Jets Group's experience with AI in quotations and orders

Frode Stenstrøm: How can technology be used to create real economic growth in a traditional manufacturing company? In collaboration with Prosesspilotene, Jets Vacuum has created a completely new sales solution on Dynamics 365 Sales, where artificial intelligence is used to create quotes. Listen to Trond Jarnes from Jets explain how this happened. Trond Jarnes: Hi, my name is Trond Jarnes. I work as Business System Director at Jets Group. Jets Group is a Norwegian company based in Sunnmøre outside Ålesund. It is in a municipality called Hareid. Like everything else in Sunnmøre, it started here in the maritime sector by making sanitation systems for the other companies in the shipbuilding industry. And we are nearly 40 years old. We supply vacuum toilets and treatment plants for ship- and land-based solutions. We have supplied over 570,000 toilets. 85 percent of the business is exports abroad. We have over 40,000 installations worldwide. And we have around 200 employees. We have a global presence with a network of representatives and agents around the world. We are proud to say that our system uses significantly less clean drinking water than a traditional toilet system. So we account for daily savings of 27 million liters of drinking water every single day. We are divided into three main areas. There is the maritime sector, which we started with, where we supply all kinds of vessels, everything from large cruise ships with several thousand toilets on board to small recreational boats or other tankers and the like. We have also developed a land-based market. In Norway, it is better known in the cabin sector, what we call cabin and home. There is also a market for it in Sweden. But we also see that this is beginning to gain a foothold in other land-based buildings. With things happening in the world, with drought in Europe and California losing its groundwater, our solutions can reduce water consumption considerably. And we have several flagship projects in large commercial buildings and football stadiums and various other places. The aftermarket is also a large part of our business. That is what we will talk a little about later today. We sell spare parts, upgrade existing systems and do what we call retrofit, where we replace competing systems with Jets products. We also have our own service department that handles all incoming inquiries. Frode Stenstrøm: You have been on a journey, Trond, and as you said, you talked a little about the company's journey, 40 years of development, and then different markets and growth. But what was the basis for you beginning to invest so heavily in technology as you have done in recent years? Trond Jarnes: Well, it was really that we saw that our existing systems were becoming an obstacle to our development. We are a Norwegian company, but we also have a manufacturing company in Latvia, and we have sales organizations around the world, so entering new markets with a somewhat old sales solution made us realize that we needed to clean things up a bit and essentially rebuild the foundation. Frode Stenstrøm: Because you have, you can briefly say a little about the investment you have made overall, but I know that you have made a fairly substantial investment in quite a lot. What are the main features of what you have actually done? Trond Jarnes: Well, it really started with a need to implement an ERP system. When we started that project, really the procurement process, we quickly saw that we should also look at sales support. We had implemented a CRM system in Norway in the land-based part of the company, but we really wanted a common platform for the entire company, so an ERP project also became an ERP and CRM project. And when we started looking at it, there was the good old garbage in, garbage out, so you need good master data. We then saw that product data and engineering data were important to get in order. So we also embarked on a PLM project at the same time, meaning a Product Lifecycle Management System, because a huge effort was made to clean up technical drawings, CAD source files, and so that ran in parallel. So we have essentially replaced the entire foundation of the systems and then integrated it with various other specialist systems. Frode Stenstrøm: Trond, could you say a little about the background, because when you began automating this area where you wanted to do something about the order process, what was the processing time for entering orders before and after this type of change involving the use of artificial intelligence? Trond Jarnes: We found that we had around 7,000 quote lines on average per month, and a good deal of that comes as structured information through emails or attachments to the emails. And of course, a good deal of it comes from others contacting us by phone, where it is not so easy to automate. So we estimated that perhaps a little over half, just over 4,000 quote lines, could be automated. And we spent about 400 hours per year on manual copy-pasting and data entry. But then we saw that by being able to automate this, we could reduce that considerably. So we envisage that we can save a little over 400 hours per year by automating it. And then I think that by working with more structured input, working with our customers, working with our agents, we can further increase the number of hours saved. Frode Stenstrøm: And then, of course, an artificial intelligence robot does not get tired, so it can really just work as much as possible. There is no limit to how many orders it can process; it is more a question of money and time, really, meaning the energy used for them. Trond Jarnes: Yes, that is actually a very good point, because we are in the fortunate position of growing enormously, and then there will also be more orders to process. The more installations we deliver, the larger the aftermarket becomes, so there will be even more to handle in the future. So being able to actually increase capacity without greatly increasing the number of employees, while managing to do more with the same people. That is also important to us. Frode Stenstrøm: A great many manufacturing companies have a factory, they make a product, they have an order office, a back-office function where they receive the orders, they have a transport and logistics system to ship the goods, and very often the sales team sits and enters the orders into an ERP system. I also know that you have made some cultural changes in the way you process the orders. Could you explain that to me, Trond? Trond Jarnes: Well, in the solution we moved away from, we had spent a great deal of time and money customizing the solution, and it was a system where everyone sat and worked in one solution. But when we were now moving to a CRM platform in addition to ERP, we wanted to move quote creation into the CRM system. Then we had to expand that solution so that we could feed the ERP system with what it needed to produce it. Frode Stenstrøm: But then, that raises the big question: why? Trond Jarnes: Well, it was really a strong request from the sales departments that they should work with the sales processes in a CRM system, but also handle orders and create documents in the same system. And the fact is that once it moves over to the ERP system, the train moves on. Then it is followed up further with deliveries and that part. Frode Stenstrøm: How far have you come as an organization on that journey from the way you just described it? Trond Jarnes: Well, it is in place. Absolutely. So having that close integration between CRM and ERP was an important part of the project. We have spent a good deal of time getting it to flow as smoothly as possible. Now that you have been through the technical journey, and you have day-to-day operations, what is the mood among the sales force now? Well, it seems good. I am happy with the solution, and of course I was very accustomed to the many decades of customization of previous solutions. And that is perhaps what pushed us to automate as much as we had to as well. Frode Stenstrøm: So I think it is very good. And we have also got a solution that fits together very nicely. There is automation in place. I have spoken with customers, and everyone wants automation. Not that many actually implement it, and those who do sometimes feel that it can seem a little chaotic, because they do not quite know what is happening, when, where, or why. How has that automation, and the use of artificial intelligence in this case, been received among your people? Trond Jarnes: This is very exciting, especially since you bring that up, because this concerns human aspects, change management and that part, so it was very exciting to start with automation in a department that was incredibly Dedicated to handling spare parts sales. So there was a lot of “punching,” and we introduced a completely new way of working, where you uploaded a PDF that was then interpreted by AI, which then created a quote line. But it should have been very well received. And I think the key here is partly that we chose to divide it into phases. So we didn't fall into what I would call the trap where everything has to be automated at once. So the first proof-of-concept process was really that we would take incoming PDFs, and then we would manually upload them, and then we got the rest of the quote header information; that part, we had to do manually. And that gives the user a sense of control. They're the ones creating the quote, but they get help with this punching. Frode Stenstrøm: Interesting. Because I think that control is a pretty key word for that particular journey. If I use an IT system and don't fully understand how it works, I don't always have to understand, but there are some things I need insight into so I know I can trust it, and then I will use it. But if I don't feel that I'm in control, I'll just throw it away, and I think that's how it is for most people. They need to know what's going on. And for most people, I think artificial intelligence is a bit of a black box. What just happened? Who answered these questions of mine? So it's interesting to see how your organization has absorbed that change now. How long did you spend on that time frame, I suppose, when you started with this automation and artificial intelligence? Was it a project lasting a day, a week, a month? Could you say a little about the time frame for it? Trond Jarnes: Yes, then you may have to help me a little, Stine. Exactly how long we spent on it. We did spend time discussing what we wanted to achieve. We had some workshops on this. And the development time itself was about two weeks, is that right? Stine Sletvold Øverli: That sounds about right. Frode Stenstrøm: Yes, two weeks of consultant time in terms of the development, yes. So it's not a major investment as such, but I'm thinking, once it was on the platform, how long did it take before the organization flipped from “what is this?” to really adopting it? Trond Jarnes: Yes, we started a little cautiously there. We had dedicated superusers within the different areas. And the person who is the superuser for spare parts sales was involved in scoping this, and helped test it along the way. And then she moved on to testing this when we went live with the first version. And she also brought in a colleague who handled some of the large inquiries. Yes, that was where there was the most to save. And I think the idea was that it would be an even more controlled rollout, but then word started to spread and people were curious and wanted to start trying this. Suddenly, our entire team of just under ten salespeople was using it. So it went faster, but I think we spent a month, the month of May, slowly but surely scaling up. Frode Stenstrøm: That experience you're sharing now, I know it. I've spoken to so many companies that are sitting there scratching their heads. How do we get started with that change? They often understand that there is something about artificial intelligence for their business, but very many are left with the feeling that it's so big and so complicated, and it takes so long, so they just put it on the shelf, right? And the journey you're describing now, where you can go and accomplish some things with perhaps one to three weeks of work, and then start creating an appetite in the organization for it, I think companies need to experience that journey themselves, and then they simply have to start somewhere. But the way it's organized is also interesting, I think, because I've heard from various places. And the customers I've spoken with who I see succeeding now in using this for something scalable, they do what you've done, where you've grabbed one or two people, started small, and just got started. Not overplanned it. That's really what it does. Trond Jarnes: Now we've just ordered version 2, which is a further development that will support even more complex, large inquiries and also connect a few more data points. We're already looking toward a version 3, where we can create the entire quote document, the quote header. And we're also looking at using case management for incoming inquiries, so the target state here is that we should be able to fully automate the inquiries, but we will always have a knowledgeable salesperson in the loop to tweak it before it's sent out to customers. Frode Stenstrøm: And there you have another phenomenon that I also see turning what we know as ordinary software upside down, really. Today, we're used to sitting in a system, we're used to entering some things, placing an order and then sending a quote to the customer by email or in another form. But now the robots largely do a lot of the heavy lifting, the robots do the groundwork, but then the question is human in the loop, for lack of the Norwegian term: when does the human actually enter this process and quality-assure what's being done? And we're seeing now, and Microsoft has been open about this, that products like Dynamics will lose their interface to a much greater extent, going from being a register-based, established system, to becoming more about: now we're going to review what the robot has done, and then look it over and provide feedback, possibly correct it, before it goes out again. And I find what we're going to experience in terms of change in our industry now incredibly exciting, including the way people interact with technology. And you've started that journey, then, from what I hear. Stine Sletvold Øverli: And Jets' motto has probably always been that AI shouldn't create more work for salespeople. So they should spend less time validating what comes in. Frode Stenstrøm: Fantastic. Stine Sletvold Øverli: Yes, and punching in the lines in the first place. Frode Stenstrøm: I've never met a salesperson who enjoys punching in anything at all. But there's another observation in that, and I'm wondering how you've experienced it firsthand, Trond. Yes. I use artificial intelligence a lot myself in my work and my job, and what I notice as its immediate effect is that it frees up time. Time I previously spent perhaps writing a Word document, I can now spend less time writing and more time reviewing, so that's one change, but the other is that I have more time to talk to people. And in a business context, having that communication is, in my opinion, perhaps the most important thing we do, simply having time to talk to our customers too. When you've adopted this at your company, if you free up time for the staff, for example, what do you use that freed-up time for? Trond Jarnes: Well, it's exactly that we have highly skilled people who know our products and our customers very well. So actually not having to spend time on the copy-paste work they did before, and actually being able to spend more time on the complex cases where the engineer took pictures of the pump and sent them in, and actually figuring out what this customer really needs. And also being able to look at more sales opportunities. We want to become even more data-driven, and simply look at the data points we have historically, and then see where we can proactively reach out and offer spare parts. If you look specifically at a cruise ship, a cabin without a fully functioning toilet is a cabin that can't be used, and that potentially means significant lost revenue. And of course, if you compare that with a cargo ship, it's a little different. So we also have to look at the different segments, because there are differences here. Frode Stenstrøm: Yes, that's exciting, because then you use the freed-up time for upselling, really. Mm, absolutely. But then I'm very curious about the solution, because here you've taken a standard sales application, and then added an agent that takes incoming documents and fills in the quote lines. But you mentioned something, Trond, about some of what I see as business rules, the order of them, for example— how has that actually been handled in a world of artificial intelligence? I'm getting a little curious about the solution—can we say a little about it? How does it work? Stine Sletvold Øverli: How technical do you want to get? Frode Stenstrøm: No, as much as you like. No, but just for inspiration, because if someone is sitting there and... Let's say there was a builder sitting there now, and they know that we're going to build a garage, then there will be product lists galore. If they're building a kitchen, then they need that. If they're building 24 apartments, then there are quite a lot of product lines that need to be included in that kind of quotation process. How do you actually build something like that now? Stine Sletvold Øverli: Then you should really ask Vegard. Frode Stenstrøm: Then you should ask Vegard. Because these are the business rules that are built into this. And Trond, when you sat down and specified this, how did you know that these were the rules you wanted the solution to handle? Trond Jarnes: When it came to order, that was something that was important and easy to understand. It was a matter of finding a line number. If you get a PDF with 41 lines of spare parts, then you need to check the column containing the line number. But there were things that we couldn't fully anticipate, so we had to take a somewhat iterative approach. Among other things, we can see that products in it—we're fortunate or unfortunate enough to have leading zeros. That's a legacy we've carried with us for many years. Everyone who works in Excel knows that leading zeros are a nuisance, because they have a tendency to disappear due to formatting. And that can also happen when our customers' systems send it in to you, and then the zero disappeared. So we added an operation to the agent here that meant that if it didn't get a match here, it tried again by adding a zero in front. And that made the match rates significantly higher. Frode Stenstrøm: So in a way, you've embedded those kinds of self-repairing rules in the business logic? Yes. One of the weaknesses of using artificial intelligence has historically been hallucinations, that it simply makes things up. Have you noticed any of those kinds of things in a solution like this? That it simply invents something similar? Trond Jarnes: No, I haven't felt that there have been any instances of that. I don't think it has happened to my knowledge. Frode Stenstrøm: No, that is very, very good. Stine Sletvold Øverli: It will usually add it as a write-in line if that were to happen, and then it's still up to someone to monitor and validate it. Frode Stenstrøm: Trond, now that you've been on that journey, both in terms of investing in technology as a whole, changing the sales culture, and then taking the time for artificial intelligence. What would you have done differently if you had started by rewinding, turning the clock back four years— what would you have done differently then? Trond Jarnes: Well, with everything I know now, it would of course be to look at what we can automate. Traditionally, we've spent years discussing what that button should look like and what it should do. But what we get with AI now is that we remove the button. The button isn't there. It happens automatically. That's something exciting. We've spent a great deal of time discussing solutions, reducing the number of clicks in sales solutions, whereas now we're suddenly leaping very far ahead because we can automate it. But then again, there's also the human aspect, that we have to take it step by step, and, in good Norwegian, ease into it, which I think is important. So I don't know, it's hard to say. Frode Stenstrøm: I think the realization you have in all such projects is that technology can do a great, great deal, and it can also happen very, very quickly, but we can't move faster than the organization is able to, I think, in what they've done. I also think that perhaps our advice when it comes to artificial intelligence is to try, instead of making it so big, to do a small project like this, invest, what, perhaps 50-60-100 hours externally, and 50-60-100 hours internally, and then simply try it and see what value creation you get out of it. Then I think perhaps it creates that appetite in the organization to move forward with the technology choice. Before we finish completely and wrap up the thread, I'd like to ask you a question, and that is: if you were to give some advice to other colleagues who work in a role similar to yours, what would it be? Trond Jarnes: It is simply to involve internal resources, get buy-in all the way up to senior management and, not least, spend time finding the good business cases, and then break the elephant into smaller pieces. Take it piece by piece, and iterate your way forward. And consider the human side of it, and bring the user along throughout. And don't create the perfect technical solution in the first version. Stine Sletvold Øverli: That has also been one of the biggest success factors in the collaboration between Prosesspilotene and JETS behind the scenes as well, that they have allocated enormous internal resources to move the project forward. Frode Stenstrøm: And it's only recently, I think, that I've answered a number of tenders where there has been a question asking how companies can succeed, and I think it simply comes down to collaboration. That you sit together and work together and solve a problem and don't make it so big. And with those words, we say thank you to Trond from Jets and Stine from Prosesspilotene for their contributions to today's broadcast. If you're wondering about anything, ask your question in the comments section below “Was it useful?”. Like and share the video with those you think will benefit from it. If you want more professional content, subscribe to our channel and follow us on LinkedIn. This broadcast can be found on our YouTube channel as well as the website prosesspilotene.no. The broadcast is produced by Prosesspilotene, which increases the value of your customers using Microsoft technology. Prosesspilotene's analytics team can help you create solutions like the ones we saw today from Jets Vacuum, using artificial intelligence as well as insights and analytics. Prosesspilotene's collaboration team can help you create customer clubs, websites and customer portals. The person responsible for this episode has been me, Frode Stenstrøm. The project manager is Oddbjørg Berge. Wishing everyone better work processes. See you again.