Episode 112 - Ishant Kulshreshtha

Erik Martinez: [00:00:00] I've been watching some agencies build their AI capability around one person. Let's call them the AI champion. They are the person who figured it out, who built the custom GPTs or agents, who got curious and went deep. And it sounds great until you realize that if that person decides to leave, everything they know walks out the door with them.

Your agency's progress with AI is literally limited to one resignation. And I felt that risk in my own business, so I started asking, "What are the most important things our team knows and is still locked inside people's heads, buried in meeting transcripts or in email threads that nobody's ever looked at?"

Ishant Kulshreshtha from White Label IQ has been seeing the same thing inside the agencies he works with, and he's got a name for it that I love: sitting on a pile of gold. Ishant, welcome to the show.

Ishant Kulshreshtha: Thank you, Erik. It's a pleasure to be here.

Erik Martinez: Ishant, before we get going into our topic, can you give the audience one-minute version of who you are [00:01:00] and what you do?

Ishant Kulshreshtha: Okay. I'm Ishant Kulshreshth. I work as a AI strategic analyst at White Label IQ. I'm a writer. I've written two books, so that's how it all began. And, this AI transition is something that I want, to be, writing on, and this is one wagon I wouldn't want to miss, and here I am.

Erik Martinez: That's awesome. I have written exactly zero books, so you're already winning that game. So, let's dive in. I think when we talk about AI adoption, there's so many different ways to think about AI adoption, but you're talking to agency owners every day and agency leaders. What are you really seeing in terms of AI adoption in agencies, and are there any specific challenges that you're seeing pattern?

Ishant Kulshreshtha: So, yes, we get to talk to a lot of agency owners, and there is one consistent pattern that we see all the time. Just to be clear, there are two kinds of agencies, one that are beginning their AI journey and the ones that they have begun and they have started learning and understanding how [00:02:00] AI works for their agency.

So the ones who have just begun learning, they are focused on, " What tool should I use? Is it Replit? Is it Lovable? Do I use Claude Code or is it Codex?" So, that's the, problem that new AI adopters are talking about, and that's not the real thing that should worry them because it's never the tools. I remember clearly that about two months ago, we had a lot of custom GPTs.

I personally had about twenty plus custom GPTs in my GPT account, and now we have moved to Claude, and everything sits in Claude. Now it's skills. Earlier it was custom GPTs. So it's never about the tool, but always about, how you can progressively, create an infrastructure that works in your favor and not in favor of a particular tool.

Erik Martinez: I think that's true. I have a friend who owns a home remodeling business here in the, Lawrence, Kansas area, and he and I have been geeking out over AI together for the last six months. And one of the things that's really interesting, one day I [00:03:00] asked him, "Oh, are you using ServiceNow or any of those tools for your business?"

He goes, " No. I use Claude. I use ChatGPT." I'm like, "Okay, so tell me a little bit more about that." And he's like, "I really have built my system using Claude and moving files in and out of Google Drive folders, and our whole process is built around that, not software."

If you had had this conversation with somebody three years ago, that was inconceivable, right? Because, you're either completely doing it manually, right? But what he's done is he's engineered a system using existing tools,

Ishant Kulshreshtha: Yeah.

Erik Martinez: but not building and buying software.

He's using the native AI tools to help do that, and I thought that was very, very, very cool.

Ishant Kulshreshtha: Yes.

Erik Martinez: One question I have for you, just kinda talking about these agencies that are focused on tools, some are starting to figure out how to scale, right? Go from, one-off chat conversations to custom [00:04:00] GPTs, then to, Building maybe one-step agents and then maybe some multi-step agents, right?

I have this framework that I work with, right? Multi-step agents and then multi-step, multi-source agents, and then hopefully you get to the holy grail of, fully autonomous agents, right? While some of those do exist, I don't think a lot of us are anywhere near that, right? We're still kind of closer to these single step, one-off type scenarios for most of the businesses.

But one of the questions I'm kind of curious about is, Do the agencies that you're talking to,

Do they try to help their people learn how to go from just chatting to maybe using custom GPTs? And there's plenty of use cases where that's completely appropriate, Like, writing questions for a podcast.

I do that all the time. I don't need anything fancier than a custom GPT because it's that type of process, right? What are you seeing in terms of training within agencies as part of the AI adoption cycle?

Ishant Kulshreshtha: So, [00:05:00] I'll begin with a rather unsettling statement, which would be that AI on its own is slop for agency work. And any AI just on its own sitting idle, a newly bought Claude account or a ChatGPT account is going to be slop for agency work unless you customize it for yourself.

It knows how you work, it understands your processes, it is trained for your audience, and it is basically an AI built for you. Until you achieve that phase, an AI is going to give you the exact same answers that it gives to your friend and to their friend, and basically it makes everybody in the room the same person.

That's one thing that agencies should focus on. What we are seeing in terms of adoption and training is, people are curious. Most agencies are still writing emails, using ChatGPT. And we are talking about a large number. don't have a perfect number to quote here, but a majority portion of the agency owners that we talk to, they feel glad when they build a custom chat-- GPT and a custom GPT is a thing of the past. And It's still great [00:06:00] if you do it nicely. If you understand how to train a custom GPT, then it's great. But yeah, people are stuck there.

Erik Martinez: You know, I think that's one of the challenges that we have is like, learning how to paint, right?

You got this blank canvas in front of you,

Ishant Kulshreshtha: Yeah.

Erik Martinez: and it's likely if you've never been trained to paint, like you know what a brush is, right?

Ishant Kulshreshtha: Yeah.

Erik Martinez: But, that's why there's art schools. They teach you how to use, the tools. Why is that important? Well, in the AI world, I think we're gonna teach you how to ask questions of the AI. We call that prompt engineering or context engineering, but it's really learning how to ask questions of the AI. Do you know how to ask questions of the AI, Ishant?

Ishant Kulshreshtha: Yes, kind of.

Erik Martinez: Yeah, kind of, right? Because as a beginner, what type of questions were you asking the AI?

Ishant Kulshreshtha: As a beginner, it was simple. It was, " Can you rewrite this in my voice? Can you rewrite this email? Can you make it better? Can you fix the mistakes?" Could you help me plan my day?" It began with all those things. Most of the [00:07:00] learning begins internally, and when you start experimenting with it and you start learning and reading the answers that an AI gives you back, you understand its patterns. You see them, and then you start modifying your prompts, and that is how this prompt engineering thing works, and that is how you learn it.

So, yes, it began by small steps, and then it transformed into something that proved that it works, and then you adopted it.

Erik Martinez: But today, if I were a brand-new employee at White Label IQ, and I've never used AI. You would now know, hey, you can start with some simple questions. So start there. Now, take that simple question and do this.

You would learn faster if you had some of that guidance.

That brings me to the concept of the AI champion. And the agencies I've talked to, pretty much all of them have an AI champion. And what I mean by that, that's the person who's the most curious about AI and experiments with it the most in the company,

Ishant Kulshreshtha: Mm-hmm.

Erik Martinez: There might be more than one, [00:08:00] depending on the size of your organization. But they're the people who are like, " AI this, AI that. I tried this with AI, and it didn't work. I tried this with AI, and it did work. Oh, and when I did that, I found out this." They're enthusiastic about it. They're going home at night, and they're still playing with it.

From your perspective, when you talk to agencies that have an AI champion.

What do you think are the benefits and the downsides of having that person on your team?

Ishant Kulshreshtha: So, an agency is like a well-built machinery. It can work really well if it's serviced well, if it's maintained well. If all the cogs and wheels are in places, and the lubrication is in place. So, your AI champion is just one portion of, this machinery. They are either the cog, the wheel, or the lubricant.

And for your agency to work properly you need all the things in place. You need all the components of this machinery running smoothly in cohesion with everything else. So, an AI champion, what he does to your agency is, he empowers the entire agency, and he creates the curiosity within the agency of what AI is.

[00:09:00] So, he is knowingly or unknowingly creating this bias and creating this movement towards learning this new thing. So your team might just be asking questions, and suddenly one day this man shows up and talks about a vibe coding tool that he built that does a lot of these jobs. And that puts motion into your entire organization, and they start experimenting, start building.

Basically, when you see somebody close to you sitting next to you actually build it, you realize that it's humanly possible to actually do it. It might seem daunting from far, from watching it on YouTube or watching it on your Instagram reels. But, when someone next to you built it, you suddenly feel the power to be able to build it yourself.

What would it take? A thirty-minute YouTube video? And then you suddenly are ready to invest that time. So that's how an AI champion boosts it. But what an AI champion does bad. Okay so, if your agency relies completely on one AI champion, the agency's progress is limited to one resignation.

[00:10:00] So, it will be decided or halted with one resignation, and that's not what you want. So it's a group of champions. It's basically a team that works together and works for the agency, not for themselves. The single big motivation for AI champions is they are getting to learn a lot of the different tools.

They are doing a lot of the different tasks. And this variety in what they are building and what they are doing is what keeps it in their head that they are a champion in AI. And by the way, I'd like to bring something up here.

This quote was, "Jack of all trades, master of none." And I thought that being a jack of all trades and not being a master of one is a bad thing. A jack of all trades is not valuable because a master of one is more valuable. Only two years ago, I realized that this is an incomplete quote.

The entire quote is, "A jack of all trades, master of none is still better than master of none."

Erik Martinez: Yeah.

Ishant Kulshreshtha: It changed my entire perspective two years ago, that being a jack of all trades can sometimes work in your favor, [00:11:00] and it is a good thing to be. You know the bandwidth or the spread that your skill has and the spread your agency can, rely on you.

Erik Martinez: Yeah, I think that's right. You know, as I think about AI champions, I think what you described of them helping push an agency or an organization forward is correct. I think the challenge that I've seen with the AI champions is a lot of knowledge gets locked into one person's head, and I think you stated that pretty well. The risk is, one person has the key to all the knowledge of how to do those things, and that sounds great until you realize that, hey, they may be leaving. Or, they need to be gone for two or three weeks. Any of those situations. It's funny because I think AI has given us the ability to be a jack of all trades, and to actually show some skill in areas that you didn't have skill before.

So, before AI, I couldn't build a website. I can build a website now.

Ishant Kulshreshtha: Yeah.

Erik Martinez: I [00:12:00] understood how to build a website. I've been very much in the project management of that throughout my career, but I've never been the one who's coded, right? And never one who's decided all the features, and I can do all of those things.

The question I think that's starting to come up though is, because the technology's accelerating so fast, it's pushing people back into specialties again.

Ishant Kulshreshtha: Yeah.

Erik Martinez: Because you can't learn it all. So, back to your quote, I think you're right. Being a jack of all trades is better than being a master of none. But I would also argue that if you have a specialty or you have a skill for a particular thing or deep knowledge, you could double down on that and do really, really well with today's technology.

Would you agree with that statement or do you have a different perspective?

Ishant Kulshreshtha: I completely agree. But, again, for an agency to work, it's a machinery and with all the cogs and wheels, you cannot just rely on unicorns. So, your unicorn is your specialist, but you need somebody who understands the broader [00:13:00] perspective.

Your agency is not working in one lane. No agency is working in one lane. It's multiple lanes, multiple feats, broader direction, and that's where these jack of all trades come into play. And yes, you need specialists, these unicorns that push you. Unicorns give you the speed and, the jack of all trades give you the breadth of what you could actually do.

So that's how I would frame it.

Erik Martinez: I think that's a great framing.

Let's pivot to the idea of bringing in outside help. And, I'm in that business. You're in that business. We're the outside help that people bring in when they have a need. From your perspective and the people you've talked to, when does it make sense to bring in outside help, and when is it too early?

Ishant Kulshreshtha: You can always look for an outside help, but most problems are better understood from within. And an outside perspective is almost always required. But yeah, it should begin from the inside. For an example, all the agencies that have been operating for more than a year are already sitting on a pile of gold that [00:14:00] they never realize that they have.

They have had countless meetings with clients. They have had countless internal meetings. They have had threads, hundreds and probably thousands of threads of emails that they live on. And they have no idea on what their clients have moved from. So, they don't know their client's positioning on day one and what's their position today.

What do they feel? What are the problems that they face? Has it evolved? Have their needs evolved? They don't know it all. The problem identification, the initiatives always have to begin internally. Yes, an external help can make you realize or, probably give you, insight on what things you should be looking on.

Basically, it has always been on your desk with dust piled on it. An external perspective can just blow that air out, and you would be able to see the problems and the solutions more clearly. That's how an external help would work.

And when does it make sense to include them? Begin internally. Decide a number. When you have failed that number of times, you can [00:15:00] look for an external help. Your agency has been working. I don't know why they fear AI. If they have passed the internet age. If they have understood how to live with SEO, website building, back-end, front-end, and all of these technical things, AI is not difficult. AI is just helping you do your work faster if you understand your work. So, understanding your work comes first, and then an AI would be able to help you do it.

Agencies shouldn't fear it. Start internally, fail X number of times, and then move to an external agency, and most agencies wouldn't need that.

Erik Martinez: Yeah. You know, I had a conversation with a small agency a couple of months ago and they're like, "how can you help?" And I'm like, "You know, I think the very, very first thing you should do is go think about what it is you're trying to accomplish.

What are the things that are, causing you the most pain? Or, the things that you can't get to that you and your team can't get to? And when you figure those things out, then let's have a conversation about that. Because that's [00:16:00] where I might be able to help you the best."

It's really interesting because, I've had this conversation multiple times in the past month or so, where I've been talking with potential clients and they say, "Well, I can just build it myself." Absolutely, you can. If you're willing to invest the time and you're willing to fail a bit, you can absolutely build those things.

And I think the opportunity to bring in some external help is when the opportunity cost of not doing whatever it is when you need to accelerate it is one of the things that you should use as your criteria.

You all can absolutely build these things in-house with the existing tool sets. It comes down to a little bit about time.

Ishant Kulshreshtha: Mm-hmm.

Erik Martinez:You know, let's pivot to the concept of an AI council. And I have had the pleasure of collaborating with a few agencies that have started to implement an AI council.

And basically, the concept of that is hey, we're having a multi-person, multidisciplinary team come [00:17:00] together and figure out what problems we can solve. I had, Ryan Shanafelt on, I believe it was episode 107 from De Novo Marketing, and they have an AI council. And the one thing that they like, " Oh, my God, we so hate doing reporting."

Our clients need it, so it's not going away, but we so hate doing the reporting. So, they decided that was one problem that affects the entire agency, affects client satisfaction, that we're gonna go solve that problem. So, I guess the question is, besides that kind of use case, what other things are you seeing agencies that have AI councils doing? Or, are there not that many that actually have an AI council?

Ishant Kulshreshtha: Again, coming back to the cogs and wheels and the machinery thing. That's how we talk internally, and that's how I like to frame an agency. So, your organization is made up of different people. There is a leadership that knows the direction that you want to go in.

There are people who understand your agency really well, who have worked long [00:18:00] enough to understand the problems and to have a potential idea of where the solution is. They at least know a direction in which the solution lives. And then there are people who will actually use the solutions that you build.

One of the major things that we see, when we talk to agencies is, you can build a thousand beautiful tools, but adoption is something that can only happen, when everybody has been bought in. Everybody has been included. Your leadership cannot build tools for the people who are actually going to use it.

They have to be involved. They have to have their say, because they need to understand the tools well enough. They need to, trust the tool well enough that they begin using it. And that's where an AI council would come into play. It should include a mix of all the people. And what I can say from experience is it always begins with a group of people that you shouldn't be calling an agency council, and it evolves into something that becomes an ideal agency council for your agency.

So, it evolves over time. You learn by your mistakes, and that's how it begins. It needs a lot of people. From the [00:19:00] people who will be using it, to the people who understand the company's direction. To the people who understand AI really well to be able to build a solution, and the leadership who knows the direction.

An AI council, if it works in this format, it should do well.

Erik Martinez: When this group of people comes together,

after every meeting, and let's say the group meets once a month. What should they come out with? What should they produce?

What's the work that they should be doing that helps advance the cause?

Ishant Kulshreshtha: When a council meets, and this is a group of people who understand different portions of a problem statement, they can come up with a hundred solutions. They can build a hundred solutions. The real power and the real effort lies in what needs to be built right now. Creating the difference between what needs to be built right now and what needs to be pushed to later, and what needs to be shelved forever.

So, a thousand different ideas can come, but after a meeting, should come out is, what is the immediate need? And a list of shelved ideas. It's more important to shelve [00:20:00] ideas that pop up, because if you don't, cancel these ideas out, everything gets built and none of it works.

So, a priority list is always important. Because, you cannot move in a thousand different directions. You should be solving one problem at a time with AI, at least. What we have also experienced is building complex solutions all at once, it is very difficult to gain the trust of the user.

They don't trust AI enough that it could do such complex works. So you build it in slices, and then you join them together. When each slice has earned its trust. So that's the AI council's job.

Erik Martinez: Yeah, I think that's perfect. Solve one problem at a time, make it simple. Don't try to solve the entire system all at once. Solve things that actually help your team advance. That's kind of my paraphrase of what I heard you say.

Ishant Kulshreshtha: Yeah.

Erik Martinez: So, if an agency owner is sitting here going, "Well, that's great. I'm still not sure where to start." How do they choose that right first path for them? What Questions should they ask themselves to say, " How do I get [00:21:00] started today?"

Ishant Kulshreshtha: To be able to answer that, any agency owner needs to understand how an AI would work. For any AI to work, it needs great quality input. If the input is scattered, if it's unorganized, if your data looks like rough notes, an AI's output would look like rough notes.

So, where an agency owner needs to begin is start organizing your data. And your data lives everywhere. It lives in the meeting transcripts. It lives in your project management tool. It lives inside of your emails and your internal communication.

It lives in all those places. And if it lives in all these places, it's a good sign that you can organize it. But most agencies have the data living in people's heads. So it's some people who have, everything inside of their head. People walk to them, ask them for SOPs, ask them for procedures, and that is when they speak it out, and that is how it works right now.

If I could give one thing to do right now that any agency owner could do is sit down, just [00:22:00] open the voice-to-text on any LLM, and for every day for the next thirty days, speak for thirty minutes about the direction of your agency, the procedures, the SOPs, your mission, vision. What do you want with AI? Anything and everything.

For the next thirty days, for thirty minutes every day, just sit down and talk about AI. Talk about your agency, and that should give you a habit. Plus an infrastructure to build upon. So that's what I would suggest.

Erik Martinez: I think that's a great suggestion. We have the ability to record something, easily transcribe that into a document, and that document can be the input into an AI, which you can then use to analyze and structure what you just talked about. In a very, very easy format.

So I think that's a really practical first step that people could take today to start working towards the next thing, which is, "Hey, how do we now start to leverage all this knowledge to help our business out?" I think that's fantastic.

Moving along to proving [00:23:00] progress. So, I've done my 30 days of recordings. I've got all this knowledge.

Ishant Kulshreshtha: Mm-hmm.

Erik Martinez: What is the next step? I've put all this stuff together. I've got some knowledge. Now where do I go from there?

Ishant Kulshreshtha: For thirty days, if you have talked to an AI, you have an infrastructure in place. Your AI now understands how you work, how you speak. What does your brand guideline look like? What do you feel about your clients, how does your deliverable look like?

And when all of that is defined, you can leave the course to an AI, and AI would pick it up from there, make sense of all of that combined. And they'll be able to make sense of huge chunks of data and be able to take away most of the work off your plate.

So, that's when you begin. It's still experimenting because you have just collected the data. You have collected it, but not connected all of them. So your data still lives in either Google Documents, Notion, Obsidian, or some other tool of preference. But, now there needs to be connections made between these databases. And yes, it will be a step-by-step process, but you have done half the job of [00:24:00] collecting everything.

Now, when you start using an AI to analyze all of this, anybody can realize how does this work better. These three documents in sync work better. I handed off the content guidelines plus the brand guidelines. Plus I tell it that this is an email that I need to send out. It understands everything, and it is able to build something that used to take hours or at least it used to take multiple teams.

So it's easier now.

Erik Martinez: Yeah, I think that's a very good practical next step. I've now got thirty days of recordings. I've also got history in my email and all these other places, and how do I bring that together into AI? And you might just pick one problem to solve, right?

Ishant Kulshreshtha: Yeah. And, just to continue where, what you said. It's quite simple once you've collected all of this data. I'll repeat what I said earlier, that every agency that has been operating for more than one year is sitting on a pile of gold. We never realize it, but when agency owners or agency salespeople talk to other agencies. They are talking to a wide variety of people who have a wide variety of [00:25:00] problems, and their positioning or their statements have changed and evolved over the past one year.

If we could just give an AI all the meeting transcripts of one year and ask it, " How have my conversations or the questions of my clients, how have they evolved over one year? What is it that they were asking one year ago and what is it that they are asking right now?"

This is enough to give you something that could give your brand strategy a direction. Potentially, it could also, weigh in on your mission and vision statements. So, you are sitting on a pile of gold, and you just have to realize it.

Erik Martinez: I think that's right. There's so much information that we've been able to collect but not use. I think that's a really, really important point because there is a lot of opportunity to leverage all this data that we're collecting now to help make our businesses better, stronger, faster, more responsive to our customers.

Ishant, I want to be respectful of your time. We're [00:26:00] drawing close to a close. I guess the next question is, you've given some really good practical steps for getting started, and I think even agencies that are further along the curve and may be solving some very specific problems probably could listen to this episode and go, "You know, we haven't applied that in account management," or, "We haven't applied that in project management," or, "We haven't applied that to our digital marketing teams in some way, shape, or form."

You could literally do this department by department depending on the size of your agency. Is there any last piece of advice you'd like to leave the listening audience with?

Ishant Kulshreshtha: I don't want to sound motivational or something, but AI is not difficult, and agencies are the last group of people who should be afraid of AI. Agencies, they solve the most complex problems. They are solving a wide variety of problems, and if they have been able to do it over the past few years AI isn't something that should, give them fears.

The only thing, that would change in how agencies operate, and it would change drastically, which [00:27:00] is, you can no longer charge by the hour. pre-AI era, I've read hundreds of blogs on why you should not sell by the hour or you shouldn't, talk about an hourly, rate.

It's always the value prop-proposition that you provide. And now more than ever, even if agencies don't want it, it's coming for us. That the hourly rate is going to go and we need to figure out a different selling model. So, one thing I would, say to agencies to think about.

Start thinking about. Start changing about how they charge.

Erik Martinez: Ishant, thank you so much for your time today. I always have fun talking to you. I always get some new ideas every time I talk to you on what to do. If somebody wants to reach out to you, what's the best way to get in touch?

Ishant Kulshreshtha: Anybody who wants to reach out to me, they can reach out to me on my email address, which is Ishant K, I-S-H-A-N-T K, @whitelabeliq.com. Or you could reach out to White Label IQ directly.

Thank you, Erik, for being such a great, thinking partner. It's a pleasure to think with you.

Erik Martinez: Thank you, Ishant. I really [00:28:00] appreciate it.

Well, folks, you, you definitely should reach out to White Label IQ. They're fantastic. They're good friends and collaborators of mine, and I absolutely enjoy, talking not only with Ishant, but with Brian and other members of the team.

Here's something you can do on the drive home. Ask yourself, if the person on your team who's figured out AI the best didn't show up for two weeks, what would break? Whatever comes to mind first, that's where your most valuable knowledge is trapped in one person's head. And here's the thing, it's not just you who needs to start capturing what you know.

That person, the one you just thought of, they need to do it too. And honestly, so does everybody else on your team. Thirty minutes a day just talking to an LLM about how you do your work, your processes, your decisions, and how you think about problems. And in thirty days, that knowledge isn't locked in anyone's head anymore. It belongs to the organization.

Thanks for listening to this episode of the Digital Velocity podcast. Have a fantastic day.

Narrator: