AI Is Rewriting Professional Services: When Faster Implementation Moves the Bottleneck

Help me think through how this episode applies to my situation. Start by asking what I am trying to change. Separate the episode's claims from your suggestions, and say when the notes do not support a claim. Use the transcript to find passages, then check the audio before quoting. Transcript: https://yuvalyeret.com/podcast/episodes/ai-is-rewriting-professional-services-when-faster-implementation-moves-the-bottleneck/transcript.md ## Published episode notes AI can significantly compress enterprise software implementation timelines. That does not eliminate the constraint. It moves it. Julie Hodge, AI Strategy & Operations Leader at Zuora, joins me to discuss how AI is changing professional services, customer onboarding, and the relationship between product and services. We use Zuora's Milo implementation agent and digital quote-to-cash twin as a concrete example, then follow the implications downstream: customer testing and change capacity can become the next bottleneck, customer-facing experts can become builders, and faster teams still need shared direction and lightweight governance. Key takeaways: • Start with the implementation bottlenecks customers already feel. • Digital twins create an early working model from existing business inputs. • When everyone can build, the product–services boundary starts to blur. • Faster engineering can move the bottleneck into customer testing, training, or change readiness. • Tiny teams still need shared context, priorities, and traceability. • Human effort alone is a poor test for whether an AI-generated change needs governance. Chapters: 00:00 Why AI matters in professional services 01:49 Where enterprise implementation time goes 03:27 Start with customer-visible bottlenecks 04:32 Milo, digital twins, and the first working tenant 07:02 Mapping the customer onboarding value stream 08:38 Product, engineering, and services working together 11:38 When everyone can build 13:25 Engineering stops being the only constraint 16:38 Prioritizing the services AI backlog 17:06 When delivery outpaces customer capacity 18:00 Following the moving constraint 20:14 Using AI to build AI solutions 22:19 Ownership across product and services 23:34 Speed without abandoning governance 24:08 What project management survives AI acceleration 27:02 Tiny teams and coordination costs 29:17 Start with one or two visible wins About Julie: Julie Hodge is an AI Strategy & Operations Leader at Zuora. Her work focuses on how AI operates inside the enterprise and how professional-services teams turn AI capabilities into faster, more dependable customer outcomes. Julie on LinkedIn: https://www.linkedin.com/in/jdhodge Learn about Zuora Milo: https://www.zuora.com/products/milo/ Scaling AI: From Activity to Impact , yuvalyeret.com Insights (and help/advice) on scaling AI from activity to impact , yuvalyeret.com/insights Yuval's Linkedin – https://www.linkedin.com/in/yuvalyeret/ ## Transcript Riverside's published transcript for this episode. Speaker labels are machine-generated; check the audio before quoting. Episode: https://yuvalyeret.com/scaling-ai-podcast/ai-is-rewriting-professional-services-when-faster-implementation-moves-the-bottleneck/ Source RSS GUID: 938d255f-cc27-4feb-bc83-4218e4272704 Source: Riverside podcast:transcript URL ## Transcript Yuval Yeret: Welcome to Scaling AI from Activity to Impact. I'm Yuval Yeret and today I'm with Julie Hodge. Julie and I have been chatting about a very interesting use case of using AI to improve how her company delivers services to clients. Julie. Welcome to the podcast. Julie Hodge: Thanks for having me, Yuval thanks for that introduction. So yeah, I've been leading AI within global services now for a year and a half, been focused on AI, and it's been a you know an incredible journey with lots of learning. So happy to dive in. Yuval Yeret: Cool, let's do it. So maybe a good question to start with is why? Like why are you using AI in global services in your company? Maybe tell us a bit about what's the company, what's the context, what do you do in global services, and why is AI relevant? Almost sounds like a trivial question these days, but I find it useful to sometimes get back to it. Julie Hodge: Yeah. Yeah. So Zuora is a subscription subscription billing and revenue platform and as far as why AI, it's you know, software solution like many that needs to be implemented. So my focus has been on professional services and AI is an area that can greatly, you know, improve speed, quality of deliverables, and allow us to give better solutions to our end customers. and that goes from, you know, everywhere from the pre-sale cycle all the way through hypercare. Yuval Yeret: Okay. So I'm hearing so so what I'm hearing you say is we introduced AI or we started exploring AI in order to improve speed. What what do you mean by speed in services? Let let's be more concrete. Julie Hodge: Yeah, yeah. So yeah, there's different areas of speed. It could be from if you're talking about pre-sales, creating, you know, the contract for the customer, what's in scope, what's out of scope. once you get into deliverables, you have things like design documents, being able to iterate on those fast. then you know, even with builds, you've got things, you know, configuration, all of those type things that need to be done, customizations, you've got deployment. testing is another area that tends to be very time consuming. So looking at that and seeing where that could be sped up, maybe with things such as automatic test create test case creation that then ties back to user stories. So really, you know, bringing everything together from the very beginning of the project through, you know, testing and and then, you know, supporting through hypercare, you know, with improved documentation of solutions, you know, the how the solution was built, et cetera. Yuval Yeret: So when you say speed, I'm hearing from my experience the potential for two things that are possibly related. One is the time it takes, Julie Hodge: Mm. Yuval Yeret: elapsed time, calendar time is quicker, which typically has some advantages for the customers, for the the business, less opportunity to to churn, faster recognition and all. revenue, there there could be different reasons. Julie Hodge: Mm-hmm. Yuval Yeret: And the other is typically when something takes less elapsed time, I would assume it also takes less effort. So it's more profitable for the organization. it is Julie Hodge: Yeah. Yuval Yeret: is are are those the reasons why speed was important for you? Julie Hodge: Yeah, I think all customers everywhere, no matter what software they're implementing, want it to be delivered faster, right? That's I think always been the case. And that's something with AI now that, you know, we can target, you know, not only the end to end but specific areas that might be what we call t you know, the tall temples. Areas where in an implementation previously we would we would see things slow down. And, you know, one of those areas for a lot of companies is customizations, right? A customization is something by definition, it's out of the box. And you know, you have to bring in specialists at that time to develop those customized solutions. So that was one area that we really wanted to, you know, target early. And you know, we have seen a lot of success in that area. and it's been interesting because over the last year and a half that success that we've seen has continued to improve as AI has continued to improve. And I think that's the case also that a lot of us are seeing across, you know, moving from POCs to having, you know, solutions that feel much more you know, well built out, thought out at this point, is that the speed is continuing to increase in these pockets where we've really, you know, spent time and focus, energy to increase. Yuval Yeret: So it sounds like AI is part of the current special sauce in your services organization. So probably can't share everything that you're doing with AI. But within those boundaries, can you give the the listener in broad strokes what's the current state? Like what what how is AI playing as part of that? a new customer journey, let's say, from it's a prospect that has approved the demo or whatever that looks like in your world, all the way to your system is working in production and running their billing cycle. Julie Hodge: Yeah, yeah. So Zuora re recently announced Milo and I can't go into too many details, but Milo is AI implementation support. And if you go to Zuora's website, we talk about, you know, the digital twin. And the idea is that if you're a company that wants to, you know, switch billing platforms, you know, or get off, maybe you're on Excel right now and you know you're growing very fast. Maybe you're an AI native company and you need, you know, billing, you would be able to upload all the content that you have and then That's your digital twin basically that you're creating of all your your your invoices, quote templates, product catalog, things like that. And then AI powered is getting you to a first cut tenant. And a first cut tenant is basically what it sounds like. It's it's a you know, a tenant that is, you know, you know, a large portion of the way towards being complete, taking in your your documentation. So that's kind of the goal is is speeding all of that up and using what the customer already has. And I would say that this has been you know, the the goal for a while now and it's great to see that we've, you know, we've announced it 'cause it's been something we've been working on and it's been great to see I can't talk too much about how great the results are, but we're seeing excellent results as far as how quickly our customers can implement. And I think this is something that not only, you know, new customers would be excited about, but any customers that might have been waiting to bring on, you know, a business unit. I think we've all been there where at one time you've had to work on you know, you know, another bringing in another business unit, data migration, all that can really be a hassle. so I think that's areas where AI can really get involved and reduce a lot of, you know, stress on both sides of the house, right? And not only reduce the time frame, but make things easier, reduce you know, human error and so forth and just make things easier. We've all been waiting for that. Billing and revenue is very complex. So I think this has been one of those areas that, you know, has been prime for AI to you know get involved and make things easier for everyone. Yuval Yeret: So maybe for the listener I'll try to paint a picture of how I understand some of the value streams that you're talking about. Tell me if I'm if I'm correct. So there's the product, your core product that is running a company's billing cycle. There's the process of onboarding a company into Your billing system from an existing maybe grassroots or other billing systems. They need to scale, they choose your system, they need to do some process of onboarding into the the new system. That process Julie Hodge: Yeah. Mm-hmm. Yuval Yeret: is what we're talking about here. And there's also the process that we haven't talked about so far, which is how you're working on improving this. AI-based onboarding process. And that's where I'm curious to go next. So you had this idea that you'd like to improve the the speed, dependability, make this tackle some of the critical critical long pole steps in the onboarding process with AI. Can Julie Hodge: Mm-hmm. Yuval Yeret: you walk me through? What was the process like of exploring whether I can really help their experimenting? How did you get to the point that it's currently a launched productize service Julie Hodge: Yeah. Yeah. Yuval Yeret: that is just running? Julie Hodge: Yeah. So I would say, especially if you're at a company right now where you're ready to to start this, I would look for any existing information that you have for what those long temples are. You know, go back if you have interviews with your customers after they went live where customers said this is where we struggled or we didn't have enough information or you know, we weren't we felt we weren't set up for, you know, success during testing. You know, gather all of that information and and that's where you start. Then you can find out where those those tall temples are, right? And then that those become your priorities, right? Not saying that you don't want to eventually, you know, have the entire, you know, pipeline, you know, fully AI auto you know, automated, but you need to start with certain steps, right? And then I think what's key is bringing together your product engineering group with your folks that are customer facing, right? Be it your your services folks, customer success. whoever else you've got day to day facing with the customer because they're the ones that are experts on implementing your product and you need to really loop them in on okay, so how do we solve for this, how do we solution for that? And it really needs to be a very, very collaborative, close effort. probably more collaborative than I think most organizations have ever had before with their professional services and product org. This is kind of changing the game when it comes to how closely those two groups need to work together to solution. Yuval Yeret: I think that's a great insight on the topic of forward deployed engineering. I I Julie Hodge: Mm. Yuval Yeret: I think the the key term in forward deployed engineering is that it's still part of the engineering organization. It's not a totally separate organization. when I look at I I've done work with multiple delivery professional services, customer. Oriented organizations and it's always a trade-off from what I've seen in the past. It's always a trade-off of do we keep product and delivery together and increase the cognitive load and create a lot of context switching for the same group? Do we carve them off? Separate them, decouple them, but then there's the price that you're talking about of losing touch with the reality in the trenches. How do you how do you keep both of these concerns alive at the same time? And I wonder, was there anything about AI and building this new onboarding capability with AI? They change the game for how to do this collaboration in in your experience. Is there some new secret sauce just from the fact that we can develop with AI that can break through this decades-old challenge for do we keep product and delivery together or do we keep them as separate silos? Julie Hodge: Yeah, I think that line is getting very blurry very quickly. And it's because, you know, some of the things you mentioned, AI is making it less of an impact. Cognitive load, right, is is not as much of an impact as it used to be. Context switching also a lot of that you can you know, use your your AI partners, as I like to call them, to help you out join that. So what we've seen is, you know, it's easy for easier and easier for people in G S to move to product, people in product, you know, with GS and for people in in services to build, right? Anyone can build now. so sometimes you can you can have someone who is you know boots on the ground all of a sudden be able to say, you know, I've envisioned this solution in my mind for years. This would help our customers go live faster. And then, you know, they can just go and make that a reality. And that's something that wouldn't have been possible, you know, a couple of years ago. So in in that way I think that those those boundaries are gonna blur more and more. And it'll like whatever it ends up being for deploy engineers, whatnot, I think those groups are kind of gonna be more and more brought together. Yuval Yeret: Yeah. Yeah, I th the what I'm seeing is the is the fact that you can create digital twins and you can create systems that are Safer to develop in, safer to work in, there's Julie Hodge: Mm-hmm. Yeah. Yuval Yeret: less technical debt, there's more robust testing, you have better context that's shared across people, so that more people that are, let's call it, satellites around the core engineering can do work in that area, and where some spare capacity that comes from the fact that the core engineering group. Can use agentic development, can actually go to the edges, can go forward deploy and be closer themselves to what the customers are doing with with the product. I was chatting with a cybersecurity CEO, Lior, a couple of months ago and and and he framed it in an interesting way. He said, up until now. Engineering coding was a bottleneck. It was a bottleneck, so we had to protect it. The entire process was designed around protecting this scarce resource. And one of the things that are changing right now is that it's closer and closer to the point, especially for AI native companies, where product engineering, the product teams aren't the bottleneck anymore. It's the ability of the customers to consume what we're building. It's the ability to find the right features. And one of the opportunities that we now have is rather than being on the defense, defending our product engineering teams, using the the classic processes of release planning and backlogs and roadmaps and estimation, all are designed to say no in a way. Wh which is a good Julie Hodge: Mm. Yuval Yeret: thing when you have limited capacity, you need to protect it, you need to limit the amount of work that is in process. But when you can say yes more often, you can go on the offense. You can do things I don't know if you watch the bear, for example, but the there's the notion of unreasonable hospitality where Julie Hodge: Mm. Yuval Yeret: you amaze, surprise your customers doing things that are almost unimaginable. Because there's a change in the physics or in the priorities of what you do. And that's possible. That's starting to be possible with what you're describing, with the ability to onboard customers very fast, to develop features very fast for the product. That could change the whole dynamic of what we're willing to do as a product company for our customers, what we want to do. all of that could be. could be changing. And and I'm curious whether now that you have this onboarding customer journey that is much faster, much more effective using AI, what's next in your world? Like what's the next level that you can share the vision for where to take a services organization with with AI? Julie Hodge: Yeah, I think it's continuing to improve. I think with the the AI changes, you know, and models updating so fast, a lot of things that maybe we like let's say standalone agents that we built six months ago, you know, if we didn't build them as self-healing agents, which I guess is the new thing, then those could be out of date. so it feels like, you know, with all the things that are happening in AI, you can only focus on a certain segment at a time. And then you might lose sight of a different, you know, batch of AI tools that you had developed and now those have changed. So it seems like there's always in the world of AI you know, still a quite a bit of backlog to to work on. Yuval Yeret: So how do you manage that backlog? Like what's your process as a services organization for continuously looking at opportunities and considering what's the next thing to to figure out, to to to point your AI Julie Hodge: Yeah. So the number one prior Yuval Yeret: efforts at? Julie Hodge: yeah, the number one priority is helping our customers to implement quicker. so that's, you know, guiding principle, everything needs to to focus on that. So that's how we prioritize all of our AI efforts. Yuval Yeret: Do you see a point at which Your ability to implement is outpacing your customers. Julie Hodge: Yes. I would say it's possible that that could already be happening and I think that's where you need to really change your your lens from thinking about things from your p you know point of view to the customer's point of view. And, you know, they have to deal with items such as, you know, change management, training. you know, testing can be a a you know, a heavy burden on organization. You know, if you're asking them, I need, you know, people in five different roles in your organization to take two weeks to test and provide feedback, that's hard for a lot of folks to fit in, right? They've got other things to work on. So I think that's when it becomes less about, you know, your product and how to implement more about how you can assist the customer with the specific things that they need to do. And there are, you know, AI ways that you can help them with that as well. Yuval Yeret: I love it. I love it. So I I think in your story, when I listen to it, I I I hear a couple of different rounds of finding the constraint, finding where in this flow are we currently stuck, where's the bottleneck? Doing something about that bottleneck if the bottleneck is a lot of manual configuration work. Using AI2 automate. If the bottleneck is the two weeks of UAT or user testing, whatever you want to call it, what do we do? It might not be in our product. We cannot do anything in our product to make it easier, but that's that doesn't mean we cannot do something that's outside our product boundaries. Or maybe we can with the product the deliver something that helps the customer do tests, evals, whatever. And after we do that, the bottleneck will shift to to something else. Julie Hodge: Yeah. Yuval Yeret: what what what are you using to know where is the the current bell neck? Like what are you measuring right now and how are you measuring it and who is measuring it to to know how to navigate? Julie Hodge: Yeah. Well I think we've got implied projects, you know, which are providing feedback on current solutions and then we can see how, you know, how bottlenecks might be moving at that point. And and that's again where, you know, the folks that are customer facing really need to work closely with products. 'cause they're gonna be the group that knows where exactly that that bottleneck is to say, okay, let's say for example, now testing, you know, we're seeing a lot of, you know, improvement with this aspect of testing, but maybe maybe not this aspect. You know, 'cause we really didn't think about that maybe from the customer's perspective. So let's go there and let's figure out how we can tackle this one next. You know, we tackle one part of testing but not not the other. You know, but that's that's the feedback we we need, you know, from once we start using these products with our customers. We need to provide that back to product and and work with them very closely. Yuval Yeret: And how much of that? process of finding the new constraint, focusing on it, building something that hopefully alleviates that constraint. How much do you use AI for that process? Julie Hodge: I think Yuval Yeret: How much do you use AI to build your AI solutions? Julie Hodge: It's amazing how many different components of AI are already built out. You know, like we for example we have one system where we collect feedback and then we need another system to put it in. But it's a custom, you know, AI, you know, app that now moves that. So there's a lot of different, you know, small bits of AI components that get thrown in places and it's It's just part of the day to day you barely even think about anymore how often you're plugging AI into these various things. It's just the way business is done now. Yuval Yeret: Has that changed your ways of working, your project management processes or the way you manage work in the services organization. Julie Hodge: Yeah. I think that like looking at data sets when AI's been phenomenal for that. So if you're looking at large data sets and say, Okay, I want, you know, you to find for this data set, you know, top five customer problems or whatever it is, that's where, you know, all these AI tools can really dive in and take all these data. You know, if if you've been a software company for a while, you have a lot of amazing data within your org and being able to layer AI solutions on top of that to get you know, detailed findings and aha moments that you haven't had before. that that's an an area where I think even even today we can still have more of those, right, with all the data that we have. Yuval Yeret: And so so maybe to to ask the question a bit differently, where where do you manage this backlog of AI opportunities for improving the the services experience? Julie Hodge: Yeah, just like a gear type tool. I mean for for for that backlog. Yuval Yeret: And has the way you manage that backlog changed since you've started to use AI for for the whole development and engineering processes for the services organization? I'm I'm sure there are changes happening in the product organization, but I'm really curious about the services organization. Julie Hodge: Yeah. I I would say previously a lot of product tickets would have went just to product and now it's it's collaborative. Right? It's it's being reviewed, you know, prioritized, you know, jointly with discussions. So that's where we talk about the lines blurring, right? And you know, whatever that new kind of you know, forward deployed engineers, whatever that's gonna be. but it's no longer product and silo. It you know, it's product you know, with all the folks that are customer facing, having that customer feedback and all of their, you know, years of expertise together. Yuval Yeret: En once let's say a ticket has been prioritized. Does it still go to the same product teams that worked on it recently? Does the you know, do services engineers sometimes pick it up and and work on it? What's like the life cycle of the those services improvement both? Julie Hodge: Yeah. Both. Both. Yeah. Yeah. Depending on, you know, strategy, capacity, all of those things. Yep. So lines are definitely blurring. Yuval Yeret: Cool. Any challenges that that raises that you're currently wrestling with? The fact that there are multiple teams touching these systems, or is it Julie Hodge: No, I think it's in just important to have detailed tracking. And I would say that's one thing where I know there's a lot of pressure that people are on to move faster with AI. I spoke with someone a couple of weeks ago who said he was doing four releases a day for his product Yuval Yeret: Mm-hmm. Julie Hodge: that he and I I was like, wow, that's that's that's a lot. That's very fast. but I still think it's good to have some sort of documentation and rigor. And you know, following, you know, these processes, yes, keep them lean, because we need to move very fast with AI. But that doesn't mean no no documentation, no controls, no governance. You still need to have those things. Yuval Yeret: Yeah. What what I see in I see a couple of different dynamics in clients that I work with. One is it's it's almost like a pendulum. one is we're doing spec driven development, we're bypoding with AI, we don't need to manage anything in our Jira anymore. We'll just use GitHub, TRs, whatever. It's Julie Hodge: Yeah. Yuval Yeret: a total anarchy. the A another dynamic is we're still gonna do the same process and it's tons of overhead and it feels so wasteful to manage stories and you know, have gates and approvals and all these conversations and events just because some methodology that we had used until now says so. And I I am seeing organizations that are On one hand Still thinking that it's important to manage the work, as you're saying, to maintain predictability and some level of governance and focus and priorities, but are also, while they're holding these values dear, are saying, you know what? A lot of the mechanics of how we start accomplished that in the past are irrelevant these days. So for example. stories, user stories. It was important in the past to manage those because they took us, you know, days, weeks. You know, it was a a lot of the capacity of the teams around that it was important to review. But when stories are flowing in minutes, you know, it doesn't make sense to manage them most of the time for a lot of these organizations. So I'm seeing a lot of the focus and I'm advising a lot a lot of my clients to shift a lot of the focus from stories to features, let's say, or epics, whatever Julie Hodge: Yeah. Yuval Yeret: you want to call that in in Jira or ADO. And there's there's this notion that one of my pet peeves is the notion of tiny teams. I don't know if you're seeing that in the organization, that you don't need the teams that you needed before. To develop Julie Hodge: Mm. Yuval Yeret: stuff because yeah, you can accelerate that, right? You're much faster. And there's this thinking that you can flatten the organization and everything can happen with two or three people, with their AI partners, and it's all good. And that's a it's possible, but it makes a lot of assumptions on what sort of access these people have to. the whole system, what knowledge they have, what digital twins they have, and a lot of time people are throwing away the the process that they need in order to manage these dependencies before those teams can actually be independent. So yeah so Julie Hodge: Yeah. The tiny teams can get you into trouble too. If they're not part of the the overarching goal, they can go off and develop something that isn't relevant, that's Yuval Yeret: Mm-hmm. Julie Hodge: already outdated. they can develop something that another tiny team is developing at the same time. Right? So you still have to have some overarching vision or an organization structure. And I do yes, detailed user stories for internal use, probably not really needed that much anymore. but you do need to know what what the request is, you know, what is going to be built, what is going to be added. And I think there's an area where if you're making a change that is small, then maybe tracking's not needed. But where Yuval Yeret: Yeah. Julie Hodge: what exactly does small mean? And I think that's an area where that's undefined. Is small five minutes? Is it an hour? Is it if I get it done in a day, it doesn't need to be tracked. Like where where is that boundary, right? Yeah. Yuval Yeret: And it could be five minutes, but Fable or Astra are wasting mil you know, spending millions of dollars in Julie Hodge: Yeah. Yuval Yeret: tokens. It's not just how much time it takes, it's also what's the token budget for it, which is a whole different conversation. Julie Hodge: Ex exactly. So five five minutes of humans work, then the AI writes the whole code overnight. So it wasn't that much effort, so did I not have to track it? Like I think there's still a lot of things around there that we know, we need to figure out and define for, you know, a good way of working with you know, all these different AI tools going forward. Yuval Yeret: Yeah. So I think the the message that I'm hearing is it's a process of continuous continuously learning and adapting both Both our process of onboarding and servicing our clients and finding new bottlenecks, l looking for where is the new bottleneck after we dealt with the last one and pointing our attention and our development capacity and our AI capabilities at that new bottleneck, as well as continuously improving our own development process, the way we do things so that our pace of addressing the ball necks and working through them becomes faster and faster. And and together these things create the exponential, you know, improvement that the AI native is all about. Julie Hodge: Exactly. I think that's that's exactly correct. Everything needs to be faster, but we need to go about it in a good way as well. Yuval Yeret: Cool. So I think that's a a good message to Julie Hodge: I mean taking on the entire thing is kinda daunting. but looking at those tall temples and areas, where your implementations tend to slow down and addressing those first and being able to, show show one or two wins, is important and just continuously build off of that success. And, you'll be able to find success in a small area relatively quickly. And that'll build, internal confidence as well that you know this is the path forward. Yuval Yeret: Cool. If if a listener wants to dive deeper into your perspective and point of view on how to use AI in services, where where can they find you truly? Julie Hodge: Yeah, reach out to me on LinkedIn. Yuval Yeret: Awesome. Awesome. Thank you again, Julie, for joining me for a conversation about how to use AI to improve the services and onboarding process in B2B. I'm Yuval. Join me next time for more insights on scaling AI from activity to impact. ## Source boundary These are the published show notes from the podcast feed. They are a starting point for discussion, not a verbatim record of the conversation. The transcript is machine-generated and may contain errors or unlabeled speakers. Check the audio before quoting anyone.