage some of these technologies. And I think that’s going to be even more important as time goes on to have these types of things in place. Finally, Sreeram if anybody would like to book a demo or connect with you, what’s the best place that they can get in touch.
Please. Welcome to go onto a website. We’re putting an email address and our team will get back. Otherwise, you can reach out to me and I can arrange for something. It’s www.Simfoni.com, Simfoni spelled as S-I-M-F-O-N-I.com.
Fantastic. Okay. Sreeram, it’s been a great discussion with you and learning about AI Solutions for analysis data. I’ve certainly learned a lot about how while Artificial Intelligence can be a real game-changer, you still definitely need to have the human touch and the human expertise in place. and then both man and machine together are a pretty awesome combination. So if there’s one lesson to take away from this interview, I would definitely say that is the key. Thanks again for joining me. Sreeram, it was great talking to you and look after yourself, keeping in touch. All the best.
No, thank you so much for having me. I think these strategic discussions are great. We learned quite a bit. Thank you again and look forward to catching up sometime later and discussing more about how we enable compliance and move clients through their digital transformation journey.