Agentic AI Is Here…Are You Ready for It to Work Smarter Than You?

In this episode of Executive Connect, Melissa Aarskaug sits down with Mayuri Jain, marketing strategist and digital innovation leader, to explore how agentic AI is changing how work gets done.

Unlike traditional AI that waits for instructions, agentic AI sets goals, plans, and takes action, all on its own. Mayuri breaks down how companies can harness these smart systems to handle complex workflows, boost productivity, and still keep humans in the loop where it matters most.

You’ll hear real-life examples, ethical concerns, and how leaders can prepare their teams to build trust with AI, not fear it.

If you’re wondering how to use AI without losing control, this episode will give you the roadmap.

Chapters:

01:30 A driverless car vs Google Maps analogy

03:01 Features that make AI act on its own

04:36 Multi-agent workflows explained

05:21 Ethics and putting guardrails in place

06:30 Setting up policies and traceability

07:37 Where AI is being used today

08:20 Building vs. buying AI tools

12:46 Who’s accountable when AI messes up?

15:01 How Mayuri uses different models to write a book

17:58 How teams can start building with AI

20:56 Final thoughts: Stay curious, not afraid

22:39 Mayuri’s journey from India to New York

29:00 Her new project, Voices

30:45 Where to connect with Mayuri

Mayuri

(0:00) Traditional AI is reactive, it waits for you to prompt, whereas agentic AI takes initiative, it is more proactive, it has goals, plans steps and executes them. (0:13) Think of it like, you know, (0:14) using a driverless car, you just sit on it and enter the destination and it will take care of (0:19) the traffic, navigation and everything and drop you at the right place, whereas if you compare it (0:25) with, you know, any Google Maps or something, wherein, you know, you enter the destination, (0:30) it will help you and guide you with navigation, you have to do all the stuff on your own.

Melissa

(0:34) AI used to be like an intern who waits around until you tell them exactly what to do, but now it’s more like an overachiever who schedules your meetings, books your flights and redesigns your website all before you’ve had your first cup of coffee. (0:54) Welcome to the world of agentic AI. (0:57) I’m your host and today I’m joined by Mayuri Jain, a marketing strategist, a digital innovation leader to explore how this new generation of AI is shaping not just the machines, but how they decide to do the work.

(1:16) Buckle up, it’s going to be a great podcast. (1:19) Welcome, Mayuri.

Mayuri

(1:22) Thank you, Melissa. (1:23) Thanks for having me here. (1:25) I’m super excited to talk about this topic.

Melissa

(1:29) Absolutely. (1:30) Now, we’ve all heard about chatbox and recommended engines, but agentic is a different flavor altogether. (1:38) Can you break it down for us and tell us a little bit about your background?

Mayuri

(1:44) Definitely. (1:45) I mean, traditional AI is reactive. (1:48) It waits for you to prompt, whereas agentic AI takes initiative.

(1:53) It is more proactive. (1:55) It has goals, plans, steps and executes them. (1:59) Sometimes you don’t even have to wait to tell it what to do next.

(2:04) It just does everything on its own. (2:07) Think of it like using a driverless car. (2:11) You just sit on it and enter the destination and it will take care of the traffic, navigation and everything and drop you at the right place.

(2:19) Whereas if you compare it with any Google Maps or something where you enter the destination, it will help you and guide you with navigation. (2:27) You have to do all this stuff on your own. (2:30) Just like that, agentic AI is pretty much very autonomous.

Melissa

(2:37) I love it. (2:38) I think that’s such a totally different way than I’ve been using it until recently. (2:43) Let’s get under the hood.

(2:46) What makes an agent? (2:49) Unpack that a little bit more for me, for those who are new to AI. (2:54) What enables the AI to act in this capacity with this kind of autonomy?

Mayuri

(3:01) Sure. (3:02) It’s about giving the system memory, planning capabilities and tool usage. (3:06) These agents understand context, reason about steps and execute actions, pretty much like AutoGPT and LengChain agents, which are great examples.

(3:17) But if you also add a layer of multi-agent orchestration using LengGraph, then it can also solve very complex workflows on its own. (3:30) So let’s take an example of maybe running a digital campaign where you want to perform a digital campaign and you just enter it and give it to an agentic AI system. (3:44) What it will do is it is going to identify the right personas of the agents.

(3:50) It will divide the task, it will execute the task and it will give you the final output. (3:56) And it will also see where actually the human in the loop is actually needed. (4:02) So the content manager is going to write the fine draft of the content.

(4:08) The designing specialist is going to look at basically the ad creative and it will then go to a digital campaign strategist who is going to work on planning the campaign and executing it. (4:24) So multi-agent orchestration already takes part and pretty much the execution of the task is done on its own.

Melissa

(4:36) I love it. (4:37) That’s such an easy way of doing things. (4:39) It sounds so powerful to me in this capacity, but I do know there’s a lot of concern with using agentic AI.

(4:50) But let’s talk a little bit about when the machines take the lead, where do we draw the line? (4:57) What are some concerns? (4:59) Maybe just unpacking it.

(5:01) Sometimes I think people are concerned, but they don’t really understand what they’re concerned about. (5:06) Maybe because they either don’t understand what it’s doing or tell it correctly what to do. (5:13) So where do we draw the line with leveraging agentic AI and then actually using a human for the work?

Mayuri

(5:21) That’s a big question. (5:23) Autonomy and AI should never mean unchecked control. (5:27) With great power comes in great risk, as we know.

(5:31) Goals can be misaligned and we might get hallucinated responses. (5:35) Tools can be misused and behaviors unintended, right? (5:39) So that’s why putting the right guardrails, alignment techniques, and human in the loop are non-negotiable while implementing some of these workflows.

(5:49) We need to ensure that we are actually taking care of all of that.

Melissa

(5:53) Ready to lead smarter and invest wiser? (5:57) On the Executive Connect podcast, we unpack executive strategies for wealth and influence. (6:03) Hit the subscribe button now.

(6:06) Don’t just watch, act. (6:08) So how would somebody set up policy? (6:14) Is this like a policy thing that needs to be set up?

(6:17) Is it just scheduling time to review it? (6:21) Because I find sometimes when people let it go, when do they check? (6:27) What are the checks and balances somebody should put in place?

Mayuri

(6:30) So it pretty much depends on the domain and it is very industry specific, right? (6:36) The regulations and all of those kinds of things. (6:38) But yeah, I think traceability is very important when you are kind of implementing some of these workflows, wherein you need to ensure that you have proper sources identified and you are putting in proper guardrails and you have a confidence score for many of these outputs which are coming, which gives you more trust and confidence in the implementations.

(7:02) Because then you can make out why I’m getting this output and what is the reasoning behind it and how it is kind of validating that the overall response which is coming has a high kind of a confidence score.

Melissa

(7:19) Yeah, I like that. (7:21) I think I like rules and boundaries myself. (7:25) So I’m curious to get your thoughts on where agentic AI is making the biggest impact right now.

Mayuri

(7:37) I think the first wave is already hitting the places where we see a lot of repetitive work and huge data sets that are involved, like enterprise operations, personal productivity tools and scientific research. (7:52) The trick is balancing autonomy with visibility. (7:56) The system needs to explain what they are doing and why in ways that could build the trust, which is very important.

Melissa

(8:04) So do you find that people are creating their own tools or are they leveraging other products right now? (8:14) Or is it a mix of both leveraging others’ products or creating their own?

Mayuri

(8:20) So it pretty much depends on the requirement, right? (8:23) If I’m working somewhere, then I might use an enterprise AI solution provided to me from my organization, which is basically working well on the enterprise data. (8:34) But if I’m using for my own stuff, I might use ChatGPT and OpenAI kind of a thing and use several models that it has.

(8:45) But nowadays, users with these models growing on at the speed of mushrooms, like people are becoming more curious to learn which model is best. (8:57) They are not talking about AI anymore. (8:59) They are talking about models like, have you used Grok?

(9:02) Have you used NanoBanana? (9:04) Have you used ChatGPT, O3mini, or all different models, Perplexity, SonarPro. (9:15) So they want basically a flavor of, however, see everybody is in a different journey with their learning curve around AI, right?

(9:26) So people who have explored a lot of AI, these LLM models, they are actually wanting a system wherein they have hands-on experience in kind of selecting various LLM models and then perform different tasks. (9:45) And they don’t want to be limited to one specific provider for an LLM model and use only OpenAI-related models or let’s say Anthropic-related models. (9:55) They want flavor of all the models in a simple tool.

(9:58) So that is the I see a lot of people are using enterprise AI solutions that a lot of startups are offering, which gives a flavor of multi-LLM model selection and even orchestration in their various workflows. (10:17) So personally, I would like to use that sort of interface in terms of deploying my personal or professional workflows. (10:27) But then it is also limited to how your enterprise is kind of working because again, data security is very important.

(10:37) The bigger enterprises cannot reveal the data outside to external models without proper guardrails. (10:43) And they have their own enterprise versions, which are pretty much dependent on what cloud and infra they are supported with and where their existing data is kind of, you know, lying.

Melissa

(10:54) No, and I love that. (10:55) I think it’s so true. (10:57) You know, I originally got started with Gemini.

(11:01) That was my tool of choice. (11:03) And since then, I’ve kind of pivoted to several other tools. (11:06) And I found it interesting.

(11:07) I was just in a conference recently and they were talking about how many people use AI now every day, all day long. (11:17) And I thought it was less, but it’s growing so fast. (11:21) I know for me personally, I use it every day for just miscellaneous things, personally, professionally.

(11:28) I love CoViolet for my meetings. (11:31) I love, you know, I can’t remember something. (11:34) And I use chat GPT to remind me of something and I give it some details and it, you know, there pops out what I was trying to remember that I couldn’t.

(11:43) And so, I don’t think there’s a day that goes by that I’m not using AI. (11:48) And I also don’t think that there’s a day that goes by that I’m not talking to somebody that’s telling me AI is going to, you know, take over the world. (11:57) It’s unethical.

(11:58) It’s, you know, I’ve had all kinds of different conversations about AI. (12:03) And so, the real truth I feel for me, and maybe you can unpack it more, is I use it. (12:12) AI doesn’t use me.

(12:13) I tell it what to do. (12:14) I get the output. (12:16) You know, it’s not running rogue.

(12:18) There’s no ethical constraints on, you know, asking it about a city, Central Texas. (12:24) So, but I do hear a lot of concerns about ethics, who’s accountable, plagiarism. (12:32) So, let’s talk a little bit about if an AI agent makes a questionable call, whether it’s wasting resources, doing something unexpected, or running off the rails.

(12:46) Who’s really accountable and who’s on the hook for that?

Mayuri

(12:50) I think accountability in agentic systems is pretty much shared, right? (12:55) From developers, deployers, even users, all play a role. (13:01) And, you know, that is why we need transparency logs, audit trails, and feedback systems.

(13:08) Autonomy doesn’t mean free for all. (13:10) It means carefully delegated control. (13:13) And that is very important to ensure that that is at place, right?

Melissa

(13:18) Are you someone with a demanding job, but tired of the nine to five grind? (13:23) What if you could build a profitable side hustle without sacrificing your full-time career? (13:29) Our new partner, BookProfits, can show you how to turn your limited free time into a thriving online business.

(13:37) Learn how to find, sell, and profit from used books all on your schedule from the comfort of your own home. (13:44) Not crazy tech skills needed and no major upfront investment, just simple strategies that fit into your busy lifestyle. (13:51) Visit the URL on the screen for a free training today and learn more.

(13:58) Yeah. (13:59) And I think, you know, I’m hearing, maybe you could tell me more in this area. (14:03) I’m hearing there’s some rules that are going to start happening.

(14:07) There’s some standards that are going to start being in place. (14:11) I just published my first book and I had somebody ask me about had I leveraged AI in the book? (14:21) And my answer to them was, no, I wrote the material, but I had leveraged pieces inside the book to shorten things, to make it more complex, more to the point, I guess is the right word, versus I’m very long-winded in my replies.

(14:41) And so I love how I’m able to like shorten thoughts with AI tools. (14:48) So I’d love to get your thoughts on some of the best ways just regular people are using AI day-to-day to help with efficiencies in their life.

Mayuri

(15:01) I think I would suggest that people should start exploring as many models as they could. (15:08) Like for an example, I’m also writing a book, right? (15:13) And I’ve been writing it from past few months.

(15:16) I have been using Anthropic Cloud, I think Cloud model from Anthropic. (15:23) It is pretty good at writing, but there are certain frameworks that I’m putting into book, right? (15:29) Which basically for that, basically I might use Grok, right?

(15:34) Because Grok is good at reasoning and creating more comprehensive response in terms of doing a proper reasoning and all. (15:43) So what I would suggest is that whenever you are building something or even generating an output or doing a critical workflow as good as creating a book, right? (15:52) Because you want to make sure the audience like it, it’s easily understandable and you are giving a quality and insightful content to the audience.

(16:01) You want to make sure that you are stitching everything together perfectly. (16:06) So you are not using one generic prompt, but you are using multiple prompts. (16:13) You are using multiple LLM models and using the best model for the best task, right?

(16:21) And then creating your final output because then it is like an art and it is going to create an experience overall, right? (16:29) So you want to make sure that you create the right AI experience that could deliver the output, which is very distinct and unique to you and doesn’t sound like an AI generated response, which is pretty generic. (16:43) So make sure that you understand all the models properly.

(16:48) You make the use of right models at right places for the right work based on what that model does best, you know? (16:58) So that is just what I would suggest.

Melissa

(17:02) Yeah, I love it. (17:04) And I think it’s, I heard this recently, I have some peers that are leveraging AI for meeting scheduling, for reading their inbox and doing other, I haven’t got there yet. (17:15) I haven’t had that, maybe I’m not that sophisticated yet.

(17:20) Most of my uses are, help me plan the best way to get here, the best trip or find me things for my kids. (17:31) So I’m not using it as much as I can for productivity. (17:37) I’m getting there, I’m working there.

(17:39) So let’s talk about for the executives, the leaders, the teams that are listening in today, how can they prepare their staff, their workforce on how to work alongside AI or build agentic systems to really help them today? (17:58) Do you have any suggestions?

Mayuri

(18:00) I think they should start with education. (18:04) We see a lot of content being flooded everywhere on social media around AI and how beneficial it is. (18:12) But I think we should start from the ground level, start with the education, understand how these systems work, what they need and where they can help.

(18:21) And then look at all the critical workflows and even the smaller, bigger, whatever, all different workflows within an enterprise and see how you can increase the efficiency, where are decisions being delayed, where could a smart automation move the needle? (18:38) Look at all of those things and analyze where exactly you need to infuse the AI. (18:47) And not only that, invest in the plumbing, things like data quality, APIs and governance, because an agent is only as good as the system it is connected to.

(19:00) So we want to ensure the right APIs in the third-party data sets and internal data and everything is properly connected to the agent so that they work very efficiently and give you the output that you really need. (19:14) And so it’s a critical business problem. (19:16) Also, don’t wait until it’s everywhere.

(19:20) Small pilots with tight feedback loops are the best way to start. (19:25) So start with the small, fine-tune it and ensure that the business stakeholders are always in the loop while you are building these agentic systems. (19:36) Having their buying is super important because they understand the domain and you want to make sure what the entire team is building.

(19:44) Because in AI, you will see that it’s not about having the technical expertise. (19:49) It is now about having multiple people who are curators, who have a lot of creativity, and who could bring different perspectives, who could understand the processes differently, and then create this entire agentic AI system. (20:06) So it’s a mix of technical plus the business people with functional expertise.

Melissa

(20:13) And I love that you said that because the technical expertise is so needed. (20:20) I know a lot of times when I am doing things a certain way, I’ll have somebody that works with me technically say, why are you doing it that way? (20:28) This is like a hundred times easier.

(20:30) So you’re right, leveraging both sides and kind of like leveraging both sides of our brain, it’s so important. (20:37) And they’ve even helped me make things much easier and saying, what are you doing? (20:41) That’s a hundred times harder than just doing this.

(20:44) And so you’re so spot on there. (20:46) Any final thoughts that you want to leave with our listeners that we haven’t touched on or any nuggets of wisdom about agentic AI?

Mayuri

(20:56) No, I think it’s just a very interesting space. (21:00) It’s evolving. (21:01) Don’t be shy that you don’t know much or wherever you are in your journey while you are in this race.

(21:08) Make sure that you learn things, you experiment. (21:12) It’s not very difficult. (21:13) It’s pretty easy and straightforward.

(21:15) You just have to get used to it and think of it as if it’s a kind of a collaborator. (21:26) It’s not kind of something which is going to replace you, right? (21:29) Because it will only empower users who are using it.

(21:34) So it’s better to learn early and stay, I mean, hungry, stay foolish, stay curious. (21:44) That is what I will, I think.

Melissa

(21:46) This episode’s sponsor is a couple I absolutely love, David and Alexis Keeley. (21:52) They help power couples be successful at home, just like they are at work. (21:57) If you’re crushing it in the business world, but quietly disconnected from your spouse, you’re not alone.

(22:04) Check them out at DavidandAlexisKeeley.org. (22:07) They have spent years creating a simple tool set that will help you save your marriage. (22:12) Don’t forget to use code Executive One to save $500 off their six-week coaching program.

(22:19) DavidandAlexisKeeley.org, code Executive One. (22:26) I love that. (22:27) It’s so true.

(22:28) And I often say there can be but only one Melissa R. (22:31) Skog. (22:32) There’s not an AI agent that’s going to replace me.

(22:34) So I totally agree with that. (22:35) Be curious. (22:36) I want to talk a little bit about your background.

(22:39) Tell me a little bit about you and how you got started in your interest in agentic AI.

Mayuri

(22:48) So, it’s a long story, right? (22:53) Before, I mean, in my past, I started working as an entrepreneur as well, and I had spent a good amount of time building some of the tools around civic infra utilities. (23:09) And where I was using humongous data, and that time, we were using machine learning.

(23:14) And, you know, we were using some of the APIs, like Imaga, I don’t remember the exact names, but we are using some of the APIs from Google to kind of tag the images. (23:24) So we’re using huge data sets to train the system to kind of identify, you know, which image is tagged what. (23:34) And that is where, you know, I had always an interest in AI, but we were using more of predictive AI.

(23:45) But when generative AI came in, you know, it completely changed the way we were thinking earlier using predictive AI. (23:56) And then the entire thing came in, like, you know, everybody was curious, and I myself was so curious to learn about the new things. (24:04) So I was, I had a technical as well as, you know, functional background since starting.

(24:10) So I kind of picked up a lot on it and started learning on my own, just out of curiosity that I want to learn and be with the pace, you know, people are and people are talking too much about generative AI. (24:26) Then I got a good opportunity working with a small startup into AI. (24:30) And I joined that.

(24:32) And there, you know, I really learned a lot from the founder. (24:37) While he was building up his own enterprise AI platform, and he had a couple of tools like multi agent orchestration, and, you know, a lot of other tools. (24:49) So I was closely working on the brand and growth strategy over there.

(24:56) So we kind of rebranded the entire platform and did a lot of interesting, cool stuff there. (25:02) And that is where I thought only learned everything. (25:06) So it was like, refresh mode for me, where and you know, I was learning entirely new things and just being on track with the market trends.

(25:15) And since then, I’m like into this space. (25:19) And I’ve been learning, I’ve built a couple of tools for myself on my own using wipe coding. (25:25) And I’m kind of using those tools.

(25:28) I think pipe coding is another interesting thing in generative AI. (25:31) And you know, it’s sooner or later going to replace a lot of developers because of the way how beautifully it creates an experience overall for the end user, right? (25:42) We can see the UI and we can see the code and you know, we can just as easy if we get the prompt just creates a full blown application, right?

(25:52) So, so yeah, that’s, that’s my journey around it.

Melissa

(25:57) Now, I love it. (25:58) So now you are studying engineering, was it electrical engineering? (26:04) And tell me a little bit about your journey in leaving India and coming to New York.

(26:14) New York is a really different city from India to New York. (26:20) Talk to me a little bit about that journey and maybe some tips for people that are leaving one country coming to another, maybe just lessons learned along the way.

Mayuri

(26:30) Yeah. (26:31) So yes, I did my engineering in electronics and communication. (26:36) And I, then I did my MBA and started working with the biggest, one of the biggest bank in India called ICICI for three and a half years.

(26:45) And I was with Tata group, then moved to Brazil for a year and working with a small capital placement firm. (26:52) Then I came back and saw that, you know, why not develop my own something and, you know, start a business. (27:04) And we saw a lot of opportunities in India, at least not in other countries.

(27:10) So it was limited geo-wise, but it was good. (27:13) And then when I moved to US, basically I moved because my husband was making a move. (27:20) He works with Citibank and he was moving.

(27:23) So I moved along with him. (27:26) I was pretty much excited, but when I came here, I was like, you know, there was, it’s a culture change and, you know, a lot of new things you need to adapt yourself. (27:36) And, you know, at the same time you expect your kids also to do the same.

(27:40) So it was a bit different. (27:43) And in India, I was like having good domestic support, but here it was like, you know, do it on your own, right? (27:49) Everything, right.

(27:50) From household work to, you know, everything. (27:54) So it was a bit of a struggle for me actually to adjust in the new country. (27:59) But when I started working and it became pretty normal for me, then started, you know, joining a few of the communities to help and understand people, you know, more and the culture so that I can get in tune with that and also build some sort of a professional network, which could help me groomed professionally as well as personally, right as well.

(28:29) So it was a bit different, but I’m enjoying the ride. (28:32) I’m still learning things and I’m always curious and I’m always like a learner. (28:40) And I just meet everybody around me with the same thought that, you know, that I would learn something out of this conversation or this relationship over a period of time.

(28:54) So it’s been a fruitful journey so far and I’m excited to see what’s ahead.

Melissa

(29:00) I love it. (29:02) I feel like that’s such good feedback. (29:04) I think giving yourself some time to be curious, to learn, to meet other people personally and professionally is such great advice.

(29:13) And you’re so spot on. (29:15) Not having family support makes everything so much harder and really making sure the foundation is built before we propel ourselves forward is such a good place to really focus on. (29:30) So what’s next for you?

(29:32) Tell us any interesting projects you’re working on. (29:35) I know you’re a lover of animals and is there any new projects that you’re working on that you want to share with our listeners or should we keep that until next time?

Mayuri

(29:48) Yeah, I think I’ll just briefly state that I’m working on creating a social utility called Voices during my free time over the weekends. (30:01) So that is what I’m focusing on. (30:04) It’s been a year since I’ve been working on it and I’ll soon launch it.

(30:11) And it’s going to be very unique and different. (30:15) You will see some flavor of AI behind the actual dashboard which is going to get launched. (30:23) But it is about bringing back the authentic voices and text-based conversations.

Melissa

(30:31) I love it. (30:32) It’s such interesting. (30:34) Now tell our listeners what is the best way to connect with you and maybe any final thoughts you want to share personally or professionally.

Mayuri

(30:45) Listeners can check out my work on Finextra, Medium and follow me on LinkedIn. (30:51) I will be sharing frameworks, case studies and tools to help them stay ahead of the curve. (30:58) And thank you so much for having me.

Melissa

(31:02) Yes, absolutely. (31:04) Thank you so much for sharing your time, your wisdom, your knowledge and your journey with our listeners. (31:11) That’s the Executive Connect podcast.