Engineering Excellence and AI Productivity

Julián
Hello hello! And welcome to Code[ish]. My name is Julián Duque, Principal Developer Advocate for Heroku and your host for the Code[ish] podcast. And today I have the honor to have with me as guest, Shiva Nimmagadda. Shiva is the VP of Excellence Through AI and…

Shiva
Analytics.

Julián
Analytics. Oh, what a mouthful, but amazing, that that title sounds, sounds impressive. Tell me a little bit more about it.

Shiva
Yeah. So, my team actually is responsible to bring engineering excellence with the AI tooling and analytics using and also being customer zero for all Salesforce product, because actually that’s where engineer’s excellence actually starts. And actually, we can bring a lot of excellence for every single individual in the company.

Julián
Beautiful. And how big is your team?

Shiva
I have around 120 members.

Julián
120 okay. That’s, that’s a considerable size team. Amazing. So, how is your team using AI today? I mean, you’re working about like, excellence through AI, but how are you living through that, that standards?

Shiva
So when, when, when we think about when we think about engineering excellence, we always start with the software development lifecycle. Software development lifecycle of an engineer is start from starting to plan. Plan the work items that we call the inner loop, where you start a work item. You codify it, and then you develop it, you test it, and you merge the code to a code base. That’s what we call as inner loop. The development lifecycle expands from inner loop to outer loop to scale loop. Basically, you take the piece of code, deploy it into multiple test environments, test the code, and all the way to deploy to various systems that we have in the in the product. That’s how we look at the software development lifecycle. When you really look into bringing efficiency across every single aspect of the three different loops, we actually go and look at opportunities where AI can actually accelerate and innovate in those areas. As part of those cycles, we primarily think about engineers, the majority part of them are coding. They write code to bring all various industry standard, industry defined, coding opp… coding tools, which actually fits various different use cases. Some developers write UI, some developers write migration scenarios, some developers write tech debt, some developers write new features, various set of industry standard tools. And also, we have in Salesforce, we have internal Salesforce AI coding tools which actually accelerate and innovate and bring efficiency to every single engineer based on their use cases, based on their personas.

Julián
Yeah, and those… some of those tools that we use internally now, like the public can use it like for example, Agentforce Vibes, you can like build Salesforce applications, use like what it’s called, like vibe coding, pretty much like with prompt engineering, with using a coding assistant, just like, specifying what you want and now it’s building everything. So, how has been that whole change like before AI and after AI for your team? What what have you seen that have changed the most for, for the team?

Shiva
So I think when we when we look at the before AI, team has to really write flows. They need to write a lot of Apex classes, all of them to build packages, build applications, building websites on Salesforce, all through coding manually. With AI, with especially with Agentforce Vibes and Heroku Vibes, you are actually… it creates automatically… you define, you tell the flow that you need… you tell the workflow, it actually automates and creates all the things behind the scenes for you. And creating a website, deploying a website on Salesforce, using Agentforce Vibes, is just a breeze. And not just that. We really… if we, if you think about it, we start from, always from analytics to agents, to now, actually, we are going into the world of actions. With actions, you vibe code… vibe code that with Heroku and deploy it on Heroku very, very quickly. You see the value, you show the value of the product, the idea that you are thinking about, and you see that live in Heroku, and you can drive all the actions. It’s just, just mind-blowing. And that actually brings huge developer productivity. Huge engineering excellence to every engineer who is working on bringing their ideas to life very quickly.

Julián
Yeah, this is an ever-evolving ecosystem and I’m always impressed about all the new model releases, the coding tools, the evals tools, everything that is out there. And Salesforce and Heroku have been like working on like… working internally to being able to bring these tools for their engineering and other teams to make their their jobs better, and with higher quality. But also exposing those premium to some tools, so like the… our users can also get the benefits of those. So tell me a little bit about the the tools or the engineering agents that your team has built and maintained.

Shiva
Yeah. So, we actually developed an engineering agent on Slack, because we are Slack-first. And this engineering agent is a multiplexer talking to various set of Agentforce agents, which is actually solving a particular use case. For example, I have an agent. I have an Agentforce agent, which is too… which has all the context about my Agentforce technical specs. I have an agent which actually knows all the things about all the various sets of orgs that we actually internally test. So, which is… the agent is grounded on all the context about that particular use case. Now, we built engineering agent, depending on the context of a Slack channel, it can route the question to that particular agent. That’s where we empower every engineer. Simply talk to engineering agent on Slack, which actually behind the scenes goes and talks to various Agentforce agents, and simply gets answers based on the context that they’re providing, based on the question they are answering. Previously, without all this, the team has to have… talk to another teammate. We are actually really envisioning and also living in digital teammate in the Slack here. That’s the power of bringing Agentforce, bringing Data Cloud, bringing Slack, and bringing Heroku together, which actually, we are living the dream of what digital teammate is of.

Julián
Yeah, we were talking recently about all of this AI hype. It sounds great, everything that you can do, and all the companies and everybody is trying to implement AI solutions. This is, this is like some sort of race right now. I mean, if you are not doing AI, you are you are getting left behind. But one thing you mentioned is that you are also working with the analytics. So, I guess you are measuring the results. Have you seen like that real improvement of implementing these tools? And you are continuing doing like that iteration to get like better productivity for your team?

Shiva
Definitely. I think, as I initially said, our journey starts always with analytics. As a famous saying, you can’t actually improve what you can’t measure. So we actually start measuring every single aspect of development lifecycle. We, we started looking at measurement, various set of disparate, siloed systems, brought all of them together into Salesforce Data Cloud product. We actually, there’s a lot of detail and empowered that through a Tableau as dashboards. We called it as Engineering 360. A 360-degree view around an engineer. Now, when we really look at analytics as a story there, we actually… it opens up eyes to where actually the productivity is actually not good. And this is… we started the journey almost three years ago. And we constantly iterate on it, we continually look at opportunities. We operationalize the metrics, because sometimes metrics is just like a, a nice painting on the wall. It won’t, it won’t really look good when we are not taking actions. So we have operationalized that regularly, and we have almost seen around 35% increase in cycle time of every engineer and team who are, regularly operationalizing and looking at it. So across the company, we saw a 35% increase.

Julián
That’s amazing. And and when onboarding new engineers, what has been the reception like, for example, like, there like certain organizations that they don’t have… or they are like a little bit like hesitant to start implementing these agents or like automating this way or being able to rely on A… on AI for these engineering process. So, what has been the reception with your team or the people joining the team that they get to be onboarded on these new processes because they they are new?

Shiva
Actually, that’s a good point. When we started the the measurement as well, two to three years ago, the teams, like I said, teams always initially receptive about like, what is metrics talks about? Are they going to push for performance? So I think one important thing that we have done in Salesforce engineering is, there’s no one single measure defines you. We actually have various set of metrics. We almost had a developer for 2D from 18 to 19 various metrics and expands to almost around 180, 180 different metrics across the lifecycle. Like security, availability, all of them adds and shows up how productive you are. That actually relieved a lot of teams on hey, is it a band, is it a performance band? No, there is no such thing. We have a collection of metrics which is called SPACE metrics: satisfaction, productivity, activity, and collaboration, and efficiency. If I see I’m just telling five various key bullets of category of metrics. When you add various 15 to 18 metrics, that actually gives you exactly where productivities need to be improved. That relieved a lot of teams on the anxiety. Now teams are focused on where I need to, where I need to really, really focus my energies, really need to be more productive, and every team actually has pivoted. That’s why we see like a huge improvement in work cycle times.

Julián
Nice. So, engineering is pretty well-defined, that are like processes that you can measure. We, we can say it’s a science, right? Is there any other teams that are not part of engineering that are also thinking about implementing something similar, like okay, like using agents to improve also like the type of work they are doing that are like reaching out to your team? How how how that’s happening internally, I’m interested to know.

Shiva
That’s a great question, too. Like a great point. When we think about SDLC, the Software Development Life Cycle, in parallel, we have PDLC also, the Product Development Life Cycle. We are actually not just building agents for engineering, also. Product team, the TPM, the Technical Program Management team, they have various set of use cases to really empower them. For example, in product lifecycle, what is customer‘s… customer’s feedback. The feedback is real… feedback agent is a very key important, because customers are using the product, we really need to bring back the voice of the customer and reaching their plans. So we actually leverage those kind of opportunities. We have vari… various bunch of activities starting from planning and bringing operational product, bringing back into the planning backlog all those things helps the product development lifecycle apart from the engineering development lifecycle of SDLC. So that complements each other together, which actually brings more productivity.

Julián
One thing I love, and you mentioned initially of having this approach being Slack-first, because it makes everything open, transparent, and collaborative. So you can have like multiple people interacting, working in the same type of problems and looking at the… looking at the outcome. So I think that’s a… that’s a pretty good way of, of, of using agentic applications. And this is why Slack is becoming kind of like a… like operating, operating system for, for agents. That’s, that’s amazing. So, thank you. Thank you so much for telling us this story. It’s very inspiring to start like, okay, now I’m thinking how can I make my job easier and better? How can I build my own agents to, like, simplify my, my work? Like, for example, producing the podcasts? I have a lot of things to do that I think the… a couple of agents can help me definitely here. One thing, like looking forward to the future, what do you think you are missing today that you want to see the industry implementing, or a problem being solved that it is already not yet done?

Shiva
Yeah. So I think, we are actually… my envision for the next couple of months, or maybe in the world of AI, it’s just a month because the way things are changing. So, we are slowly moving forward from use-case-driven agents to a lot of autonomous agents, background agents, where all the low entropy work that we are doing, an agent does the background. So we actually really as a human, we focus on the right key, impactful, business use cases. And they… I envision a huge Agentforce agent command center where humans are actually working closely with the use-case-driven agents, personnel-driven agents, and background autonomous agents. That’s where a command center, where are all of them playing together and becoming… I’m actually really, really, be more productive and more… bring more value to my customers very quickly. That’s what I envision.

Julián
I can, I can see that. And and I want to I want to see it happen. Shiva, thank you. Thank you so much for your time, for joining us here at the Code[ish] Podcast, and looking forward to keep talking to you and on a future opportunity maybe you can tell us all of these new things that you are working on.

Shiva
Thank you so much. Thanks for the opportunity.