🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.
Overview
📖 AWS re:Invent 2025 - Customer Story: AI Adoption with Salesforce & Amazon Bedrock (AIM267)
In this video, Shibani Ahuia from Salesforce and Sameer Vuvvuru from Capita discuss their partnership leveraging Salesforce Agentforce and Amazon Bedrock to build agentic AI at scale. Ahuia shares insights from engaging with 520 CIOs, noting the evolution from DIY AI approaches to agentic native integration strategies. Vuvvuru explains how Capita, the UK's largest public sector BPO touching 70% of the UK population weekly, implemented 44 AI use cases combining both platforms. He demonstrates measurable results including eliminating double-digit million pounds in contract penalties, improving win rates, and expanding contact center operations from 9-to-5 to 24/7. The discussion emphasizes democratizing agent development through low-code tools, enabling employees to build solutions using natural language, while maintaining security through AWS infrastructure and retiring AWS commitments via Salesforce purchases on AWS Marketplace.
; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.
Main Part
From Customer to Salesforce: Observing the Evolution of AI Strategy Among CIOs
My name is Shibani Ahuia. I'm going to share a few slides now that they're working. I'd love to introduce a friend of mine, Sameer, but I'd like to make fun of him before we start. The last time he and I were on stage together, I was in 4-inch stilettos and an evening gown, and he was dressed like a scrub. He's giving me a hard time that I dressed for my audience. I came prepared. I've worn my big data energy hat. Next time, Sameer, know your audience.
I'm Shibani Ahuia. I have been at Salesforce for one year. November 18th marked my one-year anniversary. Before being at Salesforce, I was sitting in your seat. I worked at TD Bank, one of the largest banks in Canada. I headed up their digital banking for a while and ran a number of their technologies across the globe.
How I ended up in the role that I'm in is that Salesforce loves to tell customer success stories. They invited me to an internal conference with about 450 employees. Mark Benioff was sitting where you are, and they asked me about personalization at scale and all the things we did. At the end, he asked me one question: what advice do you have for the company? I let it rip. I made it stand-up comedy. I was on a roll that I couldn't even stop making fun of the company. I said things like, it's ironic that you're helping me put my customer at the center, yet when I show up for a meeting with you, 22 of you show up. I went on and on. Next thing I knew, I was sure I was going to get kicked out. As I was walking out, someone said, "Watch out, they're coming for you."
Now I am on the Salesforce side, and I have one of the coolest jobs. I'm helping the company build out our go-to-market strategy for how we engage CIOs in the office of the CIO, CEOs, CFOs, and the entire C-suite. What that translates into is I get to travel the world and meet with enterprise CIOs of all industries, all shapes and sizes. I've hit 4,520 CIOs in the course of a year, and these are meaningful engagements. I share this because it relates to a couple of trends and patterns I've observed regarding AI over the course of the year.
Before I get into sharing any of those, I've been obligated by my legal team to share a love letter with you. This is the love letter from my legal team to say please make any purchasing decisions on generally available products today, not on anything that might be future-looking. I've gone through 520 CIOs and seen a lot of trends. One of which is the fact that AI has become the smallest, biggest word. It's become the fastest-growing thing that everyone needs to implement. They have to rush to get something into production. CIOs are in a hot seat right now. Unlike cloud, mobile, or internet, where they were squarely in the driver's seat and implementing it on behalf of their organizations, suddenly everyone with ChatGPT has become an expert at AI. Boards are putting pressure, CEOs are putting pressure. Everyone is putting all this pressure on CIOs to get something implemented.
What happened is that in the early stages of the year, CIOs would say no, I'm going to build shop. I'm going to build my own. Why I say AI is the smallest, biggest word is because I was starting to learn the differences between predictive versus generative versus agentic architecture, but the CIOs would just say no, I'm going to build my own. It would surprise me to say, you really want to build the component parts for agentic architecture? Do you actually fundamentally understand what agentic AI is and what the difference is between a co-pilot, generative, and agentic? It was adamantly this conversation: build versus buy. I'm going to DIY.
As CIOs and tech leaders started to understand that there is actually a fundamental difference between generative and agentic, the conversation shifted. The conversation started to shift to, well, maybe not build versus buy, but I'm going to start to look at maybe platforms like a SaaS platform. That's going to allow me to move with speed, and they're going to have templates and I can do it really quickly. A lot of my customer data is already within Salesforce, so let me leverage that because it seems to be a good smart place to be. There would be others that would have a different approach. They might say, well, no, I'm not going to look at Salesforce for AI. They're my CRM company. There are others that would say, I'm going to go to the Amazons of the world. I'm going to go to AWS. I want a cloud service provider. They're going to give me infinite scale and uptime, and I'm going to have governance, protection, and security. I'm going to be able to build my own customized agents.
I saw the debate shifting from DIY to build versus buy, and now it has shifted again. There's a world where both exist, and that's what we're going to share about today. I think the latest iteration, which I've started to hear more and more of over the past month and a half, is less about build or buy, less about whether it's going to be a SaaS platform or a cloud service provider. Now it has shifted to agentic native agentic integration. At TD, my Martech stack alone was 90 technologies. I couldn't imagine ripping all of that out and starting fresh. I describe agentic architecture as like the steel or the metal in Wolverine's body. No one's going to architect that from scratch. That's pretty tough to do.
Now I'm hearing a lot of CIOs say they need to look at the partners they've already invested in and the technologies they already have in their shop. They want to start looking at who is doing the work to build these agentic native integrations so they can move with speed. That is the better together story. Salesforce and AWS do compete in certain spaces, but that's not the case here. I am thrilled to introduce Sameer Vuvvuru from Capita, who is a friend of mine. Capita has become an agentic enterprise by leveraging Salesforce and Amazon Bedrock. Sameer is an example of how we have come together to drive agents at scale. When we were prepping for this, you mentioned that you have 44 use cases specifically that are linked to both Amazon Bedrock and Agentforce. Tell us a bit about those use cases.
Capita's Agentic Enterprise: Deploying 44 Use Cases with Salesforce and Amazon Bedrock
I'm absolutely thrilled to be back at re:Invent. I was at AWS until exactly this day a year ago, so it's great to come back as a customer. I've seen what Salesforce has done with AWS in the past from the inside, and now I see it from the outside as well. When I took on this new role to introduce agentic workflows into Capita, there was a lot of noise about which platform to use and who to engage with. Jeff Bezos actually said that in times of uncertainty, build on something that doesn't change. I know what won't change. AWS will be here. Salesforce will be here.
When I looked at what we needed to build across the multiple use cases we have, and we run everything from Transport for London to training nuclear submarine operators, my biggest need was to democratize the ability to build agents across my entire organization. We looked at 44 different use cases to get started. The first one is what was mentioned earlier about having a customer with 65 different applications. You could either rip and replace, or you could go over the top with MuleSoft and create that cover, then start building agents on top very easily. That's one we use all the time.
For our contact center space, the integration with Amazon Connect and that exposure into Agentforce is where we're actually building voice agents in Agentforce today and deploying them at scale using both Amazon Connect and VoiceCloud and Salesforce. We're using Mission Force for our defense applications. We're building local government in a box. All of these that used to take years and frankly tens of millions of pounds to build, we're now building in months and deploying at scale.
For the audience that may or may not be familiar with Capita, let me give a brief overview of what Capita is and what business we're in. Capita is the largest UK-based public sector BPO, business process outsourcer. If you look at the kind of services we operate, we touch 70 percent of the UK's population every single week, from running civil service pensions, which we went live with two days ago, to Transport for London, which I mentioned, to doing recruitment for the British Army. It's such a diverse use case that is very process heavy, and that's ripe for identification. That's why we're there because human labor is going to go infinite. It is going to be scalable. It is going to be available 24/7, and that's the mission we're after.
So Samir, here's another connection that we have like BPO. I used to live in the Philippines and I was setting up a captive BPO. It is a human intensive function or department. How are you looking at this? We've talked a lot about agents, human agents and digital labor coming together. How are you looking at that strategy being a very human-heavy organization? What's the strategy of bringing AI and your colleagues together?
In the immortal words of a guy called Phil, just do it. There's a lot of FUD in the market about agentification, that it's going to take jobs and reduce labor. We saw exactly the opposite when we introduced agents to our customer advisors who actually run that contact center. If you've ever been on a contact center and been passed around from department to department for hours and hours, you're super frustrated. But guess who's more frustrated than you? It's actually the customer advisor on the other end, because they did not have the tools to actually resolve your problem.
Once we gave our advisors those tools, they actually started pulling on us and giving us more use cases for agentification. That started this virtuous cycle where when we started with 10 advisors, all 1,000 plus advisors literally came to us and said, I want some of that. Because they do not want to be writing up transcripts. They do not want to be marking themselves. They don't want the supervisor looking over their shoulder. It's just a better life. They're more satisfied, they're happier, and that's translating into customer NPS improvements for our contact centers.
On that and the use cases that you've talked about, how are you measuring? Measurement was a hot topic. I hosted a dinner with CIOs, and one of the hottest topics that came up was that we're attributing a lot of benefits to AI, but tracing it back was very difficult because many organizations didn't have the baseline. How are you looking at the measurement of AI and are you telling the story?
There are three different vectors really. One, we run a lot of contracts that have KPIs. When we don't execute to those KPIs, we incur penalties, but when we exceed those KPIs, we share in the profits with the customer. Looking at that KPI and saying, hey, can we eliminate all of these penalties, we've eliminated double digit million pounds of penalties in a contract just by identifying the bottleneck. That drops straight to the bottom line. It's not new revenue, but it's even better because it's a 100 percent margin, and we share that with the customer.
Number two, we are embedding AI into every single bid that we put out there, every single one. We are seeing a win rate increase because we are cheaper, we are faster, and we're more flexible in what we're providing to the customer. Number three, just for us to become an agentic enterprise, we have democratized access to low code development of agents across all of our employees. We're actually measuring what agents they've built and what actions they're taking. At this point we are attributing a fiat currency to each one of them. But the customer and employee NPS is actually a good representation of how well they're adopting it and how happy they are with it, and it is going up.
The Better Together Story: Democratizing Agent Development and Measuring Real Business Impact
If I think about the trends and evolution that I've seen as it relates to AI, I love what you've just called out, is that not one of your measurements was how many FTEs are you reducing? None at all. This is a healthy way of starting to think about those use cases and looking at the growth opportunities or the expansion opportunities. I was actually presenting and speaking on your behalf saying that there was a contact center opportunity that you had a use case where your contact center was open 9 to 5 and suddenly when you implemented Agentforce to help support the contact center agent.
Such that the agent could handle the more mundane tasks, suddenly your contact center went from 9 to 5 to overnight being 24/7 because there were transactions that could be performed or cases that could be resolved by the agent. That's another way of looking at it to say that's a reverse benefit to your colleagues, the change management around adoption. That's a great use case, and I'm going to go even further. Historically, we used to have a customer advisor dedicated to an account because they needed to know everything about that account and how to operate within the constraints of that account.
We are now able to time division multiplex these highly capable agents across multiple different accounts just by providing on-demand knowledge base and real-time assistance using generative AI and transcription. The next step is going to enable them to do first call resolution much better.
You have a unique position here in that you clearly know Salesforce, you've got Salesforce in your ecosystem, you're familiar with it. We get you on stage, you help tell the story, but being only a year away from AWS and working at Capita, in your words, what are the benefits of this better together story? I'm a geek, right, and this is the only suit I wear. I would always love to go in and do assembly level code, but I don't want to and I can't.
One of the things that Salesforce has brought is the elevation of enabling everyone in my organization to basically go in and describe with natural language what they want to do for their customer and to test it out with their customer without my involvement or any of my team's involvement. The teams sitting over there are enabling teams now. They don't necessarily have to build for the customer anymore. All they do is advise and guide. They are the Yodas, right. So we have a bunch of Yodas, but we have a lot of Jedi Knights now, is the way I'd look at it. Every single one of my employees, Capita's employees, is a Padawan and hopefully will become a Jedi Knight in the future.
From a partnership perspective, the fact that you can now purchase Salesforce on the AWS Marketplace or the model optionality, I think as a customer I thought Salesforce was one of these big bad wolves that say you've got to have everything in Salesforce. You've got to buy every module in order to get all the benefits. But then to learn that it's an interoperable open stack that's allowing you to choose your models and in this case, Bedrock, being able to choose through Bedrock, financially for us it made a lot of sense because it runs on AWS and allows us to retire AWS commitments. It's like a no-brainer. Why wouldn't you do that?
You are the absolute experts in the Salesforce ecosystem, and you have tens of thousands of use cases that you're already identifying. Why would we recreate that? Why wouldn't we use what you already built and is available on Salesforce Marketplace and on the AWS Marketplace and have it securely governed with what I know is the best security platform on Earth, which is AWS.
Build on something that you know is going to be there for the long term. You are going to get inundated by everyone who's specializing in a single use case, but the beauty of agentic AI is that you can adapt to any use case as long as you have the right platform and the global reach, and that's what the AWS and Agentforce capability gives you. You can build your own. Start building. No one knows the process that you're doing, the job that you're doing better than you. Build for yourself. Then share it with your employees, and you will see the value.
Have fun. I literally have fun every day. I had a systems integrator come in the other day saying I'm going to charge you two million pounds to do this. I went and wiped coded it on Salesforce and I showed it to the systems integrator in like half an hour. This world is changing. Stop building perfect. If I can build, so can you. So have fun. Come visit us. You can build agents at our booth. Come check this out. Learn more about the partnership between Salesforce and AWS. If this is part of your stack, learn much more. Contact your account representatives if you have a rep. If not, come talk to us if you're interested. If you don't know what Salesforce is, I'm happy to talk about that. It's a pleasure. Thank you.
; This article is entirely auto-generated using Amazon Bedrock.
Top comments (0)