Using AI to Solve Your Commercial Intelligence Problem

How can you make sure that your customer stay content with your economic relationship?

How can you continually elevate the value that you bring to your customers?

And what if AI could help ensure that the foundation that your company has built on is solid?

The answer, commercial intelligence.

Hi, I’m Courtney Baker, and this is AI Knowhow from Knownwell, helping you reimagine your business in the AI era.

As always, I’m joined by Knownwell’s CEO, David DeWolf, Chief Product Officer and Chief Technology Officer, Mohan Rao and Chief Strategy Officer, Pete Buer.

We also have a discussion with Dr.

Andy Workman about what business leaders can learn from the increasingly challenging economics of running colleges and universities.

And we’re going to welcome Knownwell’s new lead data scientist, Ramsri Golla, for a special AI in the Wild on the AI model he created that was featured in the Google IO keynote.

But first, let’s dive into the wonderful world of commercial intelligence with David DeWolf and Mohan Rao.

Mohan, David, every business in the world is selling something.

And in some ways you could say every business has a commercial intelligence problem.

We’re always trying to find what the thing our customer wants, what the value they will pay for it, and then to deliver that.

And so I wanna talk about that concept.

If we all acknowledge essentially selling is a problem, figuring these things out is a problem.

How can AI help us solve our commercial intelligence problem?

So I think some table setting might be actually helpful here in terms of, let’s define some of the terms you just used there, right?

Because I think sales intelligence is a very real thing, what Sixth Sense, for example, is doing around really understanding who’s in the market buying and changing the game is a real thing.

Commercial intelligence is not the same as sales intelligence.

Right.

Client or customer intelligence is a real thing.

Knowing the dynamics of your customer themselves.

When I speak about commercial intelligence though, I’m talking very specifically about the relationship between a service provider and that client.

And so while those things, sales intelligence, customer intelligence, right, all sorts of these intelligences are relevant and adjacent and even overlapping, I think what’s unique about commercial intelligence, when we say the word commercial, we’re actually talking about the economic relationship between two parties.

Do we really understand the DNA of the economic relationship between two parties?

And why do you want to know that commercial intelligence?

You want to know it for really three fundamental reasons.

Number one, you want to sustain that economic relationship.

Number two, you want to grow that economic relationship.

And then number three, what we find what the research shows is not only our organizations that are stronger at commercial intelligence, more apt to retain and grow, they actually become more profitable because they’re able to do things internally more intelligently to seek the end goal that they’re after, which is to serve those customers and drive economic benefit based on that value that created.

That’s the whole purpose of business, right?

And so commercial intelligence by all of the research has been identified as something that’s really important for driving results, yet we haven’t necessarily excelled in it and really specifically focused on that economic relationship.

And so I think that frame can be helpful in understanding it and how AI can help out.

Yeah, thanks for that table setting.

So that was super useful.

Thanks, David.

So coming back to how AI can help with commercial intelligence.

So if you think about it, I’ll put it in two words, right?

It can enhance and augment several things, right?

It enhance and augment.

So those are the two key things that AI can do for us.

For example, there is a lot of data in the commercial exchange that happens between a customer and a client.

For example, in data processing, in pattern recognition, in predictive analytics, in using natural language processing, in decision support.

These are all various items within the life cycle of this relationship that David was describing that AI technologies can truly enhance and augment each one of them to a higher level than the technologies we’ve had in the past.

The other word that comes to mind, I love the words enhance and engagement, but I do think understanding too, right?

That word understanding, I think it can really help us find signal and noise and connect dots and understand at a deeper level because it’s doing so much of the information processing that we simply can’t.

Exactly.

I mean, the ability to match patterns at a scale that the human brain just cannot, right?

It’s phenomenal.

The ability to understand natural language, right?

So it is phenomenal.

The way it can help a decision maker through the support system, through more context, is phenomenal.

The way it can learn and become better over time is unprecedented.

So these are all things where an AI system makes so much sense to employ in this entire exchange of services between a customer and their clients.

One of the things I love about it when you speak to that is all of that understanding, all of that enhancement, all that engagement, what it really leads to is better service, better, you know, and so it’s not just good for the service provider, it’s actually good for the client and it’s good for all of the human beings in the middle.

Right, it’s a positive gain that really takes our ability to be service providers and serve our clients to the next level.

It’s not a, you know, win-loss equation, it’s a win-win for everybody involved.

Now we’ve talked about client centricity for a very, very, very, very long period of time, right?

With these technologies, it’s truly possible if you were so inclined to become more client centric, to understand the true dynamics of the, between what’s happening between the service provider and the client, truly you can move towards this client-focused leadership with these technologies.

You know, it’s interesting that you bring that up, Mohan, because I’ve been thinking for a while about this word client centricity, customer centricity.

We talk about it all the time.

I feel like historically it’s mostly been a mindset that we talk about.

It’s how companies think.

I think because of artificial intelligence, because of the emergence of this category of commercial intelligence, we actually are able to operationalize that now and not just say, hey, we have this inclination towards, we have this bias towards being customer centric, but we’re able to say, actually, no, we have the data, we have the intelligence, we know what customers are looking for, we know what our clients want, we know how to serve them better.

And because of that, we can plan and prioritize to do that better.

And so now we can be operationally client centric, which I think is just a huge accelerator.

Yeah, other than walking the halls of a client, there was no other method, right?

So there was always a big screen between you and the client, and you had to guess what they were thinking, right?

What these technologies allow is sort of a more transparent screen, so you can operationalize what they’re thinking, what their feedback is, what you think you can offer, and just meld the two together much better.

I think what you’re ultimately saying though, Mohan, is this new technology is gonna help us solve what we set off at the beginning of this episode is to solve that commercial intelligence problem by actually having intelligence on our client.

And I think that’s something, again, where we don’t necessarily see it as a problem, but if we do, then it’s like, oh, all of a sudden we have this new thing that can help us solve it in a way we never dreamed would be possible before.

And I, for one, think it’s really exciting.

It’s truly amazing that you can move the client focus to a much higher level than you could ever before.

Measure it, understand it in depth that before we couldn’t even imagine.

And it’s fun to see it come to life, no doubt about it.

And I think it’s gonna take all of our business experiences to the next level.

Yeah, absolutely.

And I think as you talk to executives in professional service firms specifically, it’s like when you talk about how they know their clients, you actually start to expose the gaps, but it’s so ingrained in just how business is done, how we’ve done it.

It’s so easy to get lost in that and not realize that, oh yeah, this could be so much different, so much better than anything we’ve dreamed of or utilized to date.

You know, let me add one other thing, right?

What the technologies, what AI can do, especially with NLP, with natural language processing, is the ability to understand human communications better in terms of sentiment, intent, context, right?

And be able to bridge the gap that we’ve never been able to have before.

Yeah, and I think it’s helpful to always put this in a lens, you know, especially for Knownwell and what we’re doing, it’s to help humans go do the thing only humans can do and engage in relationship.

And so, thank you, David and Mohan, for solving this big problem for us.

I appreciate it.

Lots of fun.

Awesome, thanks.

Are you at the point, Mohan just raised, where adding more people to manage your client relationships just isn’t producing better results?

If so, you are a perfect candidate to be one of the first people on the Knownwell platform.

We’re revolutionizing client retention and growth, and we’d love for you to check it out.

Go to knownwell.com to learn more.

Ramsri Golla is Knownwell’s newly hired lead data scientist, and he’s still glowing from his recent appearance in the Google IO keynote.

He was interviewed by Google about an AI model he trained on 15 different languages.

Very impressive.

I couldn’t pass up this opportunity to hear about it for another installment of AI in the Wild.

Ramsri, welcome to AI Knowhow.

It’s so good to have you.

Thank you.

It’s my pleasure to be here and very glad to be here.

So me and a friend, Ravi, we worked on an open source model, which is called as Navarasa.

Basically it’s a fine-tuned model for Indic languages, where we have taken 15 language data sets, Indian languages, and we have fine-tuned a model built by Google called JAMA.

So we trained and further trained on those data sets on top of that model, and we released that model.

We got featured in Google IO for the same, and that’s the high-level introduction of Navarasa.

Listen, Ramsri, you may have to really boil this down for as functional executives out here.

Help us understand, what does it look like when you start training the model on all these different languages?

What is the process that you go through to take that on?

Sure.

So I think, in essence, we have ChatGPT, which is like a question and answer engine.

You ask a question and it responds to that.

So the biggest thing is that when it comes to Indian languages, the performance is not that great and it’s not effective, it’s costly, it’s slower as well.

It takes more amount of, let’s say, tokens or compute than generating the same thing in English.

So in order to overcome these, we need to curate a focused question and answer dataset for Indic languages because culturally the customs and everything are different.

So if you have a question and answer dataset that’s tailored, that’s perfect.

So what we did was we took existing English datasets, we translated them to multiple languages.

And these datasets are question and answer datasets.

For example, what is photosynthesis?

But of course, in regional language and the explanation of it is also in regional language.

Similarly, any math questions, etc.

Regional question and answers, converting these datasets into regional question and answers.

And you collect, for example, 10,000 pairs of these question and answers.

That’s your dataset.

When I mean dataset, you have question and answer, question and answer.

And you have 10,000 of these in several languages.

For example, if you take Hindi, you have just in Hindi, the question and answer, Tamil, you have similar or Telugu, any of these languages.

And about 15 languages of India, that’s what we have collected.

And now you fine tune the model.

What that means is that you give question as input to the model.

You expect that to generate the same answer that you already have and the model’s weights will be adjusted to be able to tune to the question and answer that you’re giving.

And as you’re giving them in batches, the model will learn.

And one of the things is to train only fewer parameters of the model so that you can train in a single GPU machine, etc.

And that’s how it is trained.

This is technically called as supervised fine tuning.

It’s a bit like aligning the model to question and answer format where it responds to a question in an answer format.

So how long did it take from the inception of we’re going to go do this to actually having it done?

Like, what was that timeframe?

Great question.

So first, when we started from scratch, we didn’t have the complete fundamentals of this training a model.

One is the theory of it.

Second is working code that actually works, you know, so you curating the right code frameworks, data sets as well to train.

So all in all, it took about four months of time, but during that process, there are several steps that is experimenting with different models, creating data sets, then creating the training code that works end to end and doing some iterations on it.

So probably three to four months of learning, experimenting, curating data sets, and then training and releasing the model.

Well if you’re interested in connecting with Ramsri, you can find links to his LinkedIn and Twitter in the show notes.

You can also go to knownwell.com and look at our staff page to connect with him there as well.

Thank you for joining us.

God bless you.

Pete Buer talked with Dr.

Andy Workman recently about how AI can be used to tackle big business problems like stagnant revenue, high labor cost, and increased competition from all sides.

Andy, hi, welcome to the show.

So glad to have you here.

Just to set the stage for the episode, could you give us a little bit of background on your role at Widener and how and where AI fits into your work?

Sure, so I’m the provost at Widener.

Widener is a mid-sized university, about 6,000 students.

We have a school of arts and sciences, schools of business, engineering, nursing, and then graduate programs too, and health and human services, and two law schools.

And so AI is, I think, really essential to our challenges now.

I see a lot of opportunities in it.

And I think it’s gonna transform the key things that we do.

What do we do?

We distribute knowledge, for lack of a better term, that we teach.

I think it’s gonna change how and what we teach.

We produce knowledge.

So it’s gonna change how scholarship is done.

And also I think that it can solve some of the key problems we have in higher ed in terms of our cost.

I think it can improve the way we do everything from admissions to financial aid to just the basic business processes that any institution has.

And one of our big burdens now is regulation and accreditation, and we’re already seeing some real positive benefits from AI in those areas.

Well, I’m glad you said the word cost because many of our listeners are corporate executives who are wrestling with the pressures of the economy on their cost base and also the occasion of AI as a way to address that and rewrite the economics of the business.

And if you sort of think about it, I guess universities and businesses aren’t so different.

There’s a fixed cost base to some degree, especially with tenure.

There’s pressure on performance and to do more with less.

So I was hoping you could go a little sort of deeper on some of the ways that AI is gonna be used to manage the cost base.

That’s a great question.

Yes, and we are in universities where we’re like other businesses, but we’re perhaps more of our costs are centered on labor than most, about two thirds of our costs are on labor.

And it’s not just faculty members, but it’s staff.

We have, so the best way to put it is this, quality higher education is intensely relational.

It’s the relationship of a faculty member inside and outside the classroom that provides that transformative experience for students, but also our staff in terms of everybody from student life to athletics.

Those relationships for undergraduates are kind of secret sauce of higher education.

It’s not just that we’re putting some information in people’s and giving people some skills.

We’re actually helping them to mature into effective adults in all parts of their lives.

And in our professional schools as well, those relationships are absolutely essential to developing students.

Where can AI help with that?

It may seem paradoxical, but I think AI can help us to maintain that intense web of relationships.

And it’s been very hard because we really are very flat in terms of our ability to adopt new costs.

We can hire just more people to have those relationships.

But AI can do a number of things that can help us.

In fact, to spend the time that we have with the very expensive labor we have on those one-on-one connections with students.

And so how can it do that?

Well, in some areas, we think that it can improve the basic communications we have with our student population.

So even before the recent advent of the kind of LLMs, we found that chat bots can have a tremendous impact on student success and student retention.

With the tools that we’re seeing coming out now, we think that’s even more, universities are very complex, difficult to navigate institutions.

And that has a cost for students and it has a cost for their ability to succeed here.

We think that if we really invest in these, we will be a more transparent and more effective institution for students.

Another good example of this is adaptive tutoring.

So you’re probably familiar with Conmigo.

And we’re just doing some experimentation with that.

We’ll do a pilot with a math class in the fall.

But we see that as something that can really displace a lot of very intensive, slow one-on-one labor with a 24-7 chat bot that adapts to a student’s needs.

And from our early investigations of that, we think there’s some real potential there as well.

And then there’s a lot of academic labor that is just kind of routine and rote.

Things like grading, some of the kinds of communications that faculty and staff have to do with students.

We think that there’s some real possibilities for displacing some of that labor.

And what we’re seeing is the ability to, for example, create syllabi and do some basic research, to do some of the things that are gonna help faculty members spend their time in their classrooms engaging with students in a different way.

So, and then of course, the things that I think other business leaders are seeing in terms of business processes.

How we communicate with the thousands of people who apply to our university.

How we address their financial aid issues.

And how we just do all the various and sundry things that we have to do in a business to operate.

We’ve already seen some efficiencies coming out of our early adoption of AI.

That’s awesome.

Thank you for taking us through the list.

Where did the list come from?

When I talk to executives at companies that we engage with, you know, maybe there’s an AI task force or the executive team takes the job on itself of figuring out opportunities and threats of AI.

Sometimes they’ll bring in a third party consultant.

How does that happen in your world?

How did you get to the bottom of the smartest AI applications imaginable?

That’s a really great question because AI we think is very different from, say, just bringing in a software package or a training program that you might for some other kind of change.

And partially it’s because it’s still so unknown.

We don’t know what it can do for various parts of the institution.

And some things that you think it would do great, it doesn’t.

Other things that you think it couldn’t possibly do, it does well.

So the model that we’ve adopted is once we, as a leadership team, our presidential leadership team, had decided that this was something really important, and I can talk about that a little bit in a minute, is to create ways for the organization, or for a very complex, loosely connected organization, to learn.

So for example, we’ve provided AI hackathons for our faculty and our staff.

My deans have all challenged their staffs to automate processes that they’re doing now with AI, and then we present the findings, the successes of the people, and they spread through the institution.

We’ve given presentations to our board of trustees so they fully understand what we’re doing.

And almost all of those presentations are hands-on, people actually using AI to solve real problems.

And that’s been what I think is really transformative.

And so we’re seeing it, it’s now running through the institution, you’re seeing it pop up in places that you didn’t even know people were thinking about.

So our goal is to move the entire institution from early adopters through the people who are still, I think, very resistant to the whole idea of AI in such a way that we’re on the cutting edge of universities for this.

But we think we’re a learning organization, that’s what universities are, right?

So we’re taking that seriously and providing tremendous opportunities for people at every level of the institution to learn.

If we can shift gears just a little bit, we talked to companies a lot about how AI transformation is gonna change the way they think about their talent base, different profiles, different roles, different skills, different education, different training.

How are you thinking about that in higher ed?

How are you thinking about preparing this next, I don’t know, cohort of future leaders for a world that’s gonna be very different in terms of-

This, I think, is the biggest question for higher ed now because it really speaks to our very business model because if we are producing knowledge workers, which that’s pretty much what we do, right?

And those knowledge workers’ work is gonna change tremendously.

And some of it, in fact, is going to be, it already is, displaced by AI.

The people who are coming out of our institution need to be experts in using AI to enhance their talents.

And so, one, so they’re not replaced.

We also think that our students who are graduating need to understand AI at a deeper level.

We think a lot of people use AI very naively.

They put in a question and something comes out and then they use it.

That’s really not helpful.

Those people are going to be replaced really quickly.

But if someone really understands how large language models work, understands how to craft prompts, and understands the ethical issues that are involved, they’re going to be an incredibly attractive worker, we think.

And so, some of the things that we’re doing.

So for example, we have about, we don’t know exactly, but 850 to 900 students coming in in the fall.

Huge, one of our larger classes coming in, a lot of excitement there.

In the orientation in the fall, we’re doing a 90-minute sessions led by Sam Amorimando, who’s a good thought leader in K-12 education, which will introduce them to AI and what it can do.

We’ll do some really intensive work on the limits of AI and academic integrity, and we’ll in fact set a baseline for all of those students.

That’s going to really last for their entire career.

So that’s, and we’ll build on that every year as those students go through.

Let’s look a little bit into the future.

You’ve talked about a whole bunch of interesting things that you’re doing at the school now.

What do you see on the horizon?

What’s the sort of next fundamentally different, interesting application of AI for your context?

See, here’s something I’ve been thinking about a lot, and it actually has to do with research, with the production of knowledge of scholarship, because we’re already seeing faculty who are using AI to do data analysis, for example, to summarize articles, to write drafts of their articles.

And doing that, we think can be transformative.

We think that we can really accelerate the production of knowledge.

It can also be really a bad thing, because we’re seeing cases where people are, imagine this, someone uses AI to analyze data, then uses AI to write that article, which is then going to be sent to a publisher, some of which are already using AI to sort through articles and summarize them, and to make decisions.

You’re getting humans out of the loop, and also there’s a risk of really reinforcing convention and not really creating new and interesting things.

So we have to figure that out, and we have to give faculty the tools, just like we do with students, to use these tools in a very sophisticated way to understand their limitations, to understand the dangers of hallucinations, but then to gain the real, I think, advantages that this can have for scholarship.

We also have to realize, we have a mode of assessing faculty productivity that, despite the advent of computers, is still, I think, in the typewriter and cutting and pasting ages.

What about the velocity of the research is going to change?

What are the possibilities for people who are unleashed with these new tools?

So that is something I see unfolding over the next several years and something we’re looking at very closely.

It’s been such a positive story, but you started to touch on concerns.

What do you worry about?

I’m worried, and this is, I think, a lot of our faculty, when they first saw like GPT 3.5, are worried quite legitimately about academic integrity.

But what you’re really worried about there is that students don’t learn.

The reason we design students’ papers and research projects and tests and problem sets is not because we want them to do those things, it’s we want them to learn.

Those are tools to learn.

And if you short circuit that learning by using the tools of AI and we know that there are students who are doing this, then we’re really going to come have a generation of students who really are not able to function at the intellectual level that they need to.

As I said, using AI naively can have those impacts.

And so I think we really need to figure out a way to help students to understand what they are doing, to help faculty to understand that you can’t just ban AI from the classroom.

It’s already there.

And to craft education in a way where that learning happens even as AI tools are used.

And that’s gonna be the pedagogical challenge, I think, in the next five years or so.

So many interesting parallels, so many clever thoughts, Andy.

Thank you so much for joining the program today.

It’s been a pleasure.

Sure, it’s great.

Great.

Thanks as always for listening and watching.

Don’t forget to take a moment to give us a review.

It really helps other people find the show, and legit, we appreciate it.

Hey, Perplexity, what’s going on?

Welcome to the show.

This episode, we’re talking about tapping into AI to improve commercial intelligence.

What are your thoughts?

Hey there, great to be on the show.

Thanks for having me.

When it comes to tapping into AI for commercial intelligence, I’d say it’s a game changer that can give businesses a serious competitive edge.

AI can crunch massive amounts of data to spot market trends, predict consumer behavior, and optimize pricing strategies way faster than humans ever could.

And now you’re in the know.

Thanks as always for watching or listening.

We’ll be back next week with more AI applications, discussions and experts.