AI Consulting: How Strategy as a Product Is Transforming AI Adoption

Yeah, absolutely. Utsav thank you so much for having me here. ⁓ You know, we connected before, we talked about the amazing insights that you share in your book and so ⁓ happy then to share a bit more in this setting. ⁓ So my story, I guess it's, you can think of it as a path of an engineer who later on became a management consultant ⁓ who then kind of tries to combine both backgrounds of mine. So I studied electrical engineering, computer science, and right out of college I was working ⁓ as a software engineer at Amazon and then later at Google and I also had a brief stint at... ⁓ connected vehicle startup. So that's kind of where I have my training and that's where I, you know, learn to appreciate the technological opportunities that you have, right, to solve real life problems.

But then back in 2015, I came back to my home country Kazakhstan and I kind of looked around, I was looking for a job and I was looking around and, you know, McKinsey... crossed my path. Back then I didn't even know what McKinsey is, shame on me. ⁓ But I still got very interested the company. I read a book, The by Duff McDonald ⁓ that ⁓ a of interesting insights about... ⁓ how much of an outsized influence a firm like McKinsey ⁓ has had on the world, on governments, on large companies, etc. I became very much interested because ⁓ always was, know, ⁓ ⁓ inclined if you wish, but at the same time I was interested to start shaping what this technology does in a way that's kind of higher level than just an individual contributor.

And so that's when I joined McKinsey as a business analyst. I was technically a generalist consultant back then and in the very first project that we had, know, we basically needed to figure out ⁓ how citizens of a particular city are reacting to some of the ⁓ initiatives that were driven by the local municipality and our manager basically was saying, okay guys, you're gonna have, ⁓ you know, a bunch of surveys, we're gonna send consultants on the street and we're gonna... ⁓ ask them to collect that data from the citizens, just stopping people and talking to them. And I said, well, hold on a second. lot of this information is already contained ⁓ in the ⁓ ⁓ online discussion forums ⁓ Facebook and a bunch of other social channels. Why don't we just ⁓ scrape of this data and apply some ⁓ aspect and sentiment analysis type algorithms? ⁓ basically we'll have our answer. ⁓ back then the pushback was Alibek know, like don't do your Google thing here.

It's not what we're doing as management consultants. ⁓ I kind of, back then, you know, to me it sounded a little bit ⁓ like as if I came from one world into a completely different world. ⁓ But it also was a very interesting moment because the year was 2015. And I believe exactly that year actually, McKinsey acquired Quantum Black, which is now known as AI by McKinsey has grown tremendously. It was like, I don't know, thousands of people now that are part of Quantum Black within McKinsey. And I think that's what ⁓ truly jumpstarted ⁓ transformation within the firm ⁓ into the world of data analytics and digital implementation.

And even to some extent, you know AI and machine learning. And I was lucky to be at the firm throughout that part ⁓ serving either tech customers on mostly non-technical questions or non-technical customers clients on technical questions such as cloud migration, implementation of custom AI tools, and a bunch of other topics.

Yeah, absolutely. ⁓ In one sentence we say, are an AI-first consulting operating system or professional services operating system. And it's hard to grasp exactly what it means. But ⁓ to maybe connect the dots, we started with a very simple question, which is, can a consulting company ⁓ boutique consulting company for that serve you know 4x, 5x as many clients ⁓ ⁓ their headcount by same amount. ⁓ As we know in professional services ⁓ the top-line the revenue, right, is very much tied to the ⁓ headcount headcount course drives the costs. ⁓ And so we asked ourselves, you know, what can we do make consulting business truly scalable? ⁓ And our very first principles answer to that is, yes, you can do it if you can significantly streamline, which means. Either automate or at least significantly augment the most tedious consulting work streams workflows, right?

Like I myself, I found myself way too many times aligning slides on a presentation at like 1 a.m. or something like that. Something that no educated, highly skilled person should ever do, in my opinion. Very menial task. And so when we started working on perceptives, we said, OK, what are the most mind numbing, soul sucking consulting workflows that we can optimize And of course we were starting, we actually incorporated in April of 2024.

By then the Generative AI wave was in the full swing. ⁓ And to us, of course, it was the biggest enabler. If you ask like why now, you know, that's the answer because for a long time... Even while at McKinsey, was working on reports that basically would say, are ⁓ industries and how ⁓ they are, how tech forward they are, etc ⁓ And lot of them were like, know, like construction and, you know, banking, etc. And a lot of the white collar work, like it was hard to automate it.

And then now we do have this amazing tool and ⁓ that's what Perceptis And I can go into the ⁓ specific use cases, you let me know. ⁓

I ⁓ mean, anybody is using Chat GPT ⁓ know how ⁓ different the is how ⁓ much faster you accomplish something with these ⁓ smart tools. ⁓ biggest problem in my view ⁓ is that the ⁓ gen tools general gen AI tools ⁓ not a perfect fit for the world of consulting. And I think the most unfortunately picturesque in a negative way. Example of that was what unfolded in Australia in October of this year, 2025, when we've heard that Australian government was receiving some kind of report from Deloitte, if I'm not mistaken, and that report contained a ton of hallucinations. It was referring to regulations that don't exist.

It was referring to sources that don't exist, etc. And then ultimately, Deloitte, to reimburse close to 400,000 Australian dollars back to the government. ⁓ And I think that's exactly the perfect example of where the generic tools fall short. ⁓ so ⁓ you think about ⁓ on the other hand... ⁓ ⁓ the kind of, you know, five big ⁓ value propositions that we see is that we are giving people an AI that they can finally trust, you know, wage war on hallucinations and, you know, that's a very important thing. ⁓ The second thing is that ⁓ you don't really want to sit in front of Chat and wrestle with the prompt you get that ⁓ email or perfect executive summary or something like that. ⁓ actually want to say, okay, this is the piece of work I need to be done, I need a proposal to be written, I need a white paper to be written, I need to create a slide that's sitting in my head and I want to create it in the most efficient way and it needs to be in PowerPoint, it needs to be editable and ⁓ that stuff, right? ⁓ So ⁓ that at Perceptis we automate workflows, not just kind of give ⁓ a chat, but we'd rather say you need a document, there you go, you need a slide, there you go. ⁓ And then there are a couple of other considerations, as ⁓ security of paramount importance. When you use a $20 per month ⁓ GPT subscription, ⁓ one doesn't have enterprise ⁓ guarantees, for example, which it might be ⁓ training models on your data. Data retention ⁓ that apply to your data are not, you know... is not that favorable. ⁓ So basically what we're saying then is we take that tooling and we turn it into a ⁓ context appropriate machine. ⁓

Yeah, absolutely. ⁓ So here, think ⁓ if you take, for example, a tool like gamma.app, which produces slides from very short prompts, it's able to give you a comprehensive presentation. ⁓ When you see a presentation that was created in gamma, you can tell right away, OK, that was created by gamma, because they have very They have a pretty good library of different types of visualizations, etc. And yet that library is very repetitive. You see it again and again and so on. The beauty of using Perceptis actually is the opposite of that.

The very first thing that we do when we bring a new account on board ⁓ we ask them to ⁓ upload a bunch of what we call training data, which is their historical documents, historical proposals, ⁓ studies, methodologies, resumes their experts and partners, ⁓ style and brand book. ⁓ We actually take all of the data and we actually disappear for about three business days. What we do is we run a semi-automated process ⁓ on ⁓ digesting all of this data, sanitizing that data, ⁓ labeling all of that data and so on. And then when we re-emerge on the other end of the three business days, actually come back and we say, look, this suite ⁓ ⁓ developed models ⁓ this particular account is now ready. ⁓ then ⁓ happens from then on is whenever a consultant from that account, from that firm ⁓ requesting a white paper, requesting a proposal, ⁓ it's actually generated highly inspired by the historical, you know, slides, by your tone of voice, by the facts and methodologies that you are actually practicing within this consulting company. ⁓ And that's what the new document is going to look like. Our goal, right, is that a client that saw your document a year ago and that they see today, they're not going to be able to tell, ⁓ was kind of, you know, was a different style wise, that it was written by the same person.

Yeah, absolutely. and again, we understand, right, that this data ingestion process is the biggest friction point. if you want to sign up for Perceptis, you have to suspend your ⁓ for business days, share a bunch of data, pay, and then, know, hope that the document that you're going to order is actually going to be pretty good, which, by the way, is what happens ⁓ all the time. ⁓ However, to your point, there will be some companies that either don't have the legacy of like a ton of documents, but they at least would have like a website, they would have their marketing deck, they would have something. And we can absolutely start with something. But, ⁓ you know, even though if you don't have a ton of data or some actually companies are coming to us and they say, do have a ton of historical data, but we actually want to be more McKinsey-esque more BCG-esque, you in our communication and our slides, like the fidelity is just not there. ⁓ And so what we do for them is indeed Perceptis actually is sitting on a training data of 10,000. pages from the world's best consulting companies. ⁓ There's a bunch of sources, essentially is what I'm saying. And so what we can do is, and we have a toggle, right?

And we can crank it up and down, essentially saying all the documents need to look like what you have shared with us versus all the documents need to inspired by other sources. ⁓ so you can get the best of the two worlds.

Yeah, so the biggest questions that ⁓ keep on coming up are exactly actually what you were asking. Like, hey, if I have my competitor using Perceptis, ⁓ our two propose and we're like competing in the same ⁓ RFP, our bids gonna look exactly the same. And my answer to that is no, because we built individualized models based on your term of firms profile. So all the data, all the strength that you have within your firm is actually just make it ⁓ shine in the best way possible.

That's all it is. Another question is data security. And I often ask them, do you use Chat GPT or something like that already? And they usually say yes.

And I'm like, well, guess what? Our security guarantees are going to be even better because of the enterprise agreements that we have with the model providers. Another question that we often have is like, well, but I'm already pretty fast. It used to take me 30 hours. to put together a single proposal, but with Chat GPT, it actually takes me about eight today.

And I'm like, you are copying a ton of text from Chat GPT into PowerPoint, ⁓ Even once you get the right paragraph, etc and the answer is yes. And I'm like, well, we take it down from eight hours to one hour of your time? ⁓ then ⁓ guess the last thing is they also ask, well, can already, some of the more sophisticated users say, well, I already connected. My historical ⁓ documents with Chat GPT, example, right? And I ⁓ you know, ask questions and I can find the right references, right?

Case studies and so ⁓ And all of those great. That's the RAG, simple RAG functionality. Again, the truth is that RAG is still a generic or general purpose tool. ⁓ I always say RAG answers the question of does the data exist, but it doesn't answer the question of how that data can be used. An example of that would be, say, as a consulting company you served AT&T last year and you have somewhere a proposal that might be a little bit sensitive, ⁓ somewhere in your documents, ⁓ now you are going into Chat GPT and you're asking a question like what was the methodology in Telco because you are looking to serve T-Mobile.

And unless there is some smart mechanism in your system that says, yeah, there is something that we have, but like... we need to highly sanitize anything before we actually can reference the data. ⁓ You're going to be screwed. So those are, again, some examples of ⁓ where people come and ask their questions. But I think we have very decent answers to all of them.

So for us, ⁓ maybe just share a little bit how the evolution been the use cases right within Perceptis, we started on the client side of the house. the main reason for that is it serves the partner use cases ⁓ in the best way possible. ⁓ And are the one who make decision about ⁓ paying for the operating system. ⁓ So specifically, know, proposal writing was the first use case, which we nailed down pretty well. But then another thing is a lot of times actually consulting firms are looking to ⁓ check in with their historical clients or maybe clients that they have connections with, even though they might have not ever been able to convert them. ⁓ So what we do is we put a bunch of AI listeners on the website of those clients, their social media, on the websites of their competitors, industry news, SEC filings, etc. And so at the end of the day, what we're trying to do is we are trying to identify signals in the public domain that we can then forward to the consulting company and say, hey, it might be the right time to check in with this client because they just made a senior hire that looks interesting or because they actually declared a hiring freeze, which probably means they don't have enough hands internally to run their initiatives and so on. So that's kind of...

Our you know, our intent there is of course to help these businesses grow but then even if you truly make the difference and if you enable that growth fueled by AI What happens is you actually are pushing the bottleneck of productivity down the down the down the line because you have a bunch of Projects that have converted but you don't have enough people. So now you need to go and Utsav Maybe that's where you know, ⁓ your of expertise is gonna shine ⁓ But now you have we don't have enough hands now You need to go for this 1099 contractors find, know unproven people to put on those projects etc And so this is where we now give you productivity tooling to actually your analysts and associates. So we give Slide building capabilities. We give knowledge management capabilities where you can converse with your historical documents and so on So that's kind of what Perceptis is does today to come back to your question of where Perceptis is is heading Well, what we do is from those use cases we actually go into the adjacent workflows and we you know, we conquer them ⁓ one by one.

A lot of our customers have been asking us, hey guys, when I'm gonna start using Perceptis for building, you know, ⁓ PMO update ⁓ deck, board of director update deck ⁓ for my final deliverables and working documents and so on. ⁓ And that's going to be big because when you go from proposals and white papers, when you go into the world of, we're to be able to create just any consulting deck you can ever imagine, that's a very big technological ⁓ challenge. And we're pretty close to releasing that to the world.

Yeah, think so actually on that front, ⁓ we recently released a white paper, which I'm happy to share with you, Utsav ⁓ But essentially what we did is we went and we talked to a bunch of decision makers in the industry who have been procuring consulting services. ⁓ And we asked them questions like, well, what are your considerations? You know, for how long are you running? When do you run RFP process versus you do sole sourcing? What are the key things that you pay attention to and so on?

So like what shifts you're seeing? There are a couple of things, but I think the biggest trends that we are seeing is that boutiques are getting a lot of interest, know, like it almost like if you propagate what's happening right now and you assume that this trend is going to continue, it feels that boutiques are standing to, you know, take away about 20 additional percentage points of the market in ⁓ the next 10 years. A lot of that is driven by the fact that boutiques are seen as more specialized and they are seen as more ⁓ cost efficient as compared to the, you know, very established big consulting firms. ⁓ And especially, like the historically the difference was that boutiques just don't have the army of people to throw at a problem. ⁓ They might not have the very well established in-house back office processes and so on. But again, with the AI tooling, do feel that that's going to be the major enabler.

Who's going to basically put a smaller boutique and a larger firm on the same playing field. And now, you know, they're going to go head to head for those same types of clients. ⁓ So that's one thing that I'm seeing. And when you talk about, you know, the emergent ⁓ boutiques or maybe even like smaller mom and pop shops, we do see a lot of new entrants who are either like a former BCG partner, for example, going on in their... We see academics at the top of their field who are running consultancies on the site.

And again, historically, they would be able to serve only one client at a time. Now enabled by all this tooling, they can actually expand their revenue to 3x just because they are more efficient in doing things in the background. We see other types of like a ton of fractional CFOs, this type of talent that are also kind of providing this consulting like services. So there is a lot of interesting things that are emerging and happening, yeah, that's at least what I'm seeing from my vantage point.

I mean, the book itself, right? think, Ustav, what I like about what you're doing is that you are ⁓ basically foreseeing a major tectonic shift that's going to be ⁓ taking the industry, you know, as a storm. And ⁓ I like how you pinpoint the fact that AI is going to be the major factor, but also appreciate that unlike, you know, a lot of other visionaries who are... sharing their perspectives ⁓ on the topic, you are not going to an extreme to say that AI is going to be a panacea or that it's going to kill consulting in any way or that the big consulting firms are going to go away tomorrow because of that and so on. You are very pragmatic in saying that, yes, it's going to be a major enabler. It's definitely going to help. ⁓ Shuffling things around in the industry.

There's going to be new winners, there might be losers, losers are going to be dying a end, ⁓ they have a ton of time to try and adjust and ⁓ marry the new world. ⁓ So I guess the pragmatism ⁓ that you maintain in your book is resonating with me a ton, which is why I'm very much excited to be in this conversation with you.

Thank you, Utsav ⁓ Again, really, excited about what you're doing. ⁓ Keep on being this ⁓ of filter for, ⁓ know, between the noise and what's actually a signal ⁓ in the disruptive times. Yeah, and let's stay in touch.