Your Board Wants ROI in Six Months - Your Contact Center Needs Eighteen - Salesforce

Is your board expecting AI ROI in six months while your contact centre needs eighteen? You're not alone — and that gap is producing some genuinely bad decisions.

In this interview, Rob Scott sits down with Gautam Vasudev, SVP of Agentforce Contact Center at Salesforce, to get into the real story behind enterprise AI deployments — why 95% of pilots fail to deliver measurable business impact, what the organisations that actually succeed do differently, and how to close the gap between board pressure and operational reality.

They cover the data readiness problem most vendors gloss over, the "triple penalty" of deploying AI on stale or badly structured data, a practical 12-month playbook for CX leaders, and why starting small isn't a hedge — it's the only argument your CFO will actually respect.

No vendor pitch. Just the honest version.

Transcript

0:07

We're hearing the same story from CX leaders right now across every sector and every size of organization. The board has seen the AI headlines. They've seen the cost savings Salesforce has published about their own deployment, and they want to know when do we get ours? Well, the honest answer, the one most vendors won't say out loud, is that the timeline expectations in most boardrooms don't match the operational reality of actually deploying AI at scale in the contact center. And that gap is producing some genuinely bad decisions. So today I'm joined by Gotham Vatadev, Senior Vice President of Agent Force Contact Center at Salesforce to get into the real story. Why deployments stall, what the organizations actually deliver ROI do differently, and critically, how you close the gap between board pressure and operational readiness without sacrificing either. Gotham, great to have you with us. How's it going? It's going well and a pleasure to be here, Rob. Thank you so much for having me. Yeah, I'm really excited for today's conversation. So, well, let's just dive straight in. I mean, let's start with something I I think a lot of CX leaders and contact center leaders will recognise immediately. The board has a number in mind, a six months maybe, maybe a year, and the contact center team has a completely different number in mind. You know, we've got data suggesting that 53% of investors expect positive AI ROI within six months, and yet meaningful deployment of contact center AI typically runs anywhere from four to eighteen months before it's really working. So you speak to enterprises every day, Gotham. How real is that gap and how much damage is it doing? Yeah, great question, Rob. And I will say the gap is real. So boards are reading a number of amazing AI stories, and there were also early movers in this space like Klarna who really optimized their strategy for AI. And that has set expectations on really best case scenarios. Um, and even with Klarna, we know they optimized their workforce. Again, rehired some of that staff to really achieve the optimal shape for what they need in the market. But on the ground, CX teams know what the state of their data is and the complexities of integration and the change management that they need to navigate through. So the honest framing is this six months is achievable. However, you have to be really deliberate about what that first use case is, what the ROI is, so that you can tell the board a cohesive uh story that starts from your first step all the way to your final chapter. And the mistake a lot of organizations can make here is boiling the ocean in the first month. And the gap isn't always the technology, it can also be scoping. When the board speaks about hey, what is our AI in the contact center goal and strategy? There are a number of use cases that you you can talk about, and that could be fully autonomous agents helping your customer, a rep assistant helping your rep as they're handling complex uh customer engagements, uh, or fully autonomous background agents that are trying to find out the voice of the customer and signals within your enterprise to then recommend more automation. So these can be entirely different projects and programs. However, the damage is done when pilots are set up to hit a six-month timeline without a clear definition of success on ROI. And then it can be a difficult conversation with the board trying to hit on that. Yeah, some really great advice there. Thank you. And you mentioned data, and I want to stay there for a moment because I think this is the bit that gets glossed over in most vendor conversations. I mean I mean, Gartner found that 61% of service leaders have a backlog of knowledge articles to update, and more than a third have no formal process to, you know, for keeping content current. So if your AI is grounding its response in stale, incomplete, or badly structured data, I mean what happens? You know, and be honest with us, you know, how many organizations are trying to deploy Agent Force right now before they've even sorted that out? Yeah, I'll be direct here, Rob. A meaningful number of organizations are trying to deploy AI before they've really sorted out their knowledge strategy. The ones who we've seen succeed, I want to just balance this response out with the organizations that are succeeding aren't waiting to get perfect data. But they're the ones who've constrained the initial scope to data that is highly trusted. Right? So don't wait for perfect, but you need to have a solid data foundation for the initial scope and use cases that you're driving. And by the way, when you are grounding your AI in stale or un badly structured data, you pay what I call a triple penalty. Right? First, your agents respond with uh answers that might not be satisfactory. And so customers are now escalating. So those are for your first two penalties, which is you had a moment with a customer and they really felt like, man, this engagement did not go well. I need to speak to a human. Now you're paying for the agent and the human escalation. Uh the human escalation is also more pressured because the customer is coming in with a bit of frustration. Uh, they want to move, uh, they want the issue resolved quickly. And the third penalty you're paying is what happens if that agent hallucinates because the data is stale or um, as I mentioned, not structured in the right way. So this is something you have to be very careful about. And we know that customers want to use self-service and autonomous responses, but this only works if the data is actually accurate. So you have customer demand, you've got the use case, you just need to make sure that your agents are responding in the right way. And by the way, that Gartner uh stat on knowledge backlogs does not surprise us at all. However, what is surprising is how few vendor conversations include this bullet. Right? You can have amazing uh proof of concepts and the first initial kind of exploration around this, but if you don't talk about the unglamorous knowledge uh effort that you're gonna go through, you're gonna have a difficult time through that first uh initial POC. And maybe I'll I'll conclude with this. For Agent Force Contact Center, you know, our we have a certain advantage because our solution is so backed by the Salesforce platform with channels, but that also includes the CRM of the customer context and Salesforce knowledge, which is backed by Data Cloud and it can uh pull in knowledge throughout your enterprise. So it's not only knowledge in Salesforce, but it's also knowledge to enterprise through data cloud. And this was very intentional because for every project, we first work with the customer to ground that agent knowledge so that your response quality can be much, much better. And a practical piece of advice for all of our listeners here today if you take your top 20 high-volume intents and map that back to knowledge articles, that'll give you a good sense, a report card of whether you've got an A grade, a B grade, or a C grade in your knowledge coverage. So let's talk about managing the board. Here's the practical problem. A CX leader hears the message about phase deployment, starting with agent assist, proving value in a bounded use case before expanding. That makes complete sense operationally, but they go back to their CFO or CEO who's been reading about Klana's chatbot doing the work of 700 agents, and suddenly we're starting with a pile that sounds a little bit of a hedge. How do you help CX leaders make that case internally? And what's the argument that gets phased adoption past the board? Yeah, and the Klana story was such an interesting story, and Klarna was one of the first movers in the market to really seize on the possibility of this agency era. Now, they've learned, and the industry has also learned from their first mover example. Um, they since that decision did rehire a uh portion of that workforce, and they are trying to drive optimization for now what is the best scenario for them. But what we can all learn from that is when you want to start with these programs and really deploy in a way that the board can be aligned with, you want to start with what you can deliver in the next 90 days and make sure it's measurable rigorously, and use those numbers for the next phase. Now, that is not a hedge, it is something that is going to build a foundation for your future uh roadmap. And speaking of roadmap, that is a great way to have a board conversation. So very practically the framing is within 90 days, we will have, say, autonomous resolution for these three query types, or we will have a recovery of um XYZ agent or human rep hours. And that is going to set up the foundation for the next phase of the roadmap where we are pursuing these next level metrics. So for a CFO, this then becomes a conversation of not hey, we are starting a pilot. It becomes a conversation of we are de-risking a broad strategic large investment. And every CFO loves the word de-risking. When you're doing capital allocation, um it's it's a strong story to say we're gonna have a brounded proof point use case uh where we're gonna show ROI and build up from there uh versus a big bang deployment that really has multiple KPIs and multiple stakeholders, and uh, you know, there is a lack of certainty around what the path is from uh writing the check, so to speak, to seeing the ROI. So Gotham, uh we've covered a lot of ground today, but if I'm a CX leader watching this, you know, I've got board pressure, I've got a contact center team that's stretched, I've got a data estate that's probably needs a little bit of work, and I've got a procurement process that's that's going to take longer than anyone wants to admit. I mean, give me the realistic picture of what the next 12 months looks like if I make the right decisions now. You know, not the vendor pitch version, the honest version. What should I prioritize first? What should I be patient about? And how will I know at month 12 whether I've actually made progress or not? Rob, that is a great question. And here's what listeners should be excited about. There is a playbook. Right? So in month one to three, you're doing foundation work, and it won't be glamorous. You're auditing top intents and cleaning up your knowledge and confirming your integrations, like your CRM integrations, and standing up agent assistants for your human reps. But if you're doing this right, your agents are getting better context on every interaction before you've even resolved one fully autonomously with an end customer. Months three to six, this is where you start putting those autonomous AI agents into your channels with end customers. And this autonomous resolution should be with high volume, low-risk queries. So things like order status, password reset, basic account management. You're measuring here containment, resolution rate, CSAT, and you're being honest about why the AI is failing or falling short. Months 6 through 12, this is where the compounding begins. This is going to be the most fun stage where you will start seeing that ROI from phase one and two, and then start expanding into more complex agentic workflows. So things like returns processing or substance uh subscription changes or financial adjustments. Um, you know, we've got customers like Compass Working Capital start seeing 6,000 agent or repars saved by leveraging this kind of autonomous uh AI resolution. But that is all the result of the foundational conversation prep work that you've done in months one through six. It is not the starting point. So, for this playbook, what do you prioritize first? Knowledge governance for your first top 20 intents. Right? You got this on lock, you're gonna have a much smoother path. Uh, you want to find an internal owner who is not just IT or not just in the service department. The technology will move as fast as your organization will let it. And then what do you want to be patient about? Be patient about full agentic autonomous resolution. There is complexity here for high-stakes interactions. It's not a 90-day bet, it is a journey, and your early wins will help you kind of keep the momentum throughout the journey. So, how will you know when your journey is in the right direction and coming to the close of its first chapter because you will keep evolving it? You'll know when your autonomous resolution is measurable and improving, your agent reps are getting your human reps, are getting interactions with full context and they are driving to similar resolution rates that they were even before the autonomous resolution piece, and with higher employee satisfaction as well. So from these kind of metrics, you'll know you're absolutely building something real. And you'll get this first signal at those milestones of 90 days, 180 days, et cetera. And one of the most important things you can do is use a uh knowledge discovery or self-learning knowledge tool. So for Salesforce in our Agent Force Contact Center, we have a self-learning knowledge module which can literally go through all of the conversations that uh your customers are having with your organization and find those gaps for either automation or knowledge, which is then going to give you much better clarity into those first 20 uh use cases you want to target. Well, you you've given some great advice there, and it sounds like a very helpful tool. So thank you so much, Gotham. And um, it's been a really honest and practical conversation. So lots for our audience. So thank you so much for joining me. Absolutely, Rob. Thank you so much for having me. It's been a pleasure. Thanks again. So, what stands out for me, folks, is a framing around speed versus readiness and the idea that the fastest path to genuine ROI isn't the fastest start, it's the right start. So getting the data in order, starting whether you can where you can prove value quickly and managing board expectations uh with evidence rather than optimism. So if you're a CX leader navigating this right now, we'll have a full write-up of this conversation on CX Today, along with the research behind the numbers we've discussed, and links are all going to be down below. And if you've got a perspective on this, if your timeline looks different or you found a way to bridge the board floor gap that actually worked, we'd love to hear from you in the comments or over on the CX Today LinkedIn community. I'm Rob Scott from CX Today. Thanks for watching,