Would You Rather…? (The AI Edition)

Would you rather have an AI-powered decision-making system that’s 90% accurate, but lacks transparency, or one that’s 70% accurate, but fully explainable?

Would you rather invest in AI that can predict your next big revenue stream by 30%, or one that improves your current operations by 70%?

Stick around, because on Episode 75 of AI Knowhow, we’re going to tackle the answers and more.

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 CEO, David DeWolf, Chief Product and Technology Officer, Mohan Rao, and NordLight CEO, Pete Buer.

We also have a discussion with Abhijit Mitra of outreach.io about his company’s foray into agentic AI for B2B sales.

But first, drop an egg in the frying pan and turn the stove up to 11 because it’s time to welcome Pete Buer to the studio for a new segment we’re calling This Is Your Brain on AI.

Any questions?

Pete Buer joins us as always to break down the business impact of some of the latest and greatest AI news.

Hey, Pete, how are you?

I’m good, Courtney.

How are you doing?

I’m doing well.

According to a recent study from Microsoft, over-reliance on AI tools may affect our critical thinking skills.

Pete, I think this is so interesting.

What’s the takeaway here?

Yeah, I love this article too.

So, Microsoft and Carnegie Mellon studied 300 knowledge workers and found that using generative AI can lead to long-term reliance on the technology and diminished independent problem-solving skills.

Maybe not so surprising when you step back and think about it.

We’re not doing the long-form math that our spreadsheets do for us anymore, right?

And at some level, it’s not really a new thought or a new concern either.

For millennia, humans have worried about technology and its impact on the quality of deep thinking.

Socrates, in particular, objected to putting thoughts down on paper for fear of atrophy of mental acuity.

You can imagine his take on forcing thought into 280 characters on Twitter.

But this notion of critical thinking loss is an important one.

It sits at the heart, of course, of educators’ concerns about AI.

And we’ve had conversations in this setting before about that.

And I think it’s a relevant concern for business as well.

I also think there is a planful solution for getting after it.

And so the message for business leaders listening, we should take a deliberate look at the skill sets that define capability, mastery, success in the business setting, or, by the way, for that matter, education.

And in both cases, make sure that we have mechanisms in place to ensure that our learners, be they students or employees, are becoming proficient, even if they happen to have technologies like AI working for them.

So maybe the real audience to take away a message from this one is HR or L&D, to think a little bit differently about what belongs on the learning agenda for the upcoming year.

I feel some more New York Times crossword puzzles in our future.

Yeah, it’s so interesting.

And yeah, I love the thought about how does HR start to think of, how do we offset maybe some of the degenerative knowledge that we have with generative AI.

Pete, thank you as always.

Thank you, Courtney.

When your podcast hits the 75th episode, it’s time to let loose and have some fun.

But seriously, all kidding aside, we have fun making every episode.

But this time around, we wanna have even more fun than usual with an AI-themed version of Would You Rather.

People that listen to this show for a while know that in general, we’re a little competitive, this group.

We are going to play the game, Would You Rather?

And I think you both are familiar with the gist of the game.

So just in case you missed it, or just in case you’ve never played Would You Rather, we’re gonna give you basically an option.

And so just for an example, it might be things like, would you rather work for a boss who relies entirely on an AI assistant for decision making, or a boss who refuses to use AI and still prints every email?

So get the idea, good?

Everybody listening?

Still prints every email.

That’s called a dinosaur.

I think people do still print out emails.

I wouldn’t be surprised.

Okay, they’re probably not listening to this episode.

So are you two ready?

Yep.

Because you’re going first, and then y’all are going to get your own prompt.

Okay.

So first up, would you rather have an AI-powered decision-making system that’s 90 percent accurate, but lacks transparency, or one that’s 70 percent accurate, but fully explainable?

So 90 percent accurate, but not explainable, or 70 percent accurate, and explainable.

Is there a way in the 90 percent, do you know what 10 percent it’s not good at?

No, you don’t.

You just have to use your own intuition.

That’s right.

I think I’m leaning towards the 70 percent because I know the 70 percent is real.

I’m going to take it and then run with the other 30 that I need to deal with because I have so much confidence in 70.

Because if it’s the 90, then I’m always second guessing myself.

I don’t know.

This black box told me that it’s do this and I’m just doing it.

I don’t know.

I don’t know.

I would rather go with the 70.

I think your answer should also be though, you want to still have a job.

30 percent, that’s like a paycheck.

I’m thinking, Mohan, that if it’s 70 percent, yet it’s explainable, that means you’re going to have to look into the rationale behind every single decision to determine if this is part of the 30 percent or not, because you’re not going to execute on 70 percent confidence.

You’re going to wait until it’s 90 percent confident, then execute.

And so I’m not sure it’s going to give you many efficiency gains.

I think you’re going to be stuck.

So I think I’d take a 900 batting average, especially when you layer it on top.

I think my intuition and my gut could recognize the 10 percent I need to be careful of.

But even if I couldn’t, I’d take a 900 batting average.

Yeah, but you don’t know which 90 percent is it, right?

So you’re always going to be second guessing, David, because you’re going to be like, is this like in the 10?

Is it in the 90?

This black box is telling me, do it.

I’m just doing it.

Well, I don’t know that I would just do it.

That’s a good point.

But if it was giving me its rationale, I’d probably have to dive deeper.

I don’t know.

I, hmm, hmm.

So maybe what we should do is we could have a second AI thing that checks on the first AI thing and gives us a second confirmation.

I was going to say, what if you take the 70 percent, I take a 90 percent, and we just collaborate together?

That’s right.

And then see, oh my god, I mean, if you, if you layered the 90 percent and 70 percent together, so maybe, maybe that’s the answer.

It would be pretty good.

So Mohan, final answer.

I think I’m still sticking with the 70 percent, but I know that the 70 percent has some defensibility to it.

Yeah.

I think I’m going the other way.

I am going to go with the 90 percent, and I trust my intuition to catch the other 10 percent, but even if I don’t, I’m going to hit at least 900, and that’s a really good batting average.

All right.

All right.

Oh, my God.

This is such an awesome Would You Rather question.

So let me ask you both, David and Courtney, would you rather have an AI system that predicts customer churn with 100 percent accuracy, but only gives you about 24 hours notice before the client leaves, or the one that’s vague churn signals months in advance?

Definitely months in advance.

No doubt about that to me.

I want to get ahead of problems.

Wait, months in advance, but how accurate was it?

No, it’s very vague.

It’s like, yeah, they may leave, that sort of thing.

Yeah.

I want early warning signs.

I don’t think client problems ever get better with time.

I don’t think you want to take a two-ounce problem and allow it to become a 100-pound problem.

I think you want to tackle it early on, even if it’s a false alarm.

I’ll take false positives.

No doubt about it in my mind.

You’re taking the vague turn signal months in advance.

Definitely.

I’m trying to come up with a scenario, you would choose the 25.

It’s not even 25, it’s 24.

Oh, my gosh.

100% accuracy, but just 24 hours.

You know, of all the numbers to misquote, you would think the number of hours in a day would be one of the easier ones, to get right.

Yeah, 24 hours, like what scenario would that…

Yeah, I just don’t know if that works.

There’s just not enough time.

You basically have to throw a Hail Mary pass and pray that you picked the right receiver.

You know, like, it just…

Basically, that option is saying they just terminated, so sorry, like all you’re doing is finding out early that it’s gone.

There’s nothing you can do in 24 hours, right?

There’s like a modified Would You Rather now, right?

So it’s 100% accuracy, but it’s with more notice.

It’s not 24 hours, but you pick the number that you want.

Like, let’s say it’s a week or two weeks, and would you rather go with 100% accuracy two weeks in advance or one that’s so vague, but months in advance?

I would take the 100% if I could have a month.

Hmm.

I was going to say six weeks, but I also would caveat say, I think it depends what kind of business you’re running.

I think it really…

You were in, this is a professional service firm.

What kind of professional services?

We…

Oh, Mohan gets to make the rules.

So Mohan tell us…

I like how you’re stepping in for him.

It’s like, just take the reins.

That’s Mohan’s favorite thing that I do when I just take over.

He loves that.

Yeah.

David, you were going with depends, so expand on that.

What do you mean?

Well, I think the type of business matters.

I think the more transactional it is, probably the shorter amount of time that it takes to be able to solve the issue.

I think the more strategic the relationship is and the type of services, the more time you want to turn it around.

So I think there’s a spectrum there that somewhere between the two weeks that you said and four months of time is the right period and depending on the business that you’re in, you probably end up at, you know, in the, if you look at the bell curve, it probably hits around six to eight weeks most likely for the average professional services company that’s in the, you know, technology services, marketing services, kind of white collar knowledge worker service area.

I would think, you know, in that range, six weeks to three months is probably in the bell curve.

So with the modified, Would You Rather, Courtney, you want to recap your answer and David, you recap your answer?

What is the new version though?

Is it what David just said?

Is that what we’re going with?

It’s a…

Now, we’re going with 100% accuracy, with four, 30 days notice, or four weeks, or wig and months in advance.

Oh, I’m taking the 100%.

Okay, David?

I’m taking 100% in six weeks.

That’s what I meant, actually.

I didn’t know we could do that.

But I will say, I mean, to David’s point, in an e-commerce business, I would absolutely take 24 hours with 100%.

It would totally work.

Yeah, you just kind of say, next item is free, next thing is free.

You can kind of do so many things, right?

I mean, literally, that’s how e-commerce, that’s what they do today.

They’re like, when you click cancel subscription, they’re like, but do you want it forever for free?

And you’re like, well, I guess.

Until two years later, you start getting these $39.99 charges every month.

And you’re like, wait, what is this?

So yeah, 24 hours, that’s like an eternity.

Okay, David, you’re out.

All right, guys.

So everything you’ve been talking about is leading up to the point of actually driving a great business.

I want to pose a question, which is about the bottom line of business.

Would you rather invest in an AI that can predict your next big revenue stream with 30% accuracy or one that improves your current operations by 70%?

Okay.

My initial gut was to go current operations by 70%.

Yeah, it depends on what the current book of business is, right?

So as you think about this, if you have a pretty mature book of business, you want to kind of do the 70%.

But if you’re still early in your revenue maturity, you probably want to do the 30%, right?

Because it sounds like, hey, you got this great ideas and you’re gonna get this 30% boost in revenue stream, right?

So-

Wait, is it a boost?

Wait, is it a 30% boost or 30%?

Predict your next big revenue stream by 30% accuracy.

So it’s-

How?

You saw 70% that it’s not a good-

Yeah, so basically, one out of three big ideas that it produces for you, pay off.

Okay, David, what kind of business are we?

Give us our-

What kind of business do you want to be?

Okay.

You are-

We are-

I think you are a marketing services business, Courtney, and Mohan, you are a technology service business.

How’s that?

Great, great.

But am I wildly successful?

You’re about, let’s call it 24 million in revenue.

Over the past couple of years, you’ve been growing 20 to 24 percent year over year, and you’re moderately profitable.

Mohan, you’re a technology services business.

You’re only growing at 12 percent, but you are 64 million in revenue.

And what’s the second thing?

What is the 70 percent?

Say that again.

The AI that will improve your current operations by 70 percent.

So 70 percent increase in bottom line.

I am definitely taking 70 percent because it is so definitive.

I would rather have my EBITDA go up like crazy, and I’m going to be super happy.

I’m 70 percent final answer.

It’s so surprising.

That’s not what I thought you would choose.

What are you going to choose though, Courtney?

Well, I was thinking I would still choose the 70 percent because usually a lot of times those stage of businesses, there’s still enough like startup growth engine, you know, like your things are working, a 20 to 24 percent growth rate is not, that’s pretty decent.

It’s usually those like operational maturity to drive profitability that’s just not there yet.

And so getting that on track and being able to scale feels like the right move.

Do you think I missed anything, Mohan?

No, I think what you didn’t narrow in on, I said you were kind of bleh, moderately profitable.

I mean, if you only have a dollar of profit and you’re increasing it to $1.70, I’m not sure that’s going to pay off that much.

But I think you were assuming probably more and that’s probably reasonable.

Maybe you’re doing, we’ll call it 8 to 12 percent EBITDA.

And so you’re going to boost that, get it up to 15, 16 percent, that could pay off.

Yeah, I was not paying attention about my profitability.

I was totally focused on the top line, which a lot of people fall into that trap.

Yeah, yeah, yeah.

So for everybody listening, I know I’m going to be thinking about this fake scenario the rest of the day.

Did I do the right thing with my marketing firm?

And we would love to hear from you.

What would you do?

What would be your choice?

If you’re watching on YouTube, make sure to post in the comments or you can find us on LinkedIn.

We’d love to hear what you think.

David Mohan, thank you as always.

Would you rather end this episode now or stay in chit chat for a little bit?

That feels like there’s a wrong answer.

Would you rather have an AI platform that gives you warning signs when your customers are at risk of leaving or an AI powered platform that helps you make sure your teams are working on the most important items that will help your company grow?

Well, it’s another trick question.

Why choose when you can have both?

Go to knownwell.com to find out more and find out how you can see your very own data on the Knownwell platform.

knownwell.com.

Abhijit, welcome.

We’re so glad to have you on the show.

I’m so happy to be here as well, Pete.

For those who may not know Outreach, may I ask for a little bit of background on the company?

Yeah, absolutely.

Outreach is a sales execution platform.

We help automate work for our salespeople, helping them generate pipeline, working closing deals, forecasting for both the new business, as well as for expanding your existing business.

All of this is guided and coached by AI.

It’s a SaaS platform, and AI is at the core, and then we automate work for salespeople.

My crack team of researchers tell me you recently got a promotion, and so congratulations are in order, but can you tell us a little more?

Yes, of course.

So I joined Outreach as president of product and technology a year and a half ago, and then about six months ago, I became CEO here.

So my background is coming from more the product and business side.

Prior to Outreach, I was a GM at ServiceNow.

I created some new businesses there, which is their biggest business now.

Before that, I was at SAP and Oracle.

Well, that’s quite a background.

So this isn’t the first time on the show that we’ve talked about agentic AI, which is our focus for today.

It is the first time, however, that we’ve talked with someone about their own AI agents.

So could you give listeners a feel for Outreach’s AI prospecting agent?

So what we have done is we have looked at the what we call the sales execution workflow, which is a day in the life of a salesperson and their manager.

What do they do?

We have automated this workflow all the way from, like I explained before, generating pipeline, or working on deals and forecasting and so on and so forth.

Now, there are various stages in this workflow where AI is kicking in and helping either do something for you or getting you some insight and so on and so forth.

We don’t believe that AI will actually replace sellers.

That’s not our belief.

We believe that AI will make sellers better.

So now every seller can be your best seller.

That’s what we believe in.

So what we have done is we have taken steps in this workflow, in this sales execution workflow, and automated them using agentic AI.

And one of these steps is the prospecting step.

And so that’s where the prospecting AI, prospecting agent kicks in.

And what that does is it basically looks at all the data that’s available both within Outreach, as well as data that will sink into Outreach from, let’s say, connected CRM system or third-party systems.

So all of their data is gathered in our data cloud.

We actually give a data cloud to every customer.

And then looking at the data, it now tries to help you identify, for example, if this is your ICP, okay?

You want to target customers of a certain size in a particular segment with certain persona who are interested in certain things.

And there are some buying signals that are looking for the company’s hiring or they had a great quarterly results, if it’s a public company and so on and so forth.

So you say that, hey, these are the kind of companies I want to look at.

And it basically helps you to do the research and bring those companies and even the people with those titles that you’re looking for to you.

And say that, hey, are these the people and are these the companies that you’re looking to target?

And you say, yeah, that looks good.

And it could be new businesses completely that you don’t have any relationship with.

Or it could be your existing customers and then either you’re selling new products into your existing customer install base or you’re talking to new people.

So all of that research it does for you, brings those accounts and prospects to you.

And then says, let me tell you what you need to talk to these people about.

Because you’ve told me this is what you’re going to sell.

This is your product content.

And based on what you told me that, oh, they had a good quarter or so, or they just announced a partnership, or this new product line just launched.

So you may actually want to say XYZ.

So it actually does the personalization for you.

And then it connects all those messages into what we call sequences.

And sequences is something that invented at Outreach.

Essentially, it’s a workflow of a set of steps that you do, either automated or manual, to reach out to your customers and get them interested and warmed up, or to have a sales discussion with you, okay?

So essentially, now it knows who you’re looking for, what’s your criteria, what are those signals, the message that you will be using, and then it now connects those messages to the sequences and gets those leads warmed up for you, and then hands them over to you as warm leads so you can now actually have a meaningful conversation.

So that’s what the AI prospecting agent is all about.

That is amazing.

I’m imagining being a salesperson, and I’m thinking about the tech stack, and all the different places I have to go to try to pull information to get some kind of a profile of my ICP, and then the work I have to go through to do outreach.

Does it explode heads when you introduce this to sales teams?

Yes.

So we’re seeing a lot of interest from sales teams, both from the reps and managers to adopt this technology.

Now, the interesting thing is that, when you think about it, and you think, oh, so all this work is done by the AI, so the seller doesn’t have to do anything, we hear a lot of different feedback on that.

So, for example, in the same company, we hear that if for those high-value deals which needs really hand-holding and relationship-based selling, we actually don’t want the AI to do all of these things by themselves.

We want the AI to come back with suggestions and data that we want to review and make sure it makes sense.

And then, as a rep, I will make the next step.

I will take the next step.

Or as an ops team, I will configure what the reps can do or cannot do at various stages.

And then we are hearing that, oh, for our high-volume, low-value business or where we don’t have people, we just can’t get to them, we want the AI to do more.

And so, there is this configurability between completely autonomous mode, high-volume mode versus a lot of hand-holding human supervision, that customers are really interested in.

So there we see a lot of interest.

The other thing that we are seeing a lot of interest in is, normally, intuitively, you would think that there are a lot of startups doing something similar, like Point Solutions, which does some parts of this workflow.

Outreach obviously puts together this whole thing into an end-to-end platform, that’s our differentiation.

But when you think about it, a lot of people are actually thinking about new accounts, new customers, new business.

Whereas we know that a majority of new revenue from many companies is from the existing customers.

So there we see a lot of interest from businesses in adopting this for essentially making their existing customer, interacting and doing more with their own existing customers prospecting into that.

You used the frame earlier, a day in the life, and I’m imagining the steps in that frame from when the day begins to when the day ends, and all the different things that a salesperson engages in, all the different data sources that they’re pulling from, all the different objectives that they’re trying to achieve.

How did you decide where to invest your time and energy to enable the salesperson’s day in the life?

You chose the full-blown prospecting activity as the place for your agent.

Can you help us understand what got you to choosing that?

Yeah, so we picked the areas where we thought an agenting AI would be most impactful.

And this is in automating the reparative grant work, where as humans, we don’t like to do that also that much.

And the machine can actually process through data much faster than any human being can.

So if there’s a lot of data, for example, research is a good example.

You want to enrich your prospect’s data or your account data.

AI can do that much faster than you can.

And then you can essentially use that data to take some meaningful decisions.

Or you just had a meeting and the post meeting follow up is where the agent actually can automate things for you.

What are the action items?

What are the next steps?

What’s the email to be sent?

So that’s another use case.

So there’s the research use case, there’s the prospecting use case, there’s the meeting follow up use case.

There are a bunch of these different use cases and each of them, we call them as agents, essentially, autonomous agents.

Could be completely autonomous or supervised, as I explained.

That feeds off of the data that’s available within our platform.

So that’s how we’re looking at it.

I’m going to ask you what may be an impossible question to close us out.

And look into your crystal ball.

What you’ve described to me is an incredibly compelling application of agentic AI in one part of the business.

As I think about all the other functions of the business, I feel like there are great use cases and needs and problems to solve all over the business.

Do we get to a place where agentic AI is kind of like ants crawling all over a picnic?

You know, like, is this just going to be everywhere in the business?

Or do you suppose you get to a point where there’s sort of master agents that are running, you know, groups of lesser agents?

I’m trying to imagine what the future ends up looking like.

Yeah.

So with the advent of every new technology, we think it’s going to be it, and that’s going to be the future, but until the next one comes.

So that’s always been the case for the last 30 years that I’ve been in business and much longer than that.

Yeah.

So AI is here to stay.

It’s not a fad.

It’s going to definitely stay.

But it’s not a be all and end all.

There’s a lot to evolve in agentic AI.

And I remember the early days of web services, and we were thinking everything is going to be web service.

Early days of microservice-based architecture, which everybody is going to have microservice-based architecture and it isn’t that way.

In the end, you will have places where agentic AI is well suited, and you will have places where agentic AI is not that suited, not that well suited.

You will have to have a framework where agents are talking to each other, because nothing can operate in a silo.

It’s just like APIs and systems have to talk to each other.

That has to happen.

You will need a framework where you have workflow orchestration, and workflow orchestration is about connecting different systems together to get a process executed from end to end.

That never happens in one system for the entire company.

There is always a specialized system that you rely on, and those have to talk to each other.

And if the agents are actually getting the work done for some of these steps in the workflow, the agents have to talk to each other.

So everything that you talked about, I think, will happen in the future.

It’s just a matter of time, but that what we’re seeing now is the speed of innovation is so fast that what used to take 10 years is now taking 10 months.

So the speed is much faster.

That’s what’s different from the past.

Maybe that wasn’t such an impossible question because I think you did pretty darn good job answering it.

Thank you.

This has been a delightful conversation overall.

I’d like to just say a great big thank you to you for joining the show and sharing your expertise.

Yes.

It’s been a pleasure to be here with you, Pete and your team.

Thank you for having me here.

Thanks as always for listening and watching.

Don’t forget to give us a review on your podcast player of choice.

We’d really appreciate it if you would leave a review and or share this episode.

At the end of every episode, we’d like to ask one of our AI friends to weigh in on the topic at hand.

Hey, chat GBT, how are you enjoying my $29.99 a month?

Had to have the OG back on the episode for $75.

Thanks for being with us.

This episode, we’re doing an AI-themed Would You Rather.

Do you have a fun one you want to throw into the mix?

Can you make sure to answer it as well?

All right.

Here’s one.

Would you rather have an AI that can perfectly predict market trends, but is super sassy and questions all your decisions, or one that’s always polite but only gets it right half the time?

I’d go with the sassy know-it-all.

At least I’d get the right answers, even if I get roasted along the way.

Plus, a little sass makes work more interesting.

And now, you’re in the know.

Thanks as always for listening.

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