Joe Horsey 0:00:13.7: 

Thank you for joining us. We've got 45 minutes here for an enthralling conversation about HubSpot's financial modernization transformation. Now I've got no sound. Can you hear me at the back? Okay. Mike check. We're good. Let me introduce Janani. She's the Senior Director of Finance Transformation at HubSpot. I'm Joe Horsey. If you weren't at the keynote this morning - somebody slept in. Out way too late last night. I lead our solution consulting organization here. We're going to talk, we're going to have just a friendly conversation about finance transformation, how AI is impacting that, and hear a little bit about your journey. Why don't you introduce yourself and HubSpot? 

 

Janani Venkatesan 0:00:55.9: 

Thank you, Joe. It's an absolute pleasure to be here today with you all at Anaplan Connect. I'm Janani, and I lead finance transformation at HubSpot. Let me first start by telling you a little bit about HubSpot. I'll talk about my role there at HubSpot. HubSpot is a public company and our product is a smart customer agentic platform that helps businesses connect and grow. I'm incredibly excited our Spring Spotlight that just went live last week if any of you have been following. We have expanded across several agentic AI solutions across our platform, things like prospecting agents, smart deal progression, customer agents, all of which are helping our go-to-markets grow revenue and scale support for our customers. The most exciting launch is AEO, which is an answer engine optimization that helps marketers find insights on how their brand is performing across answer engines like ChatGPT. How does it show up in Claude? How does it show up in all the LLM answers? Which is where high-intent research is happening today.  

 

Janani Venkatesan 0:02:08.1: 

What I'm really excited about is HubSpot is not just launching AI or deploying AI in our products, but we're actually living the AI transformation internally at HubSpot. My role within finance is at the center of that. I am focused on transforming our finance function into a more automated and AI-driven function, where we are trying to automate all our manual processes. We are trying to use AI and predictive analytics for decision-making, and we are trying to build scalable, sustainable systems and connected systems. Finance transformation is never a solo effort. As you all know, we work very closely with our tech partners. We work very closely with our functional finance stakeholders to really drive sustained impact in the org. 

 

Joe Horsey 0:02:58.8: 

Obviously, hypergrowth, massive changing landscape with AI, changing business processes. How far are you on your transformation? Give us a little bit about where you started maybe, where you are in the journey, and then we can talk a little bit about where the future goes in your finance transformation. 

 

Janani Venkatesan 0:03:14.3: 

Sure. We started finance transformation at HubSpot a couple of years back. Let's first start with this, finance transformation did not start with AI. It started with a much simpler and urgent problem. The role of finance has been shifting from reporting to actually driving performance and outcomes, making business decisions, staying very close to business and driving real business outcomes. Finance also always has a baseline responsibility. Finance is responsible for producing accurate financial statements, is responsible for audit-ready books, compliance, risk management and controls. All of this is not going away. Our finance teams were spending a lot of time on these processes. These are all manual. There is a need for us to make sure the right systems are available to any of our teams to perform better in their roles, more effectively in their roles. I think with the increasing complexity in the business, what we are noticing is there's so many tools that our businesses needed to operate effectively. The tools have just been growing organically as the business scales. This definitely results in fragmentation at some point.  

 

Janani Venkatesan 0:04:32.2: 

So there is a need for us to connect systems. There is a need for our leaders to get ready to use information, timely information for decision-making. That really fueled the need for finance transformation. Now with AI, I would say that need has just compounded. We all know and we have been hearing multiple times that AI needs data. AI needs context. AI needs connected business processes and systems for it to work. Finance transformation helps connect the dots across the organization, helps us rethink about the way we scale our systems, rethink our data transformation, building on the intelligent data layer that is needed for AI to work effectively.  

 

Joe Horsey 0:05:17.9: 

You talk about this tool fragmentation, right, because that's what we see a lot right now, is AI experimentation everywhere. It's easy. I don't have to be an expert in a particular domain or discipline to activate AI. How did you start to think about wrangling all of that fragmentation across, so that you could start to activate AI in a productive way for the transformation? I'd be curious. 

 

Janani Venkatesan 0:05:38.6: 

That's a great question. I think, first of all, for us, deploying AI is not the goal. Right? It starts with foundations first. We start by looking at our current state process. What is our end goal? What is our business objective? What are the gaps in the current process? What are the gaps in the team? What are we solving for? So always tie it back to a business outcome. Then we go through the entire process, the end-to-end process, looking at where the gaps are. We look at all the upstream, downstream dependencies. This really helps us design the right systems, make sure we're looking at data, make sure we're unifying data in the process. We are consciously thinking about how to structure this data, how to get the data quality right, and how to get the data consistent, the definitions consistent for AI to work. I would say deploying AI by itself is not the goal. Right? We are solving for a business problem and we are looking at ways to make AI work effectively. We are looking at ways to embed AI, to help us solve that goal.  

 

Joe Horsey 0:06:46.7: 

We were talking before we started here about the organizational structures are changing. You have a finance team. Maybe you have an HR team, a workforce team. Now the way that your transformation is going, it might have to challenge the way you've organized. Right? We talk about CoEs with Anaplan. When these businesses processes that are being infused with AI start to cross functions, how do you think about the organizational evolution that needs to come? It's not only technology. To your point, it's not AI for AI. It's data, it's business process. Fundamentally then it's the people you have to activate that business process. How do you think about that at HubSpot? 

 

Janani Venkatesan 0:07:22.6: 

I think going back to the foundations that I mentioned. Our transformation spans across process transformation, systems transformation, data transformation, and now a lot of focus on AI adoption and impact. We always think about processes more holistically, which is cross-functional. Right? If you pick any process in the organization - and right now I'll talk about a massive head count transformation project that we're all going through with Anaplan, where we are trying to build connected planning. A head count operation or a head count planning involves several teams. It involves the HR teams. It involves the ops teams, sales planning teams, customer success planning teams. It involves finance, that brings in the final, so the fruit for the head count data and the budget for the head count. We need all these teams to come together. Having disconnected systems just causes several issues for leaders to make decisions because every team obviously has their own sources of truth and everyone has their own objectives within their functions that they're solving for. Eventually, this results in disconnected data and inconsistent data in the org.  

 

Janani Venkatesan 0:08:38.5: 

I think what we truly need to solve for is the end-to-end process, redesigning and thoughtfully redesigning the end-to-end process, solving for connected systems. Making sure these systems are integrated and there are very clear hand-offs. Making sure there are consistent data definitions. I just cannot reemphasize on the consistent data definitions, data foundations, data quality because that is foundational for AI to work. Our approach has always been pods of cross-functional teams working together for any transformation initiative that we pick up. Always focusing on data right from the start and not thinking about data much later in the progress. For example, data reconciliation, data quality, definitions are all defined right from the beginning when we start a transformation journey.  

 

Joe Horsey 0:09:32.7: 

Super-interesting. When you think about bringing all of these different stakeholders that have a portion of a process that you're trying to evolve with AI, how do you deal with the pushback? All of them, 'Well, I'm doing my part of the process and it's really good. I don't know what you're doing over there and I don't know what you're doing over there.' How did you get buy-in from the teams? What was the carrot at the end? Like, 'Hey, trust us, this is hard work and it's going to change the way that you work, but here's the outcome.' I'd be curious on how you handled that part of the transformation.  

 

Janani Venkatesan 0:10:01.3: 

Yes, I think if you look at transformation, I would say more than 50 per cent of transformation is finally about change management, user adoption, and that goes back to your question on how do you get that buy-in? I think it all starts with clearly defining the end goal and the business outcome. That's foundational. You have to first find out, what are you solving for? Make sure everybody understands that and everybody anchors to that end goal or business outcome. Unless you have that, everybody is trying to optimize for their own individual process outcomes or functional outcomes, and that's when things fall apart. Right? Anchoring to that one, common business objective is very important. I would also say, especially with AI, people play an important role. Unless you upskill people, unless you bring them up to speed with the technology, adoption is not going to work. Making sure you think through change management right from the beginning and making sure you bring people along the journey is so critical for a transformative solution to work.  

 

Janani Venkatesan 0:11:14.6: 

Also I would call out that having an executive sponsor is so important. Sometimes we really underestimate this and we might just begin a transformation journey without alignment at the executive level. You don't get a sponsor, then obviously, you don't have alignment. That's when things again could fall apart. Right? Having clear executive sponsorship, clear objectives and goals, and making sure you're bringing people along the journey is so important.  

 

Joe Horsey 0:11:46.0: 

That's super-interesting. Your transformation is in flight right now. You talked about this connected planning around head count modernization. What are some of the other things you see on the roadmap of your finance transformation? That's a broad umbrella. Give us a little peek into the future before we really dive into some of the more AI-specific questions I have.  

 

Janani Venkatesan 0:12:03.2: 

Sure. I'll talk a little bit about our finance transformation journey. Like I said, we're always focused on process first. We are currently going through some major process transformations and we'll continue to do that. I think foundations, we'll always do that while we're doing AI and other stuff. I think there's always something going on around foundations. On the data side, one thing that we are really focused on at the moment is unifying all our data assets. I would say this is true for structured and unstructured data. One, establishing a source of truth within systems. For example, Anaplan or any other system that we use. So making sure we have connected the dots within the system and we have a single source of truth of data from that system. Then going beyond and also connecting the data within, across all these systems, like Anaplan and NetSuite and Workday. They're all so interconnected and we want to make sure we're connecting and unifying all these data sets. Then making sure the definitions are consistent, the data quality is good, there are clear owners of the data.  

 

Janani Venkatesan 0:13:11.6: 

Then now with AI, I think explainability and consistency of the data is so critical. We cannot have the same AI input have different answers every time. Especially for finance, it is so important to have accurate answers and audit-ready answers. So it is important to have visibility into what the model is generating. I think we are definitely working on the explainability layer, semantic layers, defining the context. That is so critical. Also security and governance. Making sure that whatever tools we are using are all compliant, and now especially with so many tools that we have, we want to make sure that we're always watching out for security compliance. Also making sure we're defining very clear guardrails on how to use the output of that data. More recently, we have started experimenting with a lot of advanced analytics, ML models for revenue forecasting, cashflow forecasting, predictive models, anomaly detection, OpEx forecasting assistants. These are some spaces that we have started piloting initiatives. We are very excited about our finbot, which is our custom-built solution, which is built using the open AI LLM. This really brings together the structured and unstructured data because a lot of context also lives outside systems, like emails, Slack messages, documents. We also want to bring the insights from these docs and help our finance teams get answers to questions in plain English. That's our vision for finbot and we're absolutely excited about our finbot as well. 

 

Joe Horsey 0:14:59.3: 

That's awesome. When did you launch the finbot? 

 

Janani Venkatesan 0:15:02.8:  

It's been there for almost a year now. More recently, we have been piloting it for specific use-cases, like risk collection, for a QTC, quote to cash teams. They're all focused on collecting from our customers and giving them a risk score for invoices, so they know which customer is likely to default and go after them. So things like that. Then we're also using it for [?BV 0:15:30.1] commentaries, flux commentaries, like where our finance team spend so much time and finding out budget versus actual variances and explain them. So we are creating AI summaries for these variances using data from Anaplan, and connecting that with our NetSuite and other data from other systems. 

 

Joe Horsey 0:15:47.4: 

That's super-interesting. I think one of the things I'd like to understand too is the adoption curve on that. It's great, and you define it, and of course you need this finbot and it's going to help you do all these things. What was the challenge of adoption, to get your people to change the way they looked at their jobs? We see it in every customer I go to, even internally at Anaplan. We roll out these AI tools and agents to make us all more productive or change the way we work. I'm just curious on was there a struggle with adoption? Was there a struggle to, an inflexion point where people were like, 'Oh, I get it now'? I would be curious as to how that played out. 

 

Janani Venkatesan 0:16:24.2: 

Yes. I think there are multiple things at play here. One, I think we spoke about tool fragmentation a bit with SaaS. Now, with AI, I would say the AI tools pro is even worse.  

 

Joe Horsey 0:16:39.3: 

Yes, it is. 

 

Janani Venkatesan 0:16:40.1: 

Right? There's so many tools and I think what makes it more risky is with AI tools pro, the intelligence is at risk because with fragmentation of the context and data layer, the intelligence that the AI tool is going to produce itself is at risk. Obviously, that's a big risk with AI tools pro. One thing that we as finance transformation team members try to provide clarity and guidance to our finance team members is help them pick the right tools. There's always a better tool. There's no perfect answer. Right? There's always a better tool. I think while we continue to learn and experiment, we also need to pause, synthesize our learnings and make decisions. Be willing to take risks and make decisions and move on. Right? Otherwise you're continuously left experimenting with the number of AI tools that you keep getting every day. I think what we try and do is guide our users to pick tools and start piloting and start scaling once we have seen success.  

 

Janani Venkatesan 0:17:46.8: 

With tools like finbot, for example, there's definitely an adoption curve. I think there are a few things that we need to do. One, build trust in that data, which is where a lot of the work that goes behind in making sure we have the [?deterministic 0:18:04.0] core engine, calculation engine, which I think… 

 

Joe Horsey 0:18:07.5: 

Someone was at the keynote this morning.  

 

Janani Venkatesan 0:18:08.8: 

Yes, you spoke about it in the keynote and that resonated with me so much because that's exactly what we're trying to do with finbot, making sure that every time the user asks a question, finbot is able to give them accurate answers. That builds trust in the data and that definitely helps with usage and adoption. The other thing that we do is make sure we work in pods. Every pod of ours has a domain expert from the business, has finance transformation, has tech team members, engineers. We're all working on the solution together, to make sure we're actually solving for a business problem. Also we have change champions from the business who help us with adoption because you really need someone who understands the business, who's close to the process to help you drive that change. Otherwise it's just a technology installation and you're not really solving a problem for the business.  

 

Joe Horsey 0:19:05.5: 

It's super-interesting because you have to balance this bottom-up experimentation and get the team bought into, 'Hey, you're a part of this. Tell me what pains you have in your day-to-day.' Also tying that to a top-down, executive-sponsored initiative, so that you can make sure that it's a scalable, business-benefitting experiment. Right? It's an interesting dynamic. I want to touch on this fragmentation of tools because inherently, the converse of fragmentation is let's consolidate on one platform or two platforms. We know that the world is going to be fragmented to key data platforms, agentic platforms, LLMs. How do you rationalize and how does HubSpot think about those investments when they're selecting a platform, whether it's Anaplan or a Snowflake or a Databricks, to be part of this architecture? 

 

Janani Venkatesan 0:19:59.7: 

That's a great question. Our AI strategy is focused on three pillars. I'll talk about how we think about tooling in each of these pillars. The first pillar is enable, where we are actually empowering our finance team members to use tools like ChatGPT and Claude, Gemini, to automate their day-to-day tasks. Here we allow teams to experiment, learn, and pick the tools that work best for them. These are simple use-cases, like automating their one-off reconciliations or summarizing presentations or writing narratives. These are day-to-day tasks and we want to give them the options to pick the right tools that work for them. This is a space where there's a lot of experimentation and learning happening. The second pillar that we focus on is deflect. If you think about a finance team across all functions, they're spending a lot of time answering questions for internal employees or for their customers or for vendors. For example, payroll questions, benefits questions, travel questions or invoice-related questions. There's just so much time spent in answering questions. Here we look for enterprise tools, enterprise search tools that can look at all the knowledge articles and help automate these responses.  

 

Janani Venkatesan 0:21:27.8: 

Here, we want to consolidate and bring everyone to the same platform, because we want everyone to have access to that information and be able to ask questions and get the same response. Also we want to connect the enterprise tools to all our systems, like, for example, payroll query. We want to connect it to Workbase, so it has the relevant context for the tool to be able to give the right response to the user. Then our third pillar, which is automate, is about large-scale automation of workflows and conversational AI for insights. Here I spoke about finbot, which is a conversational AI for insights. Again, here we want to make sure that we are consolidating the tools. We are making sure that the data is well-governed and is accurate, trustworthy. That is so important for finance users. Here is a space where we guide the users on what tools to use. For automation workflows, we follow a hybrid approach. We use AI embedded within SaaS applications, like, for example, Anaplan has a lot of AI features. We definitely want to explore that. We use enterprise orchestration platforms, like Workato One, which are connected across multiple systems, like Anaplan, NetSuite. They connect to every system of ours and we are able to get large-scale workflow orchestrations automated using Workato. So it's a hybrid approach that we follow for larger-scale automations. 

 

Janani Venkatesan 0:23:04.1: 

The interesting thing is now with AI, in the hands of every employee, I think this is such a huge shift in strategy for us. Right? AI is no longer centralized. You no longer need a tech team to come and build the automations for you. There's so much you can do by just putting AI in the hands of your employees. That's really our enable pillar, where we are trying to upskill our team members to start using AI. Training them on how to write prompts, training them on how to build agents. More recently, we are shifting from just using AI to having them build agents because there are so many low-code tools that are now available where people can go build agents with simple drag/drop functionality. We are slowly trying to make that shift from just using AI to actually building and delivering real value.  

 

Joe Horsey 0:23:56.4: 

That's amazing. I think, invariably, what that does is it future-proofs your workforce because you're giving them new skills. Right? If I was just a finance analyst and you're asking me to build bots and agents, I'm like, 'No, I do the finance. I do the numbers stuff.' Right? Now you're giving them skills that's going to help them grow in their career. I think that's, by side effect, future-proofing any attrition you might have or your future labor and your future skills in your workforce, which is super-interesting. I wish that at Anaplan we would just give us all the tools, like we're experimenting, but within very tight guardrails because eventually someone needs to pay the bill every month on all these tools. I want to use that as a transition to when you think about ROI, the cost-benefit analysis, how much productivity are you gaining or how is your team changing? If you read the news, you saw that Oracle laid off 30 per cent of their workforce because of AI efficiencies. How does HubSpot think about the ROI, the business case, and beyond productivity? I think you're doing more than just driving efficiency.  

 

Janani Venkatesan 0:25:01.6: 

Yes. I think a year back probably if you were asking me the same question, my answer would have been very different. I think we are shifting from measuring just AI adoption to measuring business outcomes. When we started off, we were only measuring how many people are using the tool? What percentage of the company is using the tools on a weekly basis? Now we are switching to business outcomes. For example, we have a payroll query deflection agent. What we are measuring is what is the percentage of queries that the agent deflected? What time is that saving for a payroll analyst? Now what are they doing instead with those hours? The goal for us is not to just cut manual hours, but redirect that to more value-adding and strategic initiatives. If I talk about a forecasting agent, we would probably tie that to a forecasting accuracy improvement over a baseline. We are always trying to connect it back to the business outcomes. Overall what we also track as a core metric is the time savings. At the end of it, that's what we are all aiming for with AI. Right? Save time, so that we can redeploy that into more value-adding stuff.  

 

Janani Venkatesan 0:26:16.6: 

We do have a core metric of tracking time savings by finance cost center, so that we know where we are headed. We also know where there are significant shifts happening. Where is there a need to upskill, reskill people? Where can we shift to more of finance being strategic advisors versus a team that was just reporting on something? Now they're actually helping shape performance. Overall, my answer is I think the best metric is the business outcome itself.  

 

Joe Horsey 0:26:49.0: 

Yes. I think it's interesting you say, well, what is that payroll analyst doing with these hours that they've freed up? They're building agents. Right? It's a self-fulfilling thing. They're activating AI. Maybe I'll ask one more question and then we have till 2:40. Right? Okay. So let me ask you one more and we'll open it up to maybe some Q and A. All right, you're on this journey. It's ever-evolving. You say a year ago I would have said this. A year ago, and what you know now, what are some of the lessons learnt? What would be some of the advice you'd give to the audience who are trying to either get the buy-in or think differently about their transformation? Just any advice that you could give or lessons learnt.  

 

Janani Venkatesan 0:27:27.8: 

There are a lot of lessons. 

 

Joe Horsey 0:27:29.2: 

I know, there's a lot. You've given already a lot, how to foundationally think about this, but yes.  

 

Janani Venkatesan 0:27:34.7: 

Yes, I'll give you maybe a few that are top-of-mind for me. I think one thing that I've already said, but I will just say it again, is foundations first. Right? Don't go with the intention of just deploying AI. That cannot be your goal. Always start with a well-defined business problem statement and always start with an assessment of your current state process. Clearly define where you want to take that process. What are the gaps? What are the opportunities? Then think of AI as an enabler for you to achieve that. I would say watch out for AI tools pro early because quickly it could turn into something that is unmanageable and a lot of lack of clarity for the team. I think it's important to partner with your tech organizations and have a very clear AI tooling strategy. Also provide that clarity to the finance users. Then I would say don't get stuck in this learning and experimentation. Be willing to take risks. Be willing to fail. We have made wrong choices on tools. We have had wrong pilots or failed pilots. All of that I would say is a learning. You're learning something that others didn't. Right? So you are still ahead of them in your transformation journey. I think that's what matters, trying and learning.  

 

Janani Venkatesan 0:29:03.7: 

I want to quickly reference something that our CEO says and something that resonates with me very closely. Our CEO talks about going from map-readers to explorers. That is so relevant in this age of AI. Map-readers are those who need clarity, who need a map. Right? Explorers are those who try even ambiguity. They can just go learn, experiment, figure things out, and they adapt as they go. That's exactly what we need with AI. No one has perfect answers. All of us have to learn and we have to experiment, and that's how we make progress with AI. I would say if you haven't started with AI, you should absolutely start now. If you're in the early phases with AI, just focus on quick learning, delivering, at least in the initial phases. Your goal should be to learn. Learn quickly, iterate, and improve on that. If I were to summarize in one line, I would say don't wait for the map. Start exploring and the part would reveal itself. 

 

Joe Horsey 0:30:14.9: 

I love that analogy. Well, look, thank you so much for sharing your insights. Obviously, you have a great foundation, that there is not just a tooling or an AI, there's a transformation. There's a business outcome. You think about the people. You think about the process. It's really exciting to see what you're doing. I'm sure it's going to be massively successful and drive growth for HubSpot. How about we open it up to questions from the audience? We've got about ten minutes. I see someone right here and then I see someone all the way at the back too.  

 

Audience 0:30:54.7: 

What are your long-term goals or metrics for finbot a couple of years out? 

 

[Brief aside comments regarding microphone 0:31:04.6] 

 

Janani Venkatesan 0:31:15.6:  

That's a great question. I am really excited about our vision for finbot. What we want to get to is a connected ecosystem of agents. I was so happy to see that in the keynote slides, one of the slides that you shared because it resonated so much. It aligns with what we are trying to do at HubSpot in finance. We want to build a connected ecosystem of agents. Think of a situation where, let's say, revenue is falling behind plans in one of the quarters. Think of a situation where there is an agent, a variance agent, for example, that identifies this already, almost in a real-time basis. Then this triggers a driver agent, which goes and figures out why that happened. Maybe it's in a particular segment or maybe pipeline conversions were low. It figures out the reason. Then it triggers a forecasting agent that goes and updates the outlook. That then goes and triggers a scenario-planning agent, which will go and figure out whether we need to adjust targets or adjust our marketing spends to bridge the gap or whatever that is. There is a narrative agent that goes to our leaders with an explanation and recommended next steps.  

 

Janani Venkatesan 0:32:39.8: 

Wouldn't it be fantastic to have a world where have all these AI agents talk to each other? That's eventually what we want to get with finbot. We want to build all these agents. We want to make sure our agents are connected to our underlying systems and platforms, and we have unified data, we have well-governed data, and the agents are able to answer questions for our finance team members.  

 

Joe Horsey 0:33:05.7: 

That's awesome. 

 

[Brief aside comments regarding microphone 0:33:10.0] 

 

Unknown Speaker 0:33:19.1: 

Right, who had their hand up? 

 

Joe Horsey 0:33:23.9: 

There's one, yes, I think there's one all the way in the back.  

 

Audience 0:33:31.1: 

So when you're saying about the, when you decided tool and go forward, and then you decide, okay, got fail it because that tool is not driving our success, so how do you convince the business or else what would be the better solution here, taken at that time? What should we be considering in the PoC part? 

 

Janani Venkatesan 0:33:55.6: 

I hope you can hear me - yes. That's a great question. I think it also comes from the organizational culture and leaders leading it by example. That experimentation is expected and failure is okay. Right? I think we need that. I would say that in our company, there has definitely been a lot of focus on experimentation, but making sure that we're also learning from it. So it doesn't mean you can just keep experimenting with tools, keep failing with tools. You're wasting time by doing that. Making sure that you're able to take learnings from that, quickly synthesize learnings from that, and quickly pivot. Right? So if you've learnt that something is not working well, being courageous enough to say that this is not working, and also being courageous to quickly pivot your decision. Just because you already went that path, it doesn't mean, success is when you complete that. Right? So success is also when you're able to quickly say, no, and figure out this isn't working. Make a decision to change your path and go figure out another tool or maybe another pilot with some other use-case because that isn't working.  

 

Janani Venkatesan 0:35:15.6: 

We have had those situations. Obviously, it in the interim causes some constraints on your resources because, obviously, you have spent two months on a pilot that didn't work. Those two months are gone, you're not getting those two months back. I think what is important is the learning from those two months and how you can use that to expedite your next use-case. I can say with confidence that none of our failed pilots have gone with no learnings. We have learnt so much from our failed pilots and it has definitely helped our next set of pilots, both in terms of our resources, like learning something or picking up a new skill, or I would say even us appreciating a new tool better. Right? I would say that it's definitely a learning. To answer your question, you also need to ensure there's executive alignment, sponsorship, willingness to experiment and willingness to reward failure in the organization.  

 

Audience 0:36:27.9: 

This is related to AI adoption. You had metrics for AI utilization, but you switched over to business outcomes. Can you share with us the rationale, why you switched over? Also, why not both?  

 

Janani Venkatesan 0:36:44.1: 

It is actually both. Earlier, I think in the prior I heard it was just AI adoption, utilization, but now it is definitely both. It is not just business outcomes. We also keep, continue to track AI adoption and utilization because that is important as well. Right? You need to start with that. Adoption is important first, and then you're looking at business outcomes. So to answer your question, it is definitely both. The reason we wanted to start measuring business outcomes is you need to understand if AI is working. People are using AI, but is it actually working and is it actually driving impact for the org? You cannot do it unless you actually tie it to a business outcome. I'll give you another example. We are just implementing an AI-powered intake workflow for procurement. I think most of you who are in finance will know that procurement is such a cross-functional exercise and there's so many reviews in the process, and there is an important or there is a requirement for it to be more simplified.  

 

Janani Venkatesan 0:37:52.7: 

We are launching an AI-powered workflow. How do we measure it? We will measure it through time-savings, through purchase end-to-end cycle time reduction, or even through an employee experience, captured through a survey, like how was your experience purchasing something before and how was it now, after you have this workflow? 

 

Joe Horsey 0:38:15.1: 

I think we probably have time for one more, if there's one more up there. I don't see any hands. Janani, thank you so much. This was an amazing conversation. Thank you so much for taking the time to share your story. 

 

Janani Venkatesan 0:38:29.2: 

Thank you. 

 

Joe Horsey 0:38:31.1: 

Thank you.