Intro to Agent and Application Lifecycle Management (ALM) | Dreamforce 2025

And without further ado, introducing myself, Trayvon Jackson, lead technical architect at Salesforce. And joining me on the stage, we have Diego, chief marketing and technology officer at VentureCrowd. And what we're gonna talk about today is I'm gonna set the foundation for our application lifecycle management framework, and then we're gonna drill down into the specific tools and what makes that possible today. And then this is gonna set up the foundation for Diego to talk about the venture cloud story and how they're building, testing, and deploying agents on the Salesforce platform. And then we'll close with any final remarks. All right, so getting started. So developers play a key role in delivering innovation with companies like yours today. And we're seeing so much chatter around the changing role of the developers and the age of AI agents. And we're also hearing about how AI is requiring more software developers and not fewer. And this is ultimately bringing up the ideas around how AI is changing that landscape around how we traditionally would do things, how we build, how we deploy, how we interact with things. And this is bringing up an immense amount of pressure to build those generative AI experiences. And in fact, we're seeing 86% of IT leaders now expecting generative AI to play a prominent role in their organizations. And what that boils down to is two points as we deliver on that vision for enterprise AI and how that's becoming challenging, which is what the business wants, which is AI. First products, and then how that translates to developers to continue to maintain that test quality, to ship continuously and to iterate on the fly, and then do more with less. And I can say that that's no easy feat as a past developer in my time, but how does Salesforce simplify that vision and make that possible today? We do that with our agent and app lifecycle management process, where customers can innovate faster to accelerate your journey to an agentic enterprise so you can continue to optimize developer productivity. You can build agents and apps with AI powered tools and accelerate that continuous innovation. And then you can deploy quality at scale. And then lastly, customizing your solution to better meet the needs of your business as you iterate amongst your scaling. Now, what does that framework looks like as we break that down into stages? And how does Salesforce simplify that process? We, we break that down into five stages. Ideate and plan, build, test, deploy and observe. And we always keep security at the forefront, which is wrapped around our governance model. But as we break that down into specific tools and what makes that possible, this is what that looks like. And Diego's going to do an amazing job at talking through how Venture Cloud today is using these very own tools to make, build, test and deploy AI apps and agents today. But as we focus in on like some of the key tools that makes that possible today, let's start with sandboxes. So sandboxes is our secure isolated environments that allows you to test safely and release faster and secure development environments. And this decreases the risk factors of testing and production and tampering with data that's live already. And we're seeing a 29% reduced IT downtime and leveraging soundboxes today. And this is going to be foundationally speaking how we get started when we ideate and plan and then test against our agents today. Next up is the build stage. So Agent Builder is our AI low code tool that allows you to set up and deploy agents for any job and you can describe and configure any agent. And what makes that possible is you have a list of pre built skills where you can action against prompt templates, flows, apex and APIs. And then moving against that you have AgentForce testing center. So what happens when you move from building agents to now determining how these agents are gonna behave once you go into production. So AgentForce testing center allows you to deploy agents confident and you can upload test cases where you can generate to agent quality and then also compare those scenarios. And then DevOps center is our command front. So when you think about the traditional way that we would do change sets, this is our modern low code change and release management tool for everyone. It's centrally managed and you can test and promote changes from one low code pipeline to to the next. And this allows you to increase your faster deployments by 31% and then also a 28% of increased productivity that we're seeing from customers that navigate from that change process to more of an automated way to deploy agents. Now as we move to observability, this is one of our tools within shield. So event monitoring allows you to increase visibility and security across users and agent activities. You can audit how users work with agents to modify or access data. You can also block agents from unintended access to sensitive data. And this is all within our transactional security policies. You can set alerts, you can also block that intended and you can see an example at the bottom there where we're showing a user giving a prompt to an agent to give me the details about a specific account. And then the agent responds with I couldn't retrieve those details. This might be due to Access restrictions. Please contact your Salesforce Admin for assistance. And then in an extension of that, you have field audit trail. This allows you to view, track and report changes to sensitive data so you can audit who, what and when data changes for accuracy. You can also search, record and report history changes over time and to navigate those compliance requirements that make sense for your business. And without further ado, I'm going to pass it over to Diego where he's going to now take that foundation that I talked about and talk about the venture crowd story and how they're building, testing and deploying agents and putting that theory into practice. So without further ado, I pass it over to Diego.

Thank you. Pleasure being here today. I'm Diego Mogollon and I lead technology and marketing strategy at VentureCrowd, which is Australia's most ambitious investment platform. AdventureCrowd. We looked at how private capital markets traditionally worked. It was slow, it was manual and it was restrictive. So we decided to change that. We built a digital platform that lowers the barriers for issuers and investors across alternative assets like venture capital, managed funds and private credit. Since inception, We've enabled over $440 million in investments via that platform, with some of our deals delivering returns of up to 295% returns. Now, this is the key part that's all done digitally and that's all done on the Salesforce ecosystem. Across platform, marketing, cloud, sales, cloud, service, cloud, experience, cloud, Tableau, slack, and now agentforce. Or maybe I should say agentforce360. Having everything in one place gives us a single view of our users and optimal alignment between the users, the technology and the business. That puts us at a very unique position to leverage AI. So to power that mission to disrupt private capital markets, we build our own agentic AI framework. It's a unified agentic AI framework and a data layer that's powered by years of venture crowd data and expertise. And it orchestrates intelligence and actions across the whole business. For issuers, it makes capital raising much faster and easier. And for our teams, it delivers insights that improve decisions and unlock more capital. For investors, it delivers the right opportunities at the right moment. So when we talk about AI today, we're not talking about pilots or experiments. We're talking about real impact on our mission to disrupt private capital markets. Let's start with ideation. The reality of building enterprise grade AI solutions today is that success depends less on clever prompting and a lot more on disciplined software engineering and data governance. That's why having a mature ALM framework has been very important for us and it has never been more important to companies. It gives you the ability to move fast, stay compliant and deploy safely. So let's talk about ideation. With AI, everyone is experimenting faster than ever before. And innovators need a safe space to experiment. That's where Salesforce sandboxes come into play. Sandboxes have enabled our developers to prototype fast and safely, while non developers can join the process of experimentation without breaking production. And this is not just I tested it on my local sort of experimentation, this is enterprise grade level experimentation. Salesforce sandboxes reduce errors in complex orgs like ours, improve speed of validation and give our teams confidence to move fast while safely breaking things. They are the foundation for responsible speed and innovation before reaching into the real world. Now let's talk about building, which is my favorite stage. Ideation is great, but nowadays it means nothing unless you're shipping fast. And that's what I always tell my team, always be shipping. At Venture Crowd, we use Agent 4 Studio to build agents really fast. It brings everyone together to build and to ship safely with clear guardrails. As a financial services company raising millions on our platform, AgentForce Builder alone has saved us months of agentic deployments because we didn't have to build the whole trust layer and infrastructure from scratch. I personally build agentic AI frameworks myself end to end. And as much as I love doing it, it's not just a few clicks, especially if you want to do it enterprise grade or like they say here, mission critical sort of Agent Aki. And so what we can achieve with Agent 4 Studio is that we can optimize the rapid iteration process and we can leverage out of the box features. If we need deeper functionality, we can always access tools and actions like APIs, LWCs, flows and apex to get more complex jobs done. And that's actually a great thing for us because our whole tech stack is a mix of all of those. So AgentForce gives us a very quick way to connect agentic reasoning with a diverse range of code and no code tools and actions which are critical for agentic success. I'm particularly super keen to dive deeper into AgentForce Vive coding because I think that's gonna save a lot of time for our developers and they're gonna be spending much less time on toil and tech debt and a lot more time on product development and disruptive innovation. Now testing, testing is where experiments become reliable systems. With AgentForce Testing center we can test agents at scale instead of relying on manual workarounds, which I've had to do Many times it's very time consuming and you rather focus on product development. So with this testing center you can generate hundreds of test scenarios and with just a few clicks you can run detailed results that show exactly where things failed. And most importantly, you can look in the reasoning under the hood. This is one of my personal features in this whole process. It creates a continuous feedback loop and it helps you validate agents before and after they reach deployment. And it involves both subject matter experts and technical experts as well. So it increases the trust in experimentation and it reduces the path to market dramatically. Now there comes deployment, the moment of truth. Without this, nothing else matters. Do you catch my Metallica reference there? Hope you're going there tonight. Anyways, back to deployments. I think companies should look at agentic AI implementation as a new type of software engineering problem to solve. To me, building enterprise grade agentic AI solutions is 20% AI and 80% software engineering, data engineering, people change management process. And I think we saw that at today's Agent Force keynote which is now mixing the LLMs non determinism with deterministic logic like AgentForce scripts. And that's the future. So I recommend looking at agentic deployments with the same rigor. We've always looked at software deployments in traditional software engineering and here Salesforce DevOps give us that opportunity to connect secure sandboxes with live environments and to move metadata, code and changes along that pipeline. And if you think about it, agents are just a mix of metadata and code. And when you give them access to those tools and those actions that I mentioned before, then you're able to review the whole evolution of your agents, track them, review them, version control them from ideation to deployment in DevOps Center. Last but not least, observability. If there's one absolute truth I've learned in this journey is the following. At one point, agents will fail. They have to fail. Observability determines whether you catch it or you pay for it, whether you catch it before it reaches your users, before it reaches deployment. As a financial services company, this is a critical aspect for us. So we found Salesforce event monitoring and field audit trail very, very helpful in this space. It helps us trace every interaction, understand performance and fix issues fast. At Venture Crowd, that visibility allows our team to spot errors early and turn them into learning opportunities. It also helps to keep our compliance team very happy. And every agent and process that goes live at Venture Crowd has passed through this rigorous framework balancing speed, quality and security. At Venture Crowd. That's how we are scaling Agentic AI not by chance, but through system and process design. And all of this serves a bigger purpose for us, which is making private capital markets more efficient, more transparent and open to everyone. And every agent that we've successfully deployed via this process helps us bring closer the reality of a world where capital, ideas and innovation are moving forward together. That's why ALM matters for us, because it helps us turn those ideas into a reality.

Perfect. Thank you so much Diego for sharing that amazing story. We covered the foundation of what ALM Frameworks is, but Diego just shared with you and how he took that vision and applied that into their business today. So thank you for sharing that. And if you all are spinning your wheels now like I am and you're interested in learning more at Dreamforce, we actually have some additional sessions just around our platform and building on the Salesforce platform today, so take advantage of those sessions as we continue forth with our Dreamforce journey. In addition to that, we have an amazing keynote around building and governance around the Agentic enterprise. Some amazing speakers that is going to be taking place on Thursday, October 16th at 9am PT on Moscone West Level 3. You could also stream it on Salesforce plus or scan that QR code and add it to your agenda today. In addition to the free coffee, once you complete your survey today, we also wanted to leave you with something more practical as well where you can continue your learning and apply those same frameworks that Diego talked about into the stages of agents and application lifecycle management. So scan that QR code there and access that documentation. And with that, thank you all for joining our session today. Again, my name is Trayvon Jackson, Lead Technical Architect and joining me is Diego, Chief Marketing and Technology Officer at Venture Crowd. Thank you, thank you.