Welcome back to AI in action. I’m Eric Housh, joined as always by my buddy Zack Terry, best Director of AI today, rounding out our series on AI readiness with two essential pillars technology and data. So what does it actually mean to be ready on the tech side, how do you handle data in a way that sets you up for AI success?

And to help us unpack this super important guest today, we’ve got FSM Chief Technology Officer Chad Seales, the one and only Chad, welcome to the show. Hey Eric, but I have I’ve really enjoyed the last few episodes of AI in action. I’m just a little I’m a little bit giddy that I get to be here and man, I’m really excited about this conversation.

That’s awesome. And we’re going to dive right in. We’ve got a jam packed episode today. But before we get into AI readiness Zack, lots of stuff going on in the news, as always. What are we talking about this week? Yeah. Wouldn’t be another typical week without a bunch of news. So you got three topics to go through today.

The first is Microsoft. So we have been discussing Microsoft a whole lot lately, but they just rolled out two new in-house AI engines. One is called My Voice one, and this is a speech AI model that delivers speech in under a second, which is very impressive. And then the other one is called m AI one preview. And this runs on about 15,000 GPUs to help power smarter copilot responses.

So it’s a major play on Microsoft’s part for independence with AI assistants. There are some reports that it’s lower benchmarks and that they didn’t include some some information. So it’s kind of sitting in the middle in terms of performance. Right now. But definitely a good move for Microsoft to start beefing up those capabilities. Next we have anthropic a couple of things with anthropic in the news this week.

Won a lawsuit that they settled with a bunch of creators and authors. And now that that’s getting kind of turned over, it’s, you know, basically could save them a lot of money, but it also opens up the door for, well, what happens now with, with these authors, and is there an opportunity for anything to change there. So basically the books closed on that one.

And then also they came out with a major update. And if you’re a ChatGPT user, this may bring back some ghost, just because of what happened with ChatGPT sharing chats on Google and indexing that. But now by default, if you’re on a consumer model with Claude, you are going to automatically train that model. Now you can turn it off, but if you’re on that consumer model, so a pro plan or a max plan or a free plan by default, it’s now sharing the information that you’re using in the chat with its training data.

So just be aware of that. And that’s something that you may want to turn off. If you don’t want to train those models with the conversations you’re having. And then finally, I think we are seeing some caution signs flashing with AI. We’ve all kind of seen that MIT report saying 95% of gen AI projects are failing or are not successful.

Meta has paused hiring. OpenAI had a bit of a stumble with the GPT five rollout. Sam Altman himself is admitting that we may be in an AI bubble. He’s no longer emphasizing AGI is the primary goal, kind of downplaying that which he’s been very much a proponent of for quite a long time. So it’s interesting to see kind of where the winds of the market are going.

I think obviously there’s a ton of money that’s still flowing into this entire ecosystem, but this may be a sign that we’re moving a little bit from pure hype and euphoria into a little bit more of a realistic scenario. So I think that’s overall probably a pretty good thing. But we are definitely seeing some of some of these wins.

Cool. The market just a little bit. Yeah. We where we kind of got into an arms race there for just a little way out where everybody is trying to do out. I’ll do everybody and and everyone’s trying to one up each other with the news, and we’re kind of seeing that pendulum swing back to the middle and hey, what can this tool really do?

But anyway, this week, as always, big shifts and tools, user control, market confidence. It is not a boring place to be talking about AI. Not at all. And all the more reason if you’re a business owner, you need to be paying attention to what’s going on and be ready for what’s going cool. Let’s jump into the rest of the episode AI Readiness Chat.

Let’s just start really high level here. When we say technology readiness for AI, what does that mean for a business? Yeah, I want to I want to challenge you a little bit, Eric. I sincerely believe the season for readiness has has passed us. Ethan Malik in his most recent blog that came out yesterday, he has he has a quote that we’re entering into an era of mass intelligence where powerful AI.

This is the part I want to listen to, where powerful AI is becoming as accessible as Google search. And that’s just been true in in my life. So even in my little personal in my family yesterday, I was I have a hard time doing anything that’s, like real. And so I was trying to install a filter in my Hvac unit, and I couldn’t figure out which way the filter was supposed to go in.

There’s too many arrows, and it’s, I needed help. And so I took a picture of the Hvac unit, and I asked, ChatGPT is like, what do I do in ChatGPT responded with the picture I had provided it, and arrows pointing where I was supposed to put the filter in, in which direction it was supposed to go. It was just incredible.

It was absolutely incredible. It’s so helpful to me and so much so my wife, who’s not a technologist, she’s a speech pathologist. She’s been using ChatGPT to help her figure out exercises that she can do with her students. My nine year old son last night was using, some AI to help him. He’s creating images of the types of dogs, and he was showing me different pictures of, like, what the dog would look like sitting in his lap.

He’s trying to blackmail me with cuteness to, get a dog for the family. The reason I tell you this is because, like AI is ubiquitous. And so when Ethan Malik says that it’s becoming as accessible as Google search and I believe it, and so what that means for us as business leaders is that our teams, our employees, our customers, they’re using AI in their normal lives.

And if they’re using AI in their normal lives, they’re going to be bringing AI into the workspace. So we’ve got to ensure that we are assessing Fast Slow Motion. We love frameworks. The framework that we use here, as I am says, implement and multiply. So we’ve got to be assessing what technology what AI tools our team have access to and how are they using it.

Just like Zach just mentioned about the consumer version of Claude is going to use your chats with it to train its model. So if your team is using a Claude, the consumer version of Claude, anything that they put into that tool Claude can use to train it. So if if you’re putting in customer data, if you’re putting in a proprietary process about your business, right, if you’re anything about your business is now available to that model for it to be trained on, and who knows what it’s going to do with it.

So we’ve got to assess what technology is our team currently using, what technology did they have exact kind of accessible to them. And then the second question that I’m asking, I’m asking of myself and our leadership and of business leaders is I get a chance to talk to them, is what tools do you currently invest in that don’t have an AI return?

Right. We’ve got to start prioritizing investing in technology that has an AI future that’s increasing its AI footprint. And if I’m looking around my tech stack and I have a technology that isn’t isn’t screaming at me about the AI capabilities that it has, then I’m going to be reassessing that technology and finding its competitor that does. And in assessing your technology, isn’t it?

It’s not just about the the technology that you have available or the technology that your team’s actually using. Then it’s about governance. It’s about your policy. And this is really, really hard to handle right now because you’ve got to have a policy that can move at the same speed that AI is moving. So when we first put our AI policy out, we didn’t have to worry about multimodal AI, right?

Because no one on our team had access to a technology that could consume screenshots or videos. But now that’s commonplace. So our policy has to adapt at the same speed that AI is adapting and changing. So the very first sentence in our policy is these are guidelines, not rigid rules, because we know that a rigid policy, a policy full of rules, becomes obsolete the moment that we write it, because the AI is going to change.

So what we’ve been why what we’re trying to do is give our teams through our policy, a framework they can use to responsibly use AI. And that framework, it’s just to trust but verify. We’re going to trust our team to go use AI well. But then we’ve got to verify it as a company that it’s secure, that it’s compliant, that it’s safe.

Right. So our policy has clear exception. It has a clear exception process in it. If you’re a team member and you want to use a tool that you don’t currently have access to to get a job done, all you’ve got to do is raise your hand. You’ve got to tell your director or you’ve got to talk to me.

So our policy has hard written into it. If you want to do something that you don’t currently have available to you today, that’s not covered in the policy, raise your hand and let us know. And my hope is that that’s creating a partnership between us at a leadership level and the and the team member, so that we’re working together to assess and what technology do we need as a company to move us forward.

And the other aspect about the policies, really important is that I think it’s I think it’s really important to have clear guidelines, like clear guardrails for and what is what’s what’s a violation. So we’ve got a clear line. We will never upload raw customer data. So we’re not going to export our customers data out of their CRM and put it into an LLM.

It’s just too dangerous. Never do it right. Never share a screenshot that has PII in it. If you’ve recorded a call and it happens to have PII in it or confidential information, under no means can you upload that to an LLM. So we give really clear guidelines, really clear guardrails, so that our teams have a safe sandbox that they can innovate and, work in.

And when there’s exceptions to that, let them raise their hand. So here’s a summary. The season for readiness has passed us, and it’s time for us to be assessing how are our teams already using AI, and how could they be using it more effectively. And let’s ensure that we’re investing in AI that’s going to have a future, and that they’re doing it in a way that’s safe.

And I’ve got I’ve got some comments here, but I think the most important question I can ask you is, is the blackmail working? Are we going to see a new sales puppy? Yeah, probably. So, you one minute for so long, but yeah, I think it’s going to happen. It was just too realistic. So cute. I love it, I love it.

Well, I think kind of commenting on on the framework and I still come back to this MIT report, you know, talking about 95% of these corporate AI initiatives failing. And that’s definitely a big number and definitely a sign of maybe some of this hype dying down. But I think, you know, the important thing is to kind of read between the lines a little bit and look at that study and say, well, what’s the sample size?

Even statistically significant? I’m not sure that it actually was 80% or so of it. Projects that aren’t even AI related tend to fail anyway. And then also, if you kind of look into this data, something like two thirds of the businesses that choose to buy their AI technology, as opposed to building their own AI technology, tend to see success.

And so I think there’s a really big argument towards buying these tools as opposed to trying to build them yourself, because that that tends to see a lot higher adoption. It tends to just work better across the businesses. It’s easier to do right. There’s a whole lot of complex technology and expertise that you need to go and build your own models and integrate that into your systems.

But there are tools out there that are already built that allow you to hook into these tools and do that relatively simply. Right. And of course, we know that as Salesforce and HubSpot partners, you know, we see those tools and we work with those tools every day. But I think it’s important just to think through is this something I really need to be building myself, or should I go out and buy something?

Right. And then that also kind of goes into, you know, John’s been on the podcast. He talked about these skunkworks teams that we spin up and testing these new tools. So having these small teams that are able to go in and test these licenses on these new tools and make sure that they’re working correctly, but doing it in a safe environment.

So I think that’s also a really great way to approach which technology tools are going to bring value, because ultimately, you can only know if it’s going to bring value, if you test it and you experiment with it. Right. That’s how you’re able to take a list of, let’s say, ten different tools and cross seven of them out.

Now you get three that you know are going to bring a benefit. Know. And Zach, I’ll add on that the concept of build versus BI, I lean towards being a bit aggressive. And so I’ll give you an example. We built a tool that we knew was going to be replaced by, by a tool that we would be that we would be able to buy.

We knew it. It wasn’t quite the tool wasn’t quite ready out there yet. It wasn’t generally available. So we invested the time. We invest the energy, the money, the people, resources to go build the tool. And what we learned in that process, in that process of building a tool, was we didn’t have a lot of the data that we needed.

We didn’t have a lot of the context. Our users were quite primed, for to have that process in AI. So the building of the tool was a really healthy exercise for us, so that when the tech, when the tool, when the when the actual tool came along that we could buy that has all the features plus more and is going to continue to be maintained by an outside party.

So we don’t have to do it. Our team was ready for it. We had the data, we had the context ready, we had the adoption, we had the engagement of the team. And so I kind of lean towards where you can do a little bit of building, get your hands a little bit dirty, see where you can get a little bit of tension in your company and find where those walls are.

And then when the tooling becomes available to you, you’re that much more prepared to adopt it. Well, it sounds like the exercise of trying to build that yourself helped to illuminate some problems that you knew we would need to solve if we were going to actually get a benefit from investing in tools down the road. And you’re smart.

Well, guys, tech obviously that’s one side of the coin. Data is the other side chat. Where where do people get this wrong when they start thinking about their data frame? Yeah. Here’s the thing. We’ve been we’ve been doing one thing for a long time. And then that one thing is helping businesses grow and scale their businesses by adopting and leveraging CRM as well.

So we’ve had a lot of conversations over more than a decade about data, both data inside spreadsheets and file folders, and custom built CRMs inside of HubSpot and Salesforce. Right. Here’s the moral of the story. Businesses have messy data. The longer you’re collecting data, the more people more hands in the cookie jar. The more data that you have, the more likely it is that it’s going to be messy.

And what I love about AI is this is a great moment to reflect on what data we have and how can we address the cleanliness of our data. And so and so where we tell our clients to start same place that we started. It’s don’t try and get your arms around every piece of data that you have. Instead, identify what’s the job that you’re trying to do with AI and what context does that job need to know about your business and your prospects or customers?

Because when you can start narrowing down what that context is, then you can start narrowing down the data that is that applies to that context. So start small, don’t boil the ocean. Be strategic. Let me give you an example from the real world. So we had some we had some family coming over a few days ago. And my wife, I think that she had she must have ownership at Amazon because we constantly have like 4 or 5 boxes just everywhere.

And there were a whole bunch of Amazon boxes that were in our foyer and in our living room, and we had family coming over. They needed they needed to walk through the foyer. They need to sit in the living room. So I got to do something with these boxes, because I need this space. Well, what I could do is I could have spent a bunch of time opening the boxes, putting the things away, folding the boxes down, putting the boxes out and recycling.

But what I did do instead of that was I just picked the boxes up, and I put them in the hall closet because no one in the family is going to go sit in the hall closet. Okay. So the job to be done was I needed a foyer and I needed a living room. And that’s the concept that you’ve got to take to your data.

It’s what data actually matters in this moment. And that’s where you start. So start with the data that matters for the context that you need to get your job done with AI. And that’s second. And this is huge Eric. That lambs are intelligent enough now that your data doesn’t have to be perfect. So what used to be true about I was that you needed a lot of data in a very clean state to get good results.

What’s true now is that these lambs have been exposed to so much data that they can extrapolate out from a subset from a partial data set, just a little bit of your clean data. They can extrapolate out the intelligence and say, this is what our data is like. This is what our business is like. This is what our our clients are like.

This is what our prospects are like. So you just need a little bit of data to go a long way. So essentially you don’t need perfect data from the start. You just start by cleaning up enough of the data that’s important to get the job done. And then I can get that job done with data. That’s good enough.

That’s super powerful chat, because I bet a lot of people are stuck there, that a lot of people are stuck with man, Our data is a mess, and it’s got to be too much effort for us to really dig into it. So saying that, hey, you don’t have to boil the ocean, let’s let’s focus on what’s important. Let’s focus on a subset.

And that’s the starting point. Really want to drive that home for folks. And we’ve discussed this scenario with like a service side of the business. Let’s say that you’re trying to essentially integrate all the case data that you have because you’ve got a lot of resolutions stored in previous cases. And let’s say this is thousands and thousands and thousands of cases.

Right. And you want your reps to be able to easily go and retrieve previous resolutions so that they’re able to quickly resolve the next case. That’s similar to whatever previous resolution. Well, if you’re trying to use an IR model to do that, there’s kind of a couple of different approaches you can take. But the commonality is you’ve got to be able to tell the LM what a good resolution looks like, right?

There’s got to be some kind of like definition there because that’s different from business to business. And there’s all kinds of things that you could be addressing in a case. And so you can kind of take two approaches there. You can either go and clean that data. And by clean I mean go and find that example, the ideal examples of what a good resolution is.

So if you’ve got 300 cases for the same problem, chances are not all of them are going to be the model resolution that you want to use. Right? So you take that upon yourself to go in and find the model resolution and you flag those. And that’s what you’re going to use to help ground your eye on. Right.

Well, the other approach is to simply tell the AI in plain language what a good resolution looks like, and use that to inform the retrieval of that information. Right. And both approaches will probably lend similar results, but it’s two different ways of going about it. It’s taking responsibility over the data that you have and spending time to go and clean that up yourself, or it’s using AI to help assist in that process.

I think both are valid. But I think it’s is important to know that you don’t have to necessarily have everything perfectly in line before you can start getting value with AI. Yes, it’s important to have good clean data ultimately to to draw from so that you can get good results. But you can use these tools to help you actually get to that state that’s necessary in order to get those those, those good results in.

Absolutely. Let’s bring it home. You know, Chad, as Zach and I have had several conversations over the past few weeks, and it seems there’s there’s kind of this trade off between innovation and stability. So as a CTO of a company, what should tech leaders be focusing on now? How should we be thinking about this? Is relevant and this is a fundamental challenge for an established business.

I’ve been rereading The Innovator’s Dilemma, and there’s this theory in that book that you’re familiar with. Great companies fail because they do everything right. They listen to their best customers, they focus on improving their processes and their people, their services and their products. And that single focus is what blinds them to the disruptive innovation. And AI is the ultimate disruptive innovation.

And let’s say you are intentional enough as a business to take your blinders off and you begin working on how do you leverage the innovation of AI. There’s still such a big difference between being an established business and being a growth stage business, like a startup. So I have this I have a friend that’s talking to you last week, and he runs this pretty large accounting firm, and he’s living in this tension right now because he’s got to protect his core operation.

He’s got to protect his core customers and his core people. But he knows that at any point, a small startup could use AI to completely reinvent the process of accounting in that startup. It doesn’t have the business to protect, so they can go all in on the innovation in a way that it’s really difficult for him to. So if that’s the dilemma, what do you do?

And this is where this is where that skunkworks that John spoke to it, in his episode, Zach just mentioned that this is where that skunkworks approach is so critical because here’s the thing. Skunkworks teams, skunkworks projects, they don’t just appear. So I love, Lee Cockrell. He was the, he was the VP over of Disney Parks for a long time.

And he has this he has this idea that I’ve always loved. It’s that you need to have a geek on every team. And what I love about this is we as leaders, we need to be identifying who’s that geek, that internal AI advocate who’s, like, just naturally passionate, who’s like using this stuff in their life, who’s taken a picture of the Hvac to figure out how to put the filter on, is you identify who that person is, who’s that geek that loves this stuff, and you empower them in their work, on their teams and their projects in a safe way, as defined on your policy, right to go and experiment and to innovate.

And this is the AI in the game framework it’s to implement. So how do you take the AI and you actually bring it down into your business. How do you innovate with what’s available to you? How do you implement it into your daily life? And then it’s the M. So once we’ve once we’ve assessed our technology and once we have a skunkworks team that’s helped us identify how we can implement AI, how do we multiply that use out to the broader team so that we can get the value of that across the entire business?

And and I’ll tell you the last thing that I’ve been thinking about a ton and having conversations with other business leaders about is AI. It’s disruptive and in a lot of ways of our existing processes. But another way is that, first, a scaling business, the constraint of a of a scaling business is not always the capital. It’s it’s oftentimes how do we find the right people to help us scale this business in the way that we want to?

This is a conversation we used to have all the time about Cpq is because what Cpq allows you to do inside of a CRM is you can learn, you can, you can teach your system how to our most senior or most wise or most effective salespeople, how do they sell? How do they bundle? How do they price? How do they discount?

How do they communicate and documentation? How do they get the clothes right and you digitize that. You put it into a CPQ solution, and then you can bring on new people, train them up, give them the tooling and let them run the process from your senior person. Well, what’s wild about AI is that I can extend your capacity in the same way without increasing your headcount.

So your, your junior middle and your senior people, they can go further and they can do more right with the tooling because the Lims, these these AI solutions there is sophisticated they have the experience, they have the intelligence of your senior level people. And if you can give them your processes and your little bit of business intelligence and your context about your business and your data and your customers, you, your people can go even further.

So you’ve got to find a way to give your geeks right. Your people who love this stuff, give them the access to the tooling, pair them with people who know, and sometimes they’re the right person. Sometimes you got to pair them with the right person. Find the senior people who know how to do the process best. Pair those two together.

Let them innovate and experiment and implement AI solutions. And then once you get one that works, multiply it out to the business as fast as possible. Love that. So we’re almost out of time. For the for the audience, out there. The takeaway. And it’s too late to talk about readiness, right? AI’s here. Think about the tools.

Think about the right platform. Start small, get that skunk work in place. Get the policies to keep you safe. Give your people some flexibility on the data side. Don’t try to boil the ocean. It’s not about perfection. It’s about focusing, finding what matters most. Creating, auditing process, and letting AI fill in the gaps. Yeah, this has been awesome man.

Thank you so much for joining us. And thanks so much for the invitation. This is fun. Awesome. If you guys listening, if if you’re ready to get AI right. This comes from aligning your team to implementing solutions. That really move the needle. We’d love to meet you. We’d love to talk to you. Hit fastslowmotion.com/ai to connect with us and check out fastslowmotion.com/aiinaction.

For all the episodes we’ve mentioned today, designed to help you build the right AI foundation.

Yeah.