How AI accelerates enterprise software go-lives

Hi, everyone. Thank you so much for making time today. I know this is a group of very, very busy people, and we’re excited to have you here to talk about how AI is changing how fast and how you can go live on enterprise software. And specifically, we’re really excited to share our new implementation agent with you. So quick introductions of the folks that you will see on this call. I’m Tara Gottesfeld. I’m leading product marketing for Milo, our AI-powered implementation agent that we’ll be sharing a whole lot more about today. Nav is our VP of central architecture at Zuora, and he is leading the product side here. He’ll be doing an awesome demo for you in a few minutes. And extra excited to welcome Eric Bodge, who is the VP controller at ConstructConnect. And Eric is among the first customers

to experience this agentic implementation firsthand, and he’s going to actually walk you through what it looked like on a project. So very grateful Eric is here and know you will all enjoy hearing his story. Quick forward-looking statement. We’re going to focus on what’s available today. We might share some things about where we’re heading. If you have specific questions about specific future features, just check on our community ideas board or ask a Zuora team member that is working on your account. And finally, some housekeeping. So please take a moment now to use the chat tab to say hello, tell us where you’re calling in from. You’ll also see a Q&A tab at the top right of your chat box, so please use that for questions. We’ll do our best to answer those in real time.

If we don’t get to your question, rest assured, someone from Zuora will reach out and follow up with you. Finally, for additional resources, upcoming webinars related to today’s content or other Zuora content, just check out the links to resources, um, in the Click More tab. And last thing, if you can’t stay for the entire session today, though, we really hope you can, um, you’ll receive an email after with all the recording, so you can listen back to this whenever you want. Okay, let’s get started. So before we get into implementation, which is kind of the core topic today, let’s zoom out on why we’re even talking about implementation. So big picture, we’re all talking about AI all the time, and we know that AI is dramatically

changing how you are monetizing. So used to be subscriptions, that’s how Zuora got started. Now it’s much, much bigger. We see companies across industries who are doing outcome-based pricing, usage-based pricing, hybrid models, drawdowns, and it’s all happening really fast. And we know the hardest part for our customers isn’t just launching the AI pricing, it’s actually running it behind the scenes. So the way you’re monetizing directly changes the way that you have to run quote to cash. So a new pricing model doesn’t mean anything if your quoting and your billing and your revenue can’t actually execute on it, if you can’t track if you’re even making money on it over time, if you can’t be audited and feel confident about it. So when the way

that you run quote to cash has to change, that then changes what you need from the systems that are running quote to cash. Because what we see is often the old system at first can be stretched and then gets to a point where it genuinely cannot do the new thing. So we used to see customers customizing their ERPs, um, whatever it may be, billing, rev rec, both. But AI is just moving too fast for all of that customization, and so changing these systems has become a lot more urgent. You can’t take a year to implement or to tie up half your team for many, many quarters. So if AI is compressing the time it takes your business to decide on a new pricing or packaging model to just a few days or weeks, but your system still takes six months, you have a bottleneck exactly where you need

speed. And we hear the same thing from finance and IT leaders all the time, which is a slow-moving implementation isn’t just costing them money, it’s also just tying up capacity that their teams don’t have. So you have to be able to move at the speed of AI. So what is the old world, the pre-AI world of implementation? Why was it so long and treacherous? So what we’ve heard from customers is a lot of this is about the immense amount of upfront work to get a really clear view of how your systems are running today. So what products do we actually sell? Actually, so not just what the price book says, because sales quoted something non-standard eighteen months ago and nobody updated the master list. Which contracts deviate from the standard terms?

And the honest answer is often more of them than anyone even knows or wants to admit. The exceptions live everywhere, not necessarily in a system that anyone can query. And what systems are actually connected to this, and what breaks if we touch those systems? And often no one is entirely sure because the person who built that integration left in a reorg or isn’t at the company anymore. So which data can we trust and which needs validation before it goes anywhere near production? Which exceptions actually matter? Which ones can we set aside? What makes it so hard to answer these questions is that the answers are scattered. Some are in the product catalog, some are in a shared drive of PDFs, some are in a finance person’s head, an IT person’s head, a spreadsheet, a series

of many spreadsheets. And historically, getting the answers to these questions, because they are so disparate and so disconnected, has taken months before a single implementation decision can get made and before you can get that implementation off the ground. And that time isn’t neutral time, right? It’s a strategy that you can’t launch. It’s a finance team that’s doing manual workarounds because their system of record isn’t ready. It’s the g-go live date that keeps slipping. So eventually, we do see most customers get onto a working system, but the thing in question has just been how much time and how much pain will it take to get there? And historically, for all enterprise software, I’m not just talking about Zuora, the answer’s just been a lot more time and

money than anyone signed up for. What does AI change? Well, in simple terms, AI can read and reason across really, really large volumes of unstructured business information. So contracts, catalogs, transaction history, and it can do that at a speed and a scale that manually referencing those spreadsheets and their PDFs was just never ever gonna hit. So what that unlocks is a completely different order of operations. So instead of spending time gathering information before anyone can react to anything, you can actually get to a real starting point with those answers you need really quickly. And this doesn’t take away the human time, but the humans can spend time validating

and refining instead of just assembling this understanding of the business from scratch. So old world, blank page, you don’t know what you’re dealing with until you’re really deep in it. Takes a really long time, takes a lot of resources. New world, you know what you’re de-dealing with pretty quickly. You can move a lot faster, and everyone feels a lot better. Given everything we walked through, here’s how Zuora is thinking about this. Our job was never just to build software and hand it off, right? It’s helping customers actually change the way that they’re running billing, revenue, quote-to-cash, and how they’re doing it repeatedly as their business changes. So we’ve been part of thousands of these implementations across direct engagements, working with our partner

ecosystem. I know we have a lot of our partners here, so shout out to them. And AI didn’t just kinda hand us a faster button to the same process. It actually gave us a really cool opportunity to completely rethink what implementation should look like. And that rethink has a name. So really excited to introduce Milo. So Milo is Zuora’s quote-to-cash implementation agent. The Milo stands for Model, Integrate, Launch, Optimize. So Model is really two things. So first, it’s understanding your business, your systems, your data, your workflows, contract logic, pricing structures, revenue policies. I’ll show you in a second what, what that looks like. We’re calling it your quote-to-cash digital twin.

Um, and then doesn’t stop at a model. It actually becomes a configured Zuora setup. So built from your actual raw inputs, your team can log in, click through, not just see a diagram. Integrate, obviously, Milo is then mapping the systems, the workflows, the data, um, into a real integration approach. Launch is configuration, migration, validation. Historically, that phase consumed the bulk of a six-month-plus timeline. Um, now it is much, much quicker. You can go live in weeks. And Optimize is our acknowledgment that the work doesn’t stop at one go live, right? Now you have a real foundation and infrastructure to change. So say six months from now, you’re a billion customer,

and you wanna add revenue, or you’re standing up a new business unit. You’re not starting over. You’re not re-explaining your business from scratch. You’re actually extending the same model. And it’s kind of the difference between implementation being this tax that your team has to keep repaying, um, and actually just keeping pace with the way businesses, and especially AI-led businesses, are going to evolve. And one last thing that’s worth calling out at the bottom. So this is built by experts, and it’s reviewed by experts on our side and your side at every step. This is not an unsupervised AI running your implementation. It’s really Milo handling the repetitive, the manual analysis work so that our team and your team and any partners involved can spend time on the decisions that

really require strategy and human judgment. So now let’s get into how Milo actually works, how this changes the experience from evaluating software before you even buy anything to going live and beyond. So Milo starts by understanding your business and builds what we’re calling a digital quote-to-cash twin. So today, this is guided, so it’s someone from our team walking through it with you. Self-service will be next. So what happens is, with our help, you drop in a set of about ten documents, contracts, mostly as PDFs. Nothing gets reformatted, just the raw files in a secure link, all the NDAs in place that you’d expect. From that single upload, you start getting answers

to those questions that we raised. So what do you sell? Your actual product and pricing structure pulled from the documents themselves, not retyped by anyone. How you sell, the commercial terms that are sitting underneath all of that, so renewal language, overage handling, minimum commitments, the stuff that’s usually buried in a contract that maybe no one has reopened for a very long time. And finally, smart flags. This is the one I personally think is the most exciting. So immediately, you’ll start to see inconsistencies and risks that the system can kinda surface on its own. So maybe there’s a payment term that doesn’t match with the rest of the portfolio, or there’s a renewal clause that contradicts the standard, something that genuinely needs a person’s judgment before it can go any further.

So this is all really exciting. What hap– what happens next? And I’m going to turn it over to Nav to share more about how Milo works once he has this digital quote-to-cash twin and is ready to start your implementation. So Nav, I will turn it over to you. All right. Thank you. Thank you, Tara. So guys, all we have seen is how Milo gets you from almost nothing to somewhere where all the groundwork, your knowledge is already in Milo. Now, this knowledge is not gonna go away. We use this knowledge to synthesize what your environment is gonna look like, what your Zuora setup is gonna look like. I’m gonna share my screen really quickly and take you over how we do next setup of the things.

Now, before I dive in, I am going to just make sure we set up The mock environment as well. Here we go. What you have on the screen, similar to the product catalog data, I got account raw export. This has accounts in the contacts data. And similar to that, I got subscriptions raw export as well. Now, some systems do not have subscriptions. That’s okay. Give us code, code line items, order, order line items, whatever works. Um, once all this raw data is there, that’s when the magic starts. Let me hop over to Milo real quick. This is what it looks like. And whatever the files I was showing you from Acme, I have just deposited those files over here, account, subscription, product. As I said, if you have multiple files, products and pricing separate, bring it in. Subscriptions,

order, order item separate, bring it in. Once I have the files in, I’m gonna go into the mapping. This is a live workspace for migration of a legacy data through Milo. Let me explain the workspace a little bit. So on the left, you have the navigation. Then I look at the source data. Then I look at a little bit of a code, which is obviously mapping code. But what’s most interesting is the right-hand side, the extreme right. This is where I work. I don’t touch the code at all. Now, for example, there are some instructions I have to say. When you give the source data to Milo, what happens is Milo looks at the fields, infers it, and then generates a code, which is basically the mapping code automatically. No input needed for that. But then, because I know Acme needs some special handling, for example, the first instruction is if

I see the goth-growth plan, I want to make sure that product catalog also has AUD pricing as forty-nine point ninety-nine. Now, let’s say, how does Milo process it? Milo will process it, generates an instruction. This instruction, as you can see right here at the bottom middle, it gets recorded in it. On the top of the mapping that Milo has done, it takes in instructions as well, and there can be bunch of instructions that have to be run. These are repeatable instructions. I can take this whole logic and then execute it on multiple datasets across multiple iterations, which saves a ton of time. Now, obviously, I may get some errors as well. The beautiful thing is, conventionally, I will be troubleshooting the errors myself. But now I’m taking the error, bringing it back into Milo and asking Milo

to help me out. Milo generates another instruction because, for example, it says if you see discount percentage or fixed amount, there’s a special handling for discount level and what have you. All these are recorded as instructions here. This is product catalog mapping. Similarly, I have done account mapping. If I– For example, one of the instruction I gave is, if I don’t have the sold to contact, then automatically map it as same as bill to contact. Record it as the instruction right here. Just in the same fashion, I got orders as well. Now, order is a complicated thing because we can have multiple source data files in this, but it also can take mapping instruction, validation instructions, so your instructions can be if-else format as well. Once done, all of these mapping, all of these instructions are taken along. But what is more important is

I can validate right in the panel. So I’ll look at the source data on the left-hand side. This is what the source data was in the CSV, and then look at what Milo will be pushing into Zuora. Orders, order action, create subscription, what rate plans it subscribes to, all the codes, and everything including quantity. This is exactly the same in product, accounts, orders as well. So validation right in there. When I’m looking at it, I know that my data is going in right. All done. I’m ready to load. Once you’re ready to load, you will see a bunch of executions I have done. I have wiped the data sometimes. I’ve had the errors, I’ve iterated. Same way, every migration goes through bunch of iterations. Now, once I run, I may have some errors as well, right? So in this case, you see that there are some errors. In this case, the error message says, “Taxation

requirement state is required for sold to contact if the country is US or Canada.” Now, I cannot fix this because this is a source data. What we do, what Milo does, is it allows, uh, me to download the CSV, send it to Acme, and say, “These are the source records that need correction as well.” As simple as that. And then the whole process can quickly repeat. Basically, it has loaded nineteen hundred and thirty-five accounts anyway. So let’s see what Zuora tenant looks like. All right? So once the data is in Zuora through Milo, right here, this is what it looks like. All the accounts are here that I just loaded. Now, within that, I can go to Subscription. Let’s see. This is my subscription owned. I’m gonna go into the subscription and see, along with even the amendments and the versions as well. So all in all,

the data is completely loaded. Let’s take a quick look at the product catalog as well. Is it there or not? And it is there pretty much, right? So let me quickly back up. What happened? Right from where we started, we had a blank tenant, not even configured. We took some of the information that Milo gathered from the business, from Acme. What is the business about? How does it run? What products it sell? What kind of configuration they may need? We inferred that using Milo, and then Milo pushed those instructions, pushed those configurations to Zuora. Zuora tenant got configured. I took the self-service portal requirements of Acme, used Milo to do the API design, API build, including the code generation as well. We got the integration sorted. And finally, we used Milo to load product catalog, accounts,

and subscription data. And here we have our tenant fully running in almost as much tam-time as it probably would have taken me to schedule a couple of discovery sessions. Conventionally, this takes weeks, if not months. With that, I’m gonna hand over to Tara for next. So you guys just saw a really awesome demo. You now understand how Milo works. You understand the problems that it is going to solve. You understand that it’s going to make everything from deciding whether to buy to actually going live, to then adapting once you do go live much easier, more clear. You’re gonna have more confidence throughout that process. Um, but do not take our word for it. I’m really excited to introduce you to Eric Boje, who is the VP and controller

at ConstructConnect, and Eric actually has been using this technology to go live on Zuora Revenue. So really excited to have him here today. Thank you, Eric, for your time and for sharing your story. Um, would be great if you could just start off by telling the audience a little bit about you and, and your role at ConstructConnect. Sure. Thanks, Tara. Thank you so much for having me today. My name is Eric Bodge. Uh, personally, I’m the father of two young kids, three years and one years old. They keep me busy, but I love spending time with my wife and kids and watching the kids grow. Professionally, I’m the controller at ConstructConnect, a software provider in the pre-construction space. I’ve been with, uh, working with Zuora for over eight years now, and I’ve been the primary owner of the Zuora relationship for the past year and a half.

Part of my role is to optimize the order-to-cash processes, and with Zuora being the heart of order-to-cash for us and how we invoice and run subscriptions, we have worked closely with the Zuora team to optimize how we are utilizing the Zuora platform over the past year, including going live on orders, CPQX, invoice settlement, and now Zuora Revenue. This year was the right time for us to expand Zuora Revenue because we were seeing an increasing volume of revenue entries that had to be made outside of the system, and it was both no longer sustainable but also carried unnecessary risk to make out-of-system adjustments when Zuora Revenue was giving us the ability to book all of the revenue in-system. We love hearing, um, heart of order-to-cash. That’s, that’s so, so

wonderful to hear. Um, and tell us, so okay, you decide the, the pain, the pain is great enough to embark on an enterprise implementation, but you probably had some preconceived notions of what this would feel like having been through other enterprise implementations in the past before AI. Can you kind of describe what the old world felt like when you did these implementations? Enterprise implementations always require a deep understanding of the business. What teams are gonna be impacted? Where and how is the data gonna be used? Is this gonna change the availability of data that we currently rely on? What resources outside of my control am I gonna need so I can ensure that we are aligned on timing and prioritization? You do your best to

avoid surprises, but there’s always gonna be surprises, and you do your best to adapt and to minimize any negative impact from them. Maybe another team, uh, has a shift in priorities. Maybe you need a resource that you hadn’t planned on. Maybe the estimated amount of support you thought you needed from the vendor is, uh, gonna be different, and it’s gonna place some cost strain on the implementation. It requires a great deal of coordination and understanding dependencies to keep the project moving forward on time and on budget. So a lot of it, disparate teams, disparate documents. There’s a lot of context that’s living in a whole bunch of different places, and it’s really, really hard to bring it together. It takes a really long time and, and a lot of people. Um, this is something we hear from a whole bunch of customers who, who have gone through this and really

why we’ve built this technology. So once you kind of got in there and started, um, using Milo, our team started using Milo with you, I know you mentioned you had some really exciting kind of aha moments throughout this process. Can you describe some of those moments where you really said, “Hey, this is, this is different. This, this feels faster, better”? Yeah. It was clear from the beginning that the Zuora Revenue work, um, that this implementation was going to be different. As soon as we signed, we had the sandbox environment available to us with real ConstructConnect data in it. We could hit the ground running and start testing right away. We knew the scenarios that were gonna be challenging for us, and we could see how they were flowing through the system with real examples. That made it much quicker to identify where we needed to iterate and to work with the Zuora team

to make updates so that we could start testing again. The biggest game changer for us was when we got a surprise that we needed to modifly– modify a Zuora workflow to allow billing data to flow through to the revenue module the way that we needed it to. This was unplanned work, and the go-to-market systems team had other priorities that rightfully needed to be prioritized over this change. The internal estimate we received was five weeks for this to be completed. I took this timing back to the Zuora team, and they had the flow modified and ready for us to test the very next day. This kept us on track and kept us moving forward. The ability for Zuora to partner with us and keep the timeline within our control saved months. Yeah, the very next day was, was not something you were, you were used to, to hearing

in an enterprise implementation. Absolutely. And we know everyone is really excited that this technology is faster and easier, but we also hear a lot, especially from controllers and other finance leaders like yourself, how are, how are people gonna be involved? Is Milo just gonna be off running, running with my data? And I know from working with you and your team that that couldn’t be farther from the truth. Can you help people who might be considering doing something like this understand how humans were involved in this process and maybe involved differently than before? Yeah. Um, it is absolutely essential to have a human in the loop. My philosophy with AI is that AI can save time producing the what so that humans can focus on the more value-added why. It’s trust but verify.

Is this the what I expected as an output? Why am I getting any discrepancies or any unexpected results, and how do I fix them? I have to give a huge shout-out to my team. Uh, in particular, Trey Bramble, our accounting transformation senior manager, project managed this implementation and did a wonderful job. He was able to clearly communicate to Zuora what our needs were, and we had the same Zuora resources throughout the entire project. Both Trey using our AI tools and the Zuora team using Milo were able to greatly accelerate the data gathering and identifying the necessary testing scenarios so that we could focus on the results. It’s key to understand the business well enough to know what to look for in the data. AI got us most of the way there much quicker by summarizing the data, highlighting the differences, um, that needed reconciled,

and then Trey and the Zuora team used their knowledge of our business and what the output should be to correct and iterate. It allowed us to focus our time and efforts on reviewing and making decisions rather than having to build the reconciliations. Right. Yeah, it’s really important that all of this AI is two things, both built from the expertise that we have from all of these quote-to-cash implementations, so it can help to spot the patterns, spot the opportunities for efficiency, spot maybe the pitfalls of others who have gone through this. And then equally important is the, the leadership and the expertise on, on your side, and, and that is definitely why this went so well. Um, so looking back, what surprised you most about this process? So what surprised me most goes beyond just Zuora

revenue. In the past nine months, we’ve partnered with the Zuora team on several items: orders, CPQ, and invoice settlement. Zuora has been a true partner. There were several times when I thought that we were gonna be blocked for an extended period of time, but Zuora stepped in, in unexpected ways to provide a solution to remove the blocker and keep us moving forward. You only have internal resources for a certain period of time, and then they have to move on to other priorities. One delay can completely derail a project if they aren’t available when you need them to plug back in, and then you have to realign calendars and priorities once a blocker is removed. Being able to feel like you are consistently moving forward and that the team is focused on value-added work instead of just data gathering was a very pleasant surprise. It’s not, it’s not just like the speed

of any specific part of the process, though obviously that’s great, but it’s also kind of this feeling of ongoing momentum and agility that wasn’t really possible before. Yeah. What would you tell another finance or tech leader on this call who’s maybe considering going through an enterprise implementation and leveraging some of this technology? Yeah. I would say there’s never gonna be a perfect time for a system implementation. Your to-do list and your competing priorities are only gonna grow. But if you’re running manual processes out of system, every month adds risk and adds time required to support those processes. Don’t let the complexity of a migration be the reason for a delay. Zuora did three main things that would give me the confidence to do this again. First, they gave us consistency in personnel

that gave us confidence throughout the implementation. It was the same people accumulating context throughout with no knowledge lost in between phases. Second, it was the ability to start the implementation with our own data, to be able to validate early and to move fast on decisions. And third, it was the ability of Zuora to remove blockers and keep us moving forward. Zuora has proved to us that they are truly a partner, not just a vendor that we pay. That is so, so good to hear. And final thing, I know we have a, a pretty quantitative crowd probably on this call. We have a lot of, uh, technical folks. We have a lot of, um, finance folks. What are the metrics that you’ve, um, been able to gather so far, um, around sort of how this has improved, you know, speed

or, you know, use of employees, things like that? Yeah. There’s two things I would share that best illustrate the experience. Um, and the first is when we had actual Construct Connect data in the sandbox. That was day one after signing. The second was, I won’t forget the first time that I received a five-week internal estimate to resolve a blocker, and Zuora came back the day after we brought the blocker to them and said they had a solution built, ready for us to test. This happened multiple times. Our internal estimates were real. I knew the work being prioritized over this request for unplanned work, and I was lined with it. Zuora removed the tough choice of, do we put the revenue project on hold for five weeks, or do we ask another internal team to reschedule work to support us and risk putting other priorities at risk? I didn’t expect Zuora to

provide a third alternative of, “Let us provide the solution to keep you moving forward,” but they did, and it saved us months. That’s awesome. Yeah, I mean, what stands out about what you said is there, there are kind of the steps that you know are gonna happen that are happening faster, like getting a tenant in a day. That’s, that’s amazing. But there are also all the things that maybe you don’t even anticipate that are gonna come up. And I think what AI unlocks and what this technology with Milo unlocks is really just the ability to do all of those things that are maybe unanticipated or nuanced, still with a lot of momentum and, and confidence. Yeah, absolutely. There’s always gonna be surprises in an implementation, and the ability to remove blockers as fast as we could was incredible.

So great to hear. Um, it’s been so, so wonderful having you today, Eric. Thank you so much for your time. Is there anything else that you want to share with this, this audience? Just don’t be afraid. Dive in. You heard it from Eric. Don’t be afraid. Dive in. Thank you so much, Eric. Thank you for having me