Tiny Mic: What Agentic AI Can and Can't Do

There are so many people just getting started. There's so few people that deeply understand these systems. More than ever, it's incumbent upon us as business leaders to really think deeply about things that haven't been working, to have the courage to sort of challenge the workflows that need to be disrupted. Whether it's a KPI that needs to change, a team that needs to transform, find the problems that are really, really, really hard in your department and. And then think creatively about how AI can materially affect change in those problems.

Hi, everyone, and thanks for tuning in to this episode of Tiny Mics. My name is Kai Johnson and I lead our executive partnerships here at Salesforce. And today I am joined by Sanjna Parulekar.

I lead our AI marketing team here.

At Salesforce and we are in Napa, California at the Wall Street Journal Tech Live Summit. So, Sanjna, thank you for joining us today. So good to have you here.

Yeah, thanks for having.

Of course. Yeah. So, you know, we're on day two of the event. You know, I know you've been meeting with a lot of great customers while you've been on site. There's been a lot of great mainstage speakers, a lot of great programming during the course of the couple of days. What are some of, like, the key takeaways that have really made you think? Like, these are some good aha moments that you can bring back.

Yeah, I think the biggest aha is actually a very relatable one, which is that everyone is still confused about what agentic AI can and can't do.

Yeah.

And I think they're really eager to just learn from other people, be a little bit vulnerable in sharing what they don't know. And they're just looking to see what their peers are doing and to learn from them, which I think is actually pretty inspiring because we have a very high caliber audience here. You have C levels that are just trying to learn and shape their businesses in new ways.

Awesome. We just wrapped up your session, which was a masterclass on how do you build an AI agent and how do you get value from that. You joined a panel with the CIO of Indeed. And also the CTO from Open Table. And I think that there's still some people that are a little bit like, what's the difference between a chatbot versus an AI agent? How can you get more value? Tell us, what were some of the key messaging and takeaways that you got?

Yeah, I mean, we started out just by defining what is an agent, because I think that, like, with any new and cool technology, people think that they can use it for everything and that they should. Agentic AI. It's specifically meant for a job to be done. So think about like all the jobs and things that you can't complete in a day, which I'm sure there's a lot of. If you only had an AI agent that could be kind of an extension of you or an extension of your team, what would that agent do? What would their role be? What would the data that they need to do that role look like? Where would that data live? What do you want the agent to be able to do and act on? And so we really talked about the definition of an agent and some of the key use cases that we're seeing deliver a ton of value Both at indeed OpenTable and some other examples of Agent Force customers. So it was really, I think eye opening for folks to see, like, maybe I don't have any Gentek AI use case, maybe I should be thinking about it differently. You know what the folks from OpenTable and Indeed shared so clearly is that they're embracing a culture of experimentation because you don't necessarily know everything of what's coming with agentic AI, but you can't just wait for the technology to mature. You have to experiment. You have to go fast.

Absolutely. Also during the session we heard a lot of great examples customer use cases from OpenTable and indeed, I'm actually curious to, because I think a lot of people learn a lot from how are folks like actually using the product and how are they using the tool. So I'm curious, how is your team using Agent Force?

Yeah, in a variety of ways. I think there's kind of the employee facing agents that we all have access to at Salesforce. And then there's also how we use and improve the agents that are available externally. So internally, just on our phones we all have access to a ton of Agent Force agents right there in Slack. The one that I really love is our wellness reimbursement agent. I recently asked if I could get reimbursed for a new Apple watch. The answer is no, in case you're wondering. But it's great for, you know, wellness and benefits. It's kind of hard to navigate and so it's really great to have an agent that can be there to answer your questions. It's also something that's regulated in terms of what you're beholden to based on different plans and what's available at Salesforce. And so I love using that agent. It's something that I use all the time in terms of my team. Our main uses of AgentForce have actually changed the way the team operates in a sense. So we have an agent on our website on salesforce.com that agent is fed, so to speak, by marketing content. And so my website manager is constantly looking at the agent performance. When we launch new products, he will test the agent to make sure it can handle different types of questions. We look at all the different types of questions that our customers are asking of the agent. So it's kind of turned some of my marketers and my website folks into agent managers, which is a pretty interesting thing that's just sort of happened organically. I mean, agent Forest kind of silently behind the scenes has changed the way that marketing operates at Salesforce.

Yeah. And I know that the CIO from indeed was, you know, he mentioned a lot about how he encourages his team to test pilot, et cetera. And I know obviously, you know, we're in this era of everybody going out and testing AI and like really leveraging and using the product and the tool for you and your team. When you get to a certain point of, oh my gosh, this is working, or should we greenlight this project? Maybe we should halt or stall this project. Like, how do you know? At what point do you keep going with a certain project or do you exit?

Yeah, I think it's just really important to continue measuring success and checking your own biases on what you think might be happening. Because it's easy to say, we're using an AI tool for this thing. It's going to be great. You get the budget for it, you roll it out. It's harder to say, oh no, that was a failure. Like, we need to do something different. I think just that courage to call out what's not working, to experiment, to track the metrics that are you think are important at the outset and then see how those metrics evolve over time. Any other project we would do or like any employee that you would measure right, you wouldn't hire a new person, then never ask them how they're doing or evaluate their performance. You would never do that. It's the same thing with agents. So we just have to kind of stay honest with ourselves about these implementations.

So, you know, we obviously know that an AI agent is going to help with a lot of those repetitive tasks and then what's really going to be left is opportunity for employees to really be focused on like that deep focus work and actually do the strategic parts of their jobs. So talk to us about what will the future state of your workday look like?

Yeah, I think it's important first and foremost to acknowledge the fear that people have of agent tech AI. Like any other technology, they think it's going to take away their jobs. They think it's going to totally disrupt everything. And the disruption is real. But I actually think it could be a good thing if we let it be. Because quite frankly, I mean, post pandemic we totally lost our minds when it came to the workday. Right. We felt that it was normal to wake up as early as possible and just start working from our homes. And who's to say when the workday ends and when it bleeds into your evening? And so my hope for agent tech AI is that it brings us back to equilibrium in terms of our work lives, that we can actually, you know, start to reject the notion that it's okay to be in meetings from 7am to 7pm it's not okay. We can have jobs with, you know, a proper work life balance with time to actually think and do focused work versus swivel chairing across people and having meetings about meetings about meetings. And we can really reach our full potential if we let AI do all of that other kind of task, menial heavy lifting.

Love it. You recently said in an interview you can either really lean into AI or be threatened to be left behind. Tell us or give us some advice on how can you advise folks on how to lean into AI?

I mean, you can't get away from these changes in technology, but the thing you can control is learning about things, right? Learning about how these technologies work, thinking about the use cases that are going to be really, really impactful to your business. You know, a lot of what came up on the panel today actually was, well, what if you have a mandate coming top down that you have to use AI? That's the wrong question. It's not. We should have a technology in search of a problem. Find the problems that are really, really, really hard in your department and then think creatively about how AI can materially affect change in those problems. And sometimes it can be small. I mean, on my team we spend a lot of time filling out RFPs in some way, shape or form, whether it's for a customer, for an analyst report. And we have amassed an incredible amount of material on our own products. Agentic AI can help us fill out future RFPs. It sounds like the most boring use case in the world, but it's going to bring time back, it's going to help us. My people will be happier if it's going to bring back two hours in their day that they can spend with their kids or take a load off and go for a walk or a hike. That is worth it because happy employees are going to just be a happier workforce and better ROI at the end of it.

Absolutely. And actually that just goes back to, obviously, us all getting back to hopefully, what, just a regular 9 to 5 as opposed to pandemic living. So. Very good. All right, so tell us about, like, some of the things that you've heard from customers that you've just been interacting with over the past couple of days.

So I had a really interesting conversation with the CEO of Black Girls Code, who's here for the event as well. The conversation was actually not about technology. You would think that, you know, Black Girls Code, they're thinking about, like, coding skills in specific. And it was actually about all of the surrounding skills that we need to bring to folks and how we need to teach people that coding things. Vibe coding. Like, you can do things faster than ever before, but, you know, what we can't replace is the strategy of, what are you going to use that for? Right. You might be able to vibe code an application using AgentIC AI in maybe 30 minutes, but if no one uses that application, did you really do anything that great? And so what she and I were talking about were the surrounding, the softer skills that you need to bring and the kind of ideas around creativity that you can bring to these technologies to make them more useful and to just make the kind of burgeoning workforce that is kind of growing up right now to make them feel that they can really create an impact in the workforce with what they're learning about AI.

Before we move on, I want to throw a slide up here. Let's pause really quickly on Black Girls Code. I think this organization is so cool because we are always thinking about how to upsell and reskill our employees. But this organization is preparing the next generation of the workforce. They're ensuring that we have a pipeline of technologists, engineers, and ethical AI leaders. This is more than bridging the gap. It's investing in the future of innovation, starting with elementary school kids. And they're making a huge impact. They've hosted nearly 7,000 learners through their workshops in 2025 alone. They've democratized quality tech education with their free YouTube. Code Along Series 2. Code along is geared toward introducing young learners to coding. With this series, they've grown to more than 161,000 YouTube subscribers. They average 30,000 viewers per month. They were granted a YouTube Silver Creator Award and the host of the series, Kalani Jewels just won the prestigious Young Artist Academy Award. I encourage you to check the them out, head to their website to find out more. So tell us, what is a surprising way that you are actually using AI in your personal life?

Okay, this is going to sound weird and it's not agentic, but I will send myself voice notes in Slack. I will send myself a voice note. The great thing about voice notes in Slack is that they're fully transcribed automatically. If I'm walking to work, I'm walking to the gym, and I'm like, oh my God. Like, I just have all these ideas or all of these things I have to do. I will just start a Slack conversation with myself. I will record a voice note, it's fully transcribed, and then I will kind of turn it into action items using the Slack bot. So I'm like fully doing that in Slack. It is really, really, really helpful for me. It's just like a little hack. Actually, our head of product for Agent Force taught it to me, so I can't take it that much credit.

Yeah, I mean, hello, productivity, right?

Yeah, it's amazing. It's amazing.

Yes. That is an incredible way to multitask. Okay, so what's been keeping you up at night?

Oh, my God. What keeps me up at night is what's the next new thing that's going to change and totally make all of this irrelevant? I mean, agentic AI kind of came out of nowhere and changed every business and. And that's just the nature of technology, isn't it? So it's not that it keeps me up at night, but it definitely stresses me out when I think too hard on it. So I have to just look at what's. What's ahead of us, the impact we can make.

Oh, my gosh. Well, we definitely don't want you stressed, so I know we've got a big job here, but. Yes. Okay, so give us your famous last words of advice. How are you advising folks that are on their AI journey? Like, what are you. What are you telling folks?

Have some advice. I don't know if it's famous, but the advice that I would give is that it feels like the world has moved on without you. If you don't know everything about agentic AI, it's just not true. There are so many people just getting started. There's so few people that deeply understand these systems. And I think more than ever, it's incumbent upon us as business leaders to really think deeply about things. That haven't been working to have the courage to sort of challenge the workflows that need need to be disrupted and to use that business language, whether it's a KPI that needs to change or a team that needs to transform. And to think about agentic AI as an accelerator to get to your business goals versus something that we should be afraid of.

I've got one more thing before we close out. I want all of you to go and check out this new study of over 150 CEOs from the US and Canada. We commissioned this from IDC and it's called the Voice of the CEO on Digital Labor. It's no surprise that most top execs say digital labor will have a greater impact on their business than the Internet and cloud computing. But what is surprising is that CEOs are building new roles in the workforce rather than simply eliminating workers. CEOs expect nearly 80% of employees to be augmented by AI agents and work alongside them, either by having their experience, existing roles supported, or by being reassigned to new teams. I strongly encourage you to read the full report to find out more about where we're heading. Straight from the mouth of the CEO. Use the QR code to get the full report. Thanks so much for joining us today.