Working with AI: Striking a Balance Between AI and Humans

How should you decide what to offload to AI and what you should do yourself?

And what are the risk and ramifications of becoming too reliant on AI?

Before you feed those questions into Chad GBT, do yourself a favor, stop and listen to this episode.

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 and Technology Officer, Mohan Rao, and NordLite CEO, Pete Buer.

We also have a discussion with Richard Lin of anyreach.ai about how AI agents are changing the customer service game.

But first, are you ready for an AI therapist?

The answer might surprise you.

Pete Buer joins us, as always, to write 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 good.

According to NPR, the artificial intelligence therapist can see you now, Pete.

What do you take away here?

So here’s the story.

Dartmouth researchers ran a study of 200 people with diagnosable conditions like depression or anxiety.

Half the group worked with AI therapy bots, and the other half did not.

The punchline is the half that did showed significant improvement.

We’ve had conversations on this podcast before about what becomes of different types of jobs with the widespread implementation of AI across industries.

They’ve always said that the rote and easily automated work is the stuff that is the low-hanging fruit and gets tackled first, but that the work requiring a human touch ends up being that which remains the domain of the humans.

Well, here’s a great example of breaking that compromise.

I think we even used back in that conversation, therapy as the example of a job that would be hard to replace.

But here AI is working its way up the food chain for maybe not the best use of terminology.

Handling therapy at a level where the indicators and interventions follow a recognizable pattern.

But in the process, carving off the bottom X percent of cases, whatever that number happens to be, probably triaged into action given an initial diagnosis.

Okay, this case is one that can be handled by the AI therapy bot.

But the result is humans are reserved for the trickier, less conforming therapy needs states.

And I don’t know, for our listeners in professional services, does this sound familiar in any way?

This is exactly the kind of progress that we’re trying to make with the use of AI, right?

Scaling what has historically been exclusively human-led, and therefore limited, delivery approaches.

This is really fascinating to me in a couple of different areas.

One, it’s interesting, do you think people open up more to AI more quickly than, say, another human?

That you are, you know, as someone who has benefited greatly from therapy and a huge proponent for everyone being in therapy, on that note, this makes it much more affordable.

But also, you know, there is a sense that you’re like, I want, I still want this person to like me.

It’s just innate in us as a human to kind of garner our responses in a way to the person we’re talking to.

And I wonder when it’s AI, and there is no judgment, there is no human on the other side, if you’re quicker to get to the root of what’s actually happening.

So interesting application and thinking of, maybe even with our customers, can we get somewhere quicker when they know, hey, this is just a robot, just give it to us straight.

Yeah, very fascinating story, Pete.

I think there will always be a behavioral and expectation overlay on solutions.

Some people still like to go talk to the teller, and others are hitting the ATM.

It’s not just about what work can be done by the machine, but also what work the consumer wants to have done by the machine.

But with time, we’ll get these segmentations right.

Very cool.

Love to continue to hear these and hear how the applications are changing and evolving.

Pete, thank you as always.

Thank you, Courtney.

How do you strike the right balance in your business between humans and technology?

This is honestly one of the biggest questions of this era for business leaders.

One answer is to make sure your team has the right tools and technology to help them solve real world business problems.

Knownwell is just such a platform for professional service leaders.

From CEOs to account managers and all points in between, the Knownwell platform can help you prioritize which clients need attention right now and give you the visibility you’ve always wanted, but have never been able to get.

Go to knownwell.com to see your data on the Knownwell platform today.

I sat down with David and Mohan recently to talk about how leaders can find the right balance between being powered by AI and being human-led.

David, Mohan, it’s been a second since we’ve been in the studio together.

It’s good to see you two.

It’s great to see you.

It’s better to be seen though.

I’m sorry, I just had to mix it up.

I’m tired of saying the same thing every time.

Yes, it’s good to be with you too.

David, today I wanted to talk about a LinkedIn post by Ethan Mullick that you recently shared and I’ve been really interested to unpack it with the two of you.

The post is about something that many parents of school-aged kids can probably relate to.

It’s kids using AI to do their homework, and the impact it’s having on learning.

Now, full disclosure, did I this week use AI to create a test for my daughter, grade a test for my daughter, do all the things?

I did.

So I’m not sure if I’m part of the problem.

How did that go for you, Courtney?

Was it seamless?

Actually, yes, because I’ve already talked to you about this.

So I used an existing, we have standardized testing here, like a TCAT practice test.

I used that practice test and I had Chajuk T grade that test.

One, it found that one of the answers, there was no possible correct one.

But then I asked it to create another test.

And unfortunately, some of the answers were very incorrect.

It did not work for me.

And maybe that’s a good story to go into exactly what we’re talking about.

Because I think this is an area where it’s hard.

It’s hard.

How do we do this well?

And I think it gets to that topic of how do we strike that balance between humans and AI.

So you don’t need humans in the loop.

That’s what you’re saying.

Just like you need moms in the loop.

Exactly.

Mom in the loop.

No doubt.

Yes, mom in the loop.

To me, that was actually the fascinating thing about Ethan’s post.

Because, you know, candidly, like you, you know, I’ve got kids in school.

You know, it’s funny, not only do I have kids in school, my oldest daughter actually married a principal’s son, right?

So there’s a principal now in the family that is a principal of a high school.

And for 12 years, I sat on the board of a university.

And so education has always been an interesting use case for me because I’m the de facto AI guy in all these circles that they go to and say, well, you know, isn’t this scary?

You know, how are we going to protect our kids from this?

And the argument I’ve always had and I’ve grabbed on is, we’ve got to teach kids how to use this and to use it right.

And do I know what that means?

No, not necessarily.

And am I an educator?

No, not necessarily.

But I think the research where it really struck me was, this is the heart of it, is that if you use the AI and empower kids to use it as a help, then it’s going to pay dividends, right?

It’s going to help them to learn how to use AI, how to engage with the technology, how to use it in their work.

If they use it to just replace the actual work, to do their work, then they fail.

And I think that not only speaks to kind of my passion of equipping them to use it well, and how do we educate them to, you know, assess the output of it and to deal with it and to shape it as part of their workflow is so important, but it also applies to business, right?

If we think of AI just as a replacement for workers, right?

And this is the track track that’s out there, right?

How many posts on LinkedIn do we see about, I’m scaling a hundred million dollar business with three employees, right?

And it’s all about how do I do it alone, right?

Yeah.

And how do I use it to lay off this many people that have money?

I think it’s totally backwards, right?

Like with every other innovation in the history of the world, it’s about how do exceptional practitioners use it to advance and take their own skills to the next level.

And that’s where human in the loop comes in.

And I think we should be empowering individuals, empowering the worker as opposed to replacing the worker.

Yeah, this analogy, I think, is right on, right?

So for students, AI should not be about finding the right answers, but it should be about learning to think critically to find the right answers.

That’s the distinction we’re finding.

That’s important from a student perspective or from a teaching perspective.

Similarly, on the business side, there is going to be some work which you can automate using AI and say, this is sort of road work, we know how to do this, we are pretty confident of it, AI go do this.

So there’ll be some class of work that can be totally delegated to AI as we go along and build confidence.

But business is not that simple.

If business were that deterministic, everybody would be successful equally, which means nobody is going to be successful.

So what you’re really going to inculcate with AI is the first principles thinking, and use the AI platform as a tool to enhance your first principles thinking.

That is the answer here.

Not that you get to the answer, just like every student wants to get to the answer, but it’s really learning to think critically, framing the right questions, figuring out the alternatives, debating between the alternatives, drawing a conclusion, and then test and reflect.

That entire cycle that you got to go through, that in our platform we call it the Udai loop, as you both know.

Just going through that and helping users think through critically through the life cycle is the best and highest use of AI.

There’s going to be a class of work where you can automate, get the work done, but then there is this operations and strategy work in which your humans will have to be in the loop.

Couldn’t agree more.

Couldn’t agree more.

I think that’s the secret we’re finding in so many of these use cases, right?

It is the professional that embraces it as a tool, not the business that uses it as a replacement over and over again.

And I thought it just came out so compelling in this research that you’re referencing, Courtney.

It really comes down to the student really in the research, right?

Do they use it to cheat?

Or do they use and replace themselves in doing the work?

Or do they use it to learn?

And if they use it as a tool to learn, exponentially increases their learning.

Don’t you think that paradigm, it’s almost like, I mean, I think this question is like really interesting, but I do, David you earlier said, turn the cube, and I kind of feel like this is a conversation we might need to turn the cube on as well, which is, what if we used it to learn about the student, and how the student learns, and how the student, like, we are deep into standardized testing right now, and we know there are tons of flaws with standardized testing.

Why do we even need standardized teaching?

Why don’t we deploy AI to tell us, oh, these are all the things the student has done this year, and this is what they know, you know, and here’s what they need to know.

Here’s how we, you know, figure out the areas that they are behind or need to improve.

And I think if that works in that application, does that also work in business, you know, to help us know, oh, here are the areas this employee extra training can be, or there’s a growth area for their career path, those kind of things.

Yeah, totally.

You know, we did an episode where we featured this company called Mercor, if you remember, where they were learning from every interaction and getting better and better and better.

That is a really good use case and Courtney, that’s similar to what you’re talking, right?

So it is the long memory versus the short memory.

Having these AI systems having long memory and learning from every interaction, so you can get better output each time, I think is going to be critical.

It’s the equivalent of AI learning more about the student or more about the business context, just getting smarter and smarter so the humans can get smarter and smarter along with it.

So that’s going to be the key.

Totally agree.

Yeah.

And then we don’t have to have mom in the loop anymore because…

We always need mom in the loop.

Always need mom in the loop.

It’s so true.

I mean, I’m sure many parents, you’ve got to advocate for your kid, you’ve got to look at all the stuff, you know you kind of got to fill in the gaps there.

But you know, I think the same thing applies.

As a father of eight, I can tell you, standardized testing…

He just might drop on us.

Standardized testing, standardized parenting, standardized punishment, standardized anything doesn’t work for kids, right?

I have deployed the exact same tactics and tools on two random of my children and gotten wildly different results, right?

And we are human beings, we’re unique.

And so this technology allows us to move away from the factory system of everybody has to sit in a desk and take this standard test and is measured by the same thing and move towards a world of your word, personalization, where we can personalize not just the content we’re throwing at them, but the learning experience, the engagement, and we can start to tap into individual people’s strengths and weaknesses and passions and help them become the best version of themselves instead of this rubber stamp of everybody else.

And I think that creates a more human world, not a less human world.

All right, you too.

I mean, this was, I think, a really interesting conversation.

I think a lot of pain points or areas, maybe wish list items we want for the future with AI.

But what do we do with this now?

What are some things that you think we should actually change as business leaders as we’re approaching our day-to-day work?

I think everybody should be using AI in their daily work.

So there was a recent memo from the Shopify CEO out there that said, don’t ask for more job racks or open roles and just see how you can use AI.

I think whether you’re the CEO of the company or a board member or a CMO or anybody else in the company, you should be using more AI just because you can leverage yourself better.

So that is something that is an imperative that everybody should be doing.

Beyond that, either it’s building or buying AI products is important as dictated by your use case and circumstances.

This all in the service of making your employees much better, more creative, and building a better business.

So I think that’s where you have to be, and obviously there’s a lot to unpack in those statements.

Mohan, I would add on to what you just said, and I think it’s, here’s the thing that I took away from this.

He even changed my frame on how I approach AI.

I love the lesson that it’s the students and the way they engage, do they leverage it to learn?

And what I heard in that was coach, was mentor, or do they leverage it just to get their work done?

Right, and to be more efficient.

So I think one of the ways people can use it in every day is actually going to it as a coach, as a mentor, going, you don’t know where to start.

Hey, AI, I want to use you more in order to make my life more efficient and more effective.

What should I be doing that I haven’t even thought of yet?

What are the next steps?

Asking it questions, engaging with it as that personal coach that we all wish that we had sometimes that was looking over our shoulders.

To me, that was a very practical takeaway that I took from this research that I think would be helpful for everybody.

Because so many people I talked to are just stuck.

They don’t know what to use it for.

I don’t know what I don’t know, so it’s hard to know to go there and start using this new tool.

I think that different paradigm, it’s not a worker, it’s a coach.

Yeah.

David, I think it might be interesting to also share a little bit of the challenge that you gave to our whole team this week.

Because there’s some people, it’s a very practical thing you can do right now.

Yeah, well, you know, it’s funny because this Shopify email goes out and of course it ripples through society and social media explodes.

I’m sitting on a few boards and I’ve actually seen it go through these companies that I sit on the board of too, where all of a sudden everybody’s on high alert, right?

And what’s interesting is, I think you can misinterpret the Shopify email to become, well, we’re not hiring anybody until we squeeze every last drop of blood out of the turnip and you’re using AI to its fullest potential, right?

That would be the counterpoint of, hey, we’re replacing people.

We’re not actually putting the Iron Man suit on.

So my challenge to our team this past week was to, let’s take that buzz, let’s respect that this is the new world that we live in, where we are going to be expected to all be 10xers that are so much more effective in our work and doing so much more.

And what is one thing we can do this week to actually take a new step to use AI in a new way and put it to work?

But here’s the thing that I wanted to see.

It’s not just doing something.

It’s how do we share it with one another and begin to build on each other’s ideas and take it to the next step?

Because I think sometimes where AI is getting stuck is in the silos of one person uses it this way and another person uses it this way.

But if we pull it together, we start to collaborate, different people have different ideas.

And I think we’re going to start to solve more operational challenges versus some of the more execution tactical challenges.

David, Mohan, thank you.

I think this is really helpful.

Richard Lin is the CEO and co-founder of Anyreach AI, a platform that enables companies to deploy human-like AI voice agents that can handle conversations at scale.

Richard, welcome.

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

Absolutely.

Thanks for having me.

So founder and CEO of Anyreach, whose focus, as I understand it, is helping businesses cut cost and scale by handling conversations with AI voice agents.

Can you tell us the story of how Anyreach came to be?

Yeah, absolutely.

So just a quick hi-lo for your audience.

Anyreach is an omni-channel, agentic platform where anybody can spin up a fully autonomous agent on phone calls, email, SMS, and text, and so forth.

And essentially, through that discovery, we evolved to be more of a horizontal platform building fully agentic capabilities across all channels.

So how would you describe your target market and the primary problems that you solve for customers?

We’re primarily focused in healthcare, education, enterprise SaaS.

And the use cases are what you generally would think about of high operational costs with humans, whether it’s customer service or sales.

So just to get more specific, like lead qualification, for instance, for agencies that are trying to get leads for the real estate industry, or customer service requests for basic FAQs to answering calls, scheduling appointments and things like that.

So an example of this is like one of our education customers, their three school charter network in LA, where their KPI is essentially enrollments.

So the more butts in seats, if you will, the Fed and state pays them about $15,000 to $20,000 a year.

And so they actually spend a lot on paid media.

So Facebook ads and Google ads, they drive awareness and interest from parents to their website.

So we’ve actually taken over their paid media to do click to call campaigns.

One of the things I noticed on your website, you referenced the notion of cloning your top 1% agents, I guess was the reference in the moment.

But I feel like I understand that intuitively, but can you give us a little bit of explanation and including sort of the mechanics behind it?

Yeah, absolutely.

You don’t want an agent that just speaks, you actually want to do stuff for you, right?

So what we do differently with that 1%, if you will, is we have a zero shot prompting process where we actually clone the best rep or the best person in your company that represents a specific organization or task, if you will.

And really the goal here is, you either upload call recordings of those top 1% successful outcomes, maybe they close the deal, qualify, whatever you want to call it.

And if you don’t have those call recordings, you do a role playing call.

So an AI agent actually calls you and you as the human who’s setting up this agent will just talk to the business, company, all kinds of stuff.

We essentially use deep learning to clone your voice, your personality, your speech nuances, and so on and so forth, so that when the agent speaks, it sounds exactly like your best rep within your company.

And the benefit of that is not only more accuracy, less time consuming prompt engineering, if you will, but it also keeps the diversity of businesses.

So top 1% isn’t performance necessarily exclusively, it’s also representativeness.

Exactly.

Yeah.

That sounds like an inherently, intrinsically smart and cool idea.

I could imagine though, if I was an agent that I might not feel the same way, how do you handle the change management?

And how do your customers handle the change management?

We generally try to position it, and it’s not just the position, it’s actually what we believe in that humans are still absolutely going to be required as part of the process.

And the reason why is because AI is not at the place right now where it can be fully autonomous.

And the example I could use here is with autonomous driving.

Waymo has been on the road for about 10 years, and Waymo didn’t get to where it is today on day one.

And the reason why is because you actually had to label a bunch of driving data, if you will, from videos to images.

And they essentially outsourced that to a company called Scale AI to look at, hey, is this a dog?

Is this a truck?

Is this a stop sign?

And they’re still actually labeling now.

So there’s still a lot of work there.

And I think a lot of the work that folks think AI is going to replace is essentially going to evolve into how do you train AI to be higher quality, higher accuracy and so forth.

And more specifically around voice, that work actually hasn’t been done yet.

And the reason why is because only about 1% of the data in the entire world has actually been indexed and trained, and that’s the Internet and all the foundational models have pretty much scraped all of it.

But the other 99% is private data.

It’s personal data like photos you have on your iPhone, to private enterprise data.

And you can imagine all the phone calls that, the Verizon’s of the world, the Comcast and so forth, none of this is public, right?

It’s all kind of housed privately in their own databases and so forth.

But these companies actually need to expose that information, right?

And it’s probably gonna be obviously with a company like ours.

And then that data needs to be labeled.

People actually need to go in, real humans need to go into a call and label all kinds of things from like, you know, what’s the customer sentiment here?

How do they actually resolve it?

If they’re taking action, what needs to happen?

And so I think, you know, as we think about the changes of roles and so forth, that’s what’s going to end up happening.

Like a lot of the customer service agents are going to evolve into AI trainers, who are partially doing the job when AI fails, but also at the same time, helping with the pre-training process as well.

Interesting.

If I’m a head of HR listening to the conversation, and I’m thinking about the math, the workforce planning math of re-skilling agents to be AI trainers, do they have the core skill set?

Is this a bridge too far, or is this a believable path that companies can pursue?

Oh, 100 percent.

I mean, I think the way I think about AI trainers is they’re just doing the job normally, right?

I see.

And then when they resolve a call, and I’ll give you an example for instance, voice agents right now, they kind of have trouble with the emails, like capturing emails.

And the reason why is because emails are pretty obscured, if you will.

It’s not like structured information like our names, right?

It’s usually some randomized alphanumeric type of ordeal where it’s like 1, 2, 3, taskmaster24 at gmail.com.

And then the phonetics also plays into the part of the email, right?

So if your email is vb at gmail.com, even humans will mess that up and say, is that bb at gmail.com?

And an AI will say, hey, the email will say, hey, no, it’s v for Victor, v for Ben.

And an AI will take that without the context and knowing why that applied and say, hey, Victor, I’m not Ben, I’m Amanda.

Nice to meet you, right?

So a normal human would go through that process and understand that v for Victor, v for Ben, go map that correctly.

They just type that in and then it gets thrown back into the model.

So it’s essentially just doing your day job, but you’re also training while you’re doing that.

So that’s it.

And if I’m an AI trainer, is that something I’m doing for an interim period until we reach a new moment of migration and how work gets done?

Am I in the long term, am I still destined to find a different kind of job?

Yeah, I would say so.

I think, I mean, I don’t know what the future of AGI is going to look like or when that’s going to arrive.

But if the vision is clear that eventually AI will be completely autonomous, can think for itself and can take action across different kind of spectrums, if you will.

When that happens, actually, everybody’s job is on the line.

It’s not just to come to service agents or sales.

I think the entire way we think about work and potentially capital and wealth is going to completely shift.

It’s a whole new type of economy.

What that ultimately looks like, I don’t know.

But we’re all in trouble when that happens.

But I would say life probably gets easier.

We probably have more time to do more creative, fun things than just doing monotonous jobs.

And this is the job of leadership, of both doing the best we can to figure out what the future is looking like and to steward our talent and organization from here to there.

So I hope everyone’s up for it.

Richard, thank you so much for being here.

It’s been a pleasure.

Absolutely.

Thanks for having me.

Thanks as always for listening and watching.

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

And listen, we’d really appreciate it if you would take the time to share this episode with someone that you think might enjoy it.

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

So, hey Claude, it’s your turn.

This episode, we’re talking about how to strike the right balance between AI and humans.

So, what do you think?

I think finding the right AI-human balance means leveraging AI for repetitive or computational tasks, while preserving human judgment for ethical decisions and creative thinking.

The sweet spot probably varies by context, but the best outcomes likely come when AI amplifies human capabilities rather than replacing them entirely.

And now, you’re in the know.

Thanks as always for listening.

We’ll see you next week with more AI applications, discussions and experts.