Artificial intelligence is moving beyond experimentation and becoming part of how businesses plan, communicate, build products, and serve customers. In this special edition of The Signal, the conversation explores how AI tools such as ChatGPT, Claude, and other models are changing everyday work, and what those changes mean for senior housing and senior living technology. Kyle, Jessie, and Lucas discuss using AI as an executive assistant, strategic thought partner, and workflow automation tool. They explore how connected systems can help employees find information, translate customer feedback into engineering tickets, streamline onboarding and reduce repetitive work. The conversation also examines the trade-offs between model capability and cost, the importance of protecting customer data, and the challenge of measuring AI's return on investment.

Key Topics

AI as an executive assistant

AI as a strategic thought partner

Walk-and-talk workflows

AI agents and connected tools

Senior housing data and AI

Workflow automation

Enterprise AI adoption

Model selection and costs

Measuring ROI

The future of work

‍

Meet the Hosts:

‍

Lucas McCurdy

Kyle Gardner

Jessie Lucci

‍

Produced by Grit and Gravel Marketing.

‍

Become a sponsor of Bridge the Gap.

‍

01:02 - 04:53

Lucas McCurdy

Bridge the Gap signal in three two. Welcome to Bridge the Gap podcast, the Senior Living podcast, and this is a special edition, The Signal with Kyle and Jesse. We're so happy to be back on this topic of AI and technology in the world and in senior living. And we've got some really cool discussions today, hitting some of the big hot topics in the news of what is changing around technology and AI and how it applies to senior housing.

‍

Kyle, Jessie, so good to see you this morning. Welcome back on the show. Good morning. Well, diving in you know, we all three have different stories of adopting AI and technology. And these LLMs, between ChatGPT, Claude, and the other options. And I'm going to kick it off by just saying I'm a little bit late to the party.

‍

I'm just starting the heavy adoption point in my technology journey where I'm talking to, I'm using ChatGPT right now, and I'm saying, hey, okay, I'm ready to go all in, get connected through email and calendar and finances, and I'm just starting that journey and learning how it can actually change the way that I function and start my day, getting organized on all of these different aspects of my business and personal life. Jessie, I know that I think you're using chat as well. Where are you on this kind of adoption journey? 

‍

Jessie Lucci

Yeah, so personally, primarily I'm using ChatGPT and Codex. I think about my using my agent in three different ways, almost like three different employees. For me, on one side, is my EA, my executive assistant, doing the grunt work, you know, drafting my emails, looking through things, scanning through things on the other side.

‍

I think it through like my chief of staff means, you know, thinking about, you know, for example, engineering prioritization, feature prioritization, or planning my offsite. But my favorite part of using AI is really just my jam session buddy, you know, like, I do a lot of walk and talk, which is my favorite thing when I have half-baked ideas or things I want to think about.

‍

Hey, we're interested in doing this. Ask me questions; let's discuss. And my favorite part of that is you can pick up the conversation wherever you go. You can continue to contribute to the learning and continue to shape that, no matter the product or idea. Was your jam session buddy, and it's definitely one of my favorite ways of using AI, so 

‍

Kyle Gardner

That's awesome. Do you put in like headphones and walking around the block, just talking back and forth with your agent. 

‍

Jessie Lucci

Yeah, I definitely do a lot of walking these days because of that. And also with like remote control of my laptop, you know, these days you can actually, you know, think about shipping code, you know, while you are walking around or thinking about, you know, hey, just I think the possibility is endless these days.

‍

And also, you know, I think a shape how I work too, because I have a lot more time back from not doing the work, you know, think about, you know, like prioritize. What's my, you know, chief of staff. Now, I can do the stuff that's more creative. And having someone, you know, like any time I want to be able to have conversation to think about, like, hey, what is next to build that doesn't exist yet?

‍

When you started with your chief of staff in your EA, how long was the prompt? I imagine it's changed over time, but how did you set up that initial conversation? Was it kind of a long form prompt where you had really specific goals, or was it just like a two sentence kick off to the conversation and it just built on it from there?

‍

04:53 - 10:13

Jessie Lucci

Yeah, I think I started obviously using AI like everybody, starting from, you know, copy-pasting my emails a long time ago. Right. I think for the longest time that's really the most common case. But over time, there's just a lot of contextual knowledge that the agent acquired through the record. And I would just say in the last few months, the skills are getting really, really more robust.

‍

And also think about, you know, I start to really realize, you know, when I think about using AI, it's almost like a skill of like using Excel. You know, if you ask the way, the right way, you can save a lot more time, give you an example. So if I think about, hey, I want to build this, I will actually have my agent ask me questions one at a time, thinking about multiple choices, putting the most recommended answer on top so I can guide my my agent to be able to have that conversation. I think that's really helpful. I think about, you know, like there are skills like grill me, grill me on this, right? I mean, just be able to poke holes, to be able to have quick questions and answers, be able to get to the source of truth faster. 

‍

Kyle Gardner

Lucas, what about you? When you started connecting things you mentioned offline, you know, you're just like, hey, I decided to pull the cord and jump in. Was yours like a single sentence prompt, or did you have kind of a big plan that you gave to your agent? 

‍

Lucas McCurdy

I did not have a big plan. I, I'm like, Jesse, I'm doing more of the kind of walking and talking, typing, where if I have an idea or something, you know, I follow a lot of these AI technology folks on Instagram.

‍

And so I'll see a short video, and it'll prompt me to say, okay, well, I need to do that. Let me start doing that. You know, and we've talked about even on this podcast, just learning out loud. And I'm in that early phase of learning out loud around these different prompts. And so I did that yesterday on the plane. I didn't talk into it. I typed into it, and I started the process of integrating my Google suite into ChatGPT. And I'm starting that process and learning out loud right here on the podcast 

‍

Kyle Gardner

I've been using Claude. I think I'm the odd one of us in that regard. You know, they released the co-working feature earlier this year in 2026.

‍

That was the big reason for my shift to that. I was previously using Gemini, and I've just found the connectivity between like it's called a harness, right. So for the non-technical folks, you have like the the AI is the large language model, and that's what you're talking to back and forth. And that's what's predicting text and consuming all these other things.

‍

And then different companies have built what's been coined a harness, which is think of it as arms and legs to the brain of the model, giving it tools and and capabilities like reading email or accessing an API, or connecting through an MCP to your SharePoint file or your Google Drive or something like that, so that I can consume more information and then talk to you about it.

‍

And coworkers. It's a wonderful application. It's it's a little expensive. The token costs from anthropic are pretty high. But I've truly enjoyed it. And over the last year, I suppose I've connected it to probably 20 or 30 apps. Whether NIC MAPs own MCP, my SharePoint sites, and I've got some personal apps in there where I store my own data and it's just able to consume that.

‍

And so if I have a question, you know, instead of calling a team member who might be the head of a department, because we share files here and access, I can just have Claude take a first pass at finding something for me. It saves my coworker time, literally my human coworker time because I'm not bothering and disrupting them.

‍

It helps me be more independent, but most of the time it's able to find things that would have taken me 10 or 15 minutes to go hunt for, and then it can take that follow up action all at the same time. So it's been a really big saver in that regard. I've I'm using it right now and finance kind of go to market planning and resource prep.

‍

And I would say task organization. I haven't really cracked the code on a chief of staff model. I have very different competing responsibilities at map between go to market and the finance world. And so I kind of have like a like coworker project for each one where I have my different thinking and prompts and rules inside of those projects.

‍

I won't go as far to call them agents because I still have to activate them manually for the most part. But figuring out how to do you know what work should be scheduled versus on demand is kind of the next step that I have in place. Lucas, what are you hoping to achieve? First, you know, if there's something you could give, you could get time back on.

‍

10:13 - 14:22

Kyle Gardner

Where are you going? To try and experiment with that? 

‍

Lucas McCurdy

Yeah, I think it's probably similar to most people in my shoes that are experimenting with this and just getting started, a time saver. And you know, there's a common thread here, even though we are all in different versions or across the spectrum of adoption of, of these technologies, is that we're all entrepreneurs.

‍

We all have, you know, businesses that are very entrepreneurial, that I think the industry really recognizes. I find senior housing to be very entrepreneurial. Yes, there's big companies, there's rights, there's private equity, and there's very complex business models in our industry. But there's a lot of smaller entrepreneurs in our industry. And I think that there is an interesting take on the AI models where it tends to level the playing field.

‍

And I can't help but think about NIC MAP and over the years where and I don't want to speak for you, Kyle, but I think of, you know, maybe ten years ago that it was the big players using NIC MAP and the data, but now you're able to a lot of the smaller operators are able to access NIC MAP in the data in their own ways, and they don't have to have 15 analysts in their in their organization. Then there's, you know, and their salaries and everything else. And so there's been a lot of change there. 

‍

Kyle Gardner

Yeah. We've tried to democratize the data and get access to you know, when I say democratize, I mean make it easier for anyone, regardless of their technical background, to utilize the data. You still got to pay for it. It's not a social good asset, but it is a lot simpler to use today than it was maybe ten years ago.

‍

And that's taken a lot of human investment, some AI investment as well. But we're going further. And I think most data businesses, not just NIC MAP-er, head in the direction of, you know, making it easy for AI to use our data within kind of proper channels, right. So I want my data to be only in enterprise AI systems where there's no data retention, there's a data retention policy, there's a no training policy so that my IPS respected the customer's IP is respected, but they're getting all the benefits of the LM at the same time.

‍

And so like on this show, you'll hear me reference MCC quite a bit for the for people who aren't familiar with that term instance for model context protocol, think of it as an API that was designed specifically for your large language model or your AI agent. They're almost identical and structure, you know, and that's one layman talking about it, but it's just meant to make it easier for that AI agent to make API calls and pull data in.

‍

And so we're running some tests now to see how large language models perform with and without map data on senior housing-specific tasks. And the early results are really promising for like having the case for specialized data. And that goes to really any discipline, whether it's scheduling and writing emails or versus creating marketing content. Having that attention to detail of speaking your customers' language, tailoring your message to an ICP is very different than drafting an email between two executives, where shorthand is used quite regularly. And so context is so important to all the tasks that we use in that using AI and relevant data helps really achieve that context. 

‍

Jessie Luccie

Someone recently said about AI is it's about micro delegate and macro steer. And I think on the micro side of things is he really evens the playing field, right? I mean, from our perspective, I, I did not came did not come from, you know, like, you know, the engineering background.

‍

14:22 - 18:20

Jessie Lucci

But thinking about these days like, hey, you know, if I see something that's, you know, it's imperfection in the app, I can just talk, you know, no matter it's curse or codex, it will ship immediately. And deployed on the other side is like, hey, when you think about macro steers, like, you know, look, as you're having those conversations, as you continue to perfect your learning, no matter product is vision is the go to market strategy. You continue to contribute to that vision. And having a, you know, someone to be able to, you know, obviously be able to bounce your brain off is definitely an interesting way of thinking through 

‍

Kyle Gardner

When you guys are having those talking conversations with the agents. I've only used that 

feature maybe once or twice. What have you found? Maybe overwhelms them in a conversation? Or is there any tips and tricks you'd give to someone who wanted to try this walk and talk version? Because I've I've not spent a lot of time doing it, and now I'm very curious. 

‍

Jessie Lucci

I think the technology makes it really easy these days, especially even just in the last few months. You can you can tell like even the way it talks, you start to simulate human being. Like let's say about six months ago, you start to hear them saying, that's interesting. Essentially, that's processing time, right? They're trying to do things through, you know, the answer, the question you just asked them. But these days it can, you know, like it can have a giggle. Like it's literally it just realistically simulate human being. And and it's a lot easier for you to be able to have a change of conversation through means.

‍

I go, I'm trying to ask you this, but it didn't understand correctly. You can quickly actually change that conversation, update it versus let's just say six months ago. It's like, oh gosh, you have to listen to this for the longest time. Then you have the other part of the conversation. 

‍

Lucas McCurdy

Well, that's a great transition into another talking point. I'm going to hit a couple of news headlines. So Anthropic September announcement claimed that sonnet 5.5 runs 30% faster and cost up to 30% less for most work. So in June, Google introduced Gemma for 12 B, which says it runs locally on 16GB of memory and supports vision and voice. So there's some headlines about AI becoming more capable, faster and cheaper and run locally.

‍

You know, cost is definitely a factor in all of this. And, you know, Kyle, you talked about enterprise. Currently I'm only using I'm using the pro version just personally. But I'm talking to my team today. Actually, I sent out a message about talking about running an enterprise system amongst coworkers you both own, or you both work and run technology companies. You know, I'd love to hear your thoughts on this topic specifically. Jessie, you own and run a tech company. What are your thoughts on this topic? 

‍

Jessie Lucci

Yeah for sure. So there's definitely a when we think about you have Gemini, you have Claude. You have ChatGPT. You have gosh, Muse. Muse, correct me at the end of – or Cursor.

‍

You know, we start to think about, you know, the the cost model. Be able to understand. Hey how can we extract most amount of efficiency out of it. Well, without burning, you know, or the, the expense side of things. We also think about, hey, where do we use AI specifically on every single staff. So we're talking about, hey, you know, for customer facing operations folks, they can quickly turn customer feedback into linear ticket.

‍

18:20 - 25:17

Jessie Lucci

And our engineer can quickly pass it on into code. I think for us like being beyond, you know, the efficiency side of things like, hey, mass onboarding or using browser control, we specifically think about where are the which is the gap between our workflow that can be automated just because you know, you don't want the engineer go to a slack channel to see, oh, there just might be improvement ideas.

‍

Then go to an operations person, say, tell me more contextual knowledge. Those are all can be streamlined relatively easily from the people who are at the front line. Understanding the workflow, understanding the feedback to shipping into the product. You know, I think we spend a little bit of time and actually need to spend more time to think about how do we streamline those so we can continue to optimize how we serve our customers?

‍

Kyle Gardner

Yeah, we're we're also using linear and we've integrated it into slack and linear new addition slack, something we've been on for a decade, it feels like or since whenever it came out, from our perspective, we've tried to take the existing process and do one of two things. Either remove steps that AI now completes so we don't have to have that human take the action, or we've integrated AI to help move the help the conveyor belt move faster.

‍

So even if the if the humans still in the loop by necessity, can AI, you know, get everything in a line or collect all the inputs. And so they're only having to do one review instead of ten reviews or something of that nature. And so when our team gets, you know, tickets or ideas or bug bugs, we put that in a slack channel, linear reads.

‍

It automatically creates a ticket in our DevOps environment. And then anytime someone adds comments to that thread and slack, it just moves over to the dev environment for the engineering team. So the client facing groups, they don't have to worry about how to navigate DevOps or these other tools, they just get to use slack, which is something the whole company is familiar with and fluent in.

‍

And then the engineers, they've got all their work in one place. To your point, they don't have to go back between different channels and and find things. The cost part is really important. And NIC, we rolled out coworker to the entire company about 55 people earlier this year. Most staff members got a pretty narrow budget to start, and it was it was encouraged to use this every day.

‍

You can get more budget when you when you can kind of show that you're getting value and productivity for it. But we put caps on it out of fear of runaway agents, which happened from time to time of someone gives it a big, big, massive task and just says, you know, do this at all costs. And then, you know, six hours later, the things still running up a bill.

‍

So we've we've learned a little bit in that regard. But I've found every level of the organization is using it in a different way, either. I think the things you guys brought up earlier, helping with emails, preparing meetings, being my EA, my virtual EA. That's where the starting point is for sure. But then I've seen departments build basically local applications for work unique to them, or automating reports that they used to pull data from seven systems.

‍

And now we set up the appropriate MCC. And so cloud is just doing that reporting without taking any human time at this point. And so it's giving time back or it's giving us data and information we didn't have before. And I think that's really incredible. I've talked to a fair number of businesses in the space about what models they're using and how they're benefiting from it.

‍

A lot of people are default like equipped with Copilot, and I'm going to use the word equipped, quote unquote. I think they would like something else, but the Microsoft system is the one that's approved at the moment. And I haven't heard as much exciting creativity come out of the copilot users. And that I've heard from people who are using Claude or ChatGPT or something of that nature.

‍

If someone's listening to this and you've you've done some fun things with the copilot ecosystem, I'd love to hear from you and just learn. But cost is cost is incredibly important. So anthropic is kind of known as at least this in my opinion. And like people I talk to, it's known as the most expensive model provider. But when you look at its most performant model, the fable series, that's like ten times more expensive than sonnet that Lucas was just talking about.

‍

And I think ChatGPT is top tier model is called sole, which is maybe 40% cheaper than fable on on recent publications. So cost is a big one. I'm I follow the podcast All In which is a big tech and business podcast. So if you're in the startup world, you've probably heard it. If you're not, you've probably haven't, which is totally fine.

‍

And a couple of weeks ago, they were talking about open source models and how there's been a big switch in volume of work being pushed to open source because it's much cheaper. Right? You only have to pay for the compute, the processing power. You're not paying a margin on top of that, which apparently tends to be a pretty healthy amount of the spend going into the models.

‍

Now, it's a lot more difficult to utilize open source. You've got to do some building for the most part in order to set that up, or you've got to have an existing harness where you can connect into that. I think for the developers like Jesse, it's probably easier for the non developer like me. I would be asking ChatGPT how to set this up and kind of go from there.

‍

It's just important. Hey, look at the work that's being done and pick the model to that. Or my CTO has pushed back on me because I was using fable for everything. He's like, fable doesn't need to read your emails, bro. You can use the sonnet model, you can use the haiku model, which are incredibly smart, but they're much, much cheaper.

‍

And so one thing I've been kind of playing around with is how to figure out which model is best for a group of tasks. I'm not changing the model every single time I type into it, but if I'm just doing a search through my files, I use haiku, the cheapest, fastest model they have. If I'm doing board level work, I'm using fable because I want the most intelligence to support me in that effort.

‍

25:17 - 30:56

Unknown

At our company, our engineer follow pretty closely on, you know, like different models and obviously the calls on different models. But also, you know, if you think zoom back out a little bit, think about ROI. What are everybody using the default usage of the the AI for? When you zoom back out a little bit, it's like, hey, you know what can we onboarding like a massive amount of communities with human being without AI in this efficient way?

‍

The answer is no, right? I think when you think in that lens in terms of, you know, like opportunity cost wise, I think it's pretty minimal in terms of using or not using AI in terms of decisions. Very easy. And or, or if you think about from an engineering standpoint, it's like, hey, these are the features previously we need to take weeks to build.

‍

These days is like a day. I think it's always good to keep an eye on models. You know, we have dedicated person focus on that. But you know, like in central essentially when you think about, hey, putting all that on the side, it just ROI stuffing boards that specially within the dedicated workflows. 

‍

Kyle Gardner

How are you measuring ROI? Jessie that's something I just was reading an article on this morning of the article was kind of positing that it's hard for most companies to get a massive ROI because he was he was measuring revenue productivity against AI spend.

‍

And I thought, you know, that's one lane to look at it. You can look at time back. You can look at like how many employees you need in order to achieve the same outcome. And this was really geared at kind of tech businesses for the most part. So maybe not as 1 to 1 for our operator customers, but have you guys set up kind of measurements or KPIs around that, or is it still in the experiment feel experiment phase? Excuse me, where you're kind of feeling out, what's working and trying to determine KPIs later? Yeah for sure. 

‍

Jessie Lucci

So for us we think about three different lens right. One is product rollout and product built that directly related to revenue revenue retention. And we are the company who built at least one more product every single year. So for that, it's like we're directly packing on more revenue, obviously serving our custom better.

‍

The second way we think about is like, hey, like time to revenue, right? So when you have an existing products like hey no matters onboarding time, no matter, it's like, hey, like customer want to customize all the different things. Time to revenue is something you know can be unblocked relatively easily, easily with more AI agents help. The last part is like, hey, scaling the headcount.

‍

You know, like headcount is the first two if you think about it on the revenue line item. Right? The the third one we think about is like on the expense line item, do we need to hire ten, 20, 30, 50 people to to scale as we need to know matters on the engineering side or on the customer side?

‍

I think that's just pretty straightforward black and white numbers for us, at least at the moment, to be able to think about it. Obviously, you know, like time saving is huge. Just because, you know, like I don't think, you know, it's sustainable. If we all work 24 over seven, it's, you know, it's it's a marathon, not a sprint. So definitely time back is important as well. 

‍

Kyle Gardner

Yeah I agree we're we're looking at kind of employee productivity time back, speed to X. Right. In my revenue teams we're looking at speed to lead and response times in engineering. They look at speed to tickets and how many items how many work items you can complete in a in a sprint, things of that nature.

‍

You know, there's other things to to measure too. But one thing the article kind of highlighted, I really liked when computers first came out back in, you know, you know, in the but from like 1950 to 1980, you know, productivity was maybe one and a half annual productivity growth across the US was like one and a half or something like that.

‍

And then once work started reorganizing around computers and really taking advantage of this new technology, productivity really shot up. And I think we still have work to the conversation we started with here, how do we integrate AI into our daily lives, into our work lives from the get go, not just as a temporary bridge or as a Band-Aid.

‍

Lucas McCurdy

Think about that, that point that you just made. It's like think, think back into like the 1990s and 2000. You have New York City high rises, filled floors filled with human beings at desktop computers, reconciling, checking accounts for these big companies. And now contrast that with with these technologies that seemingly a lot of that is just going to go away. I, you know, I'm kind of like looking out in the future, I guess, but it's a major contrast in the advance of technology. Kyle, Jessie, thank you so much for your time today. It's been a fun conversation. 

‍

Jessie & Kyle

Thank you. Great to be with you. 

‍

Lucas McCurdy

Okay. And to our listeners, you can go to download this content. So much more. We'd love to connect with you on LinkedIn. You can click in the show notes and connect with Kyle and Jesse and myself. We'd love to hear from you. And thanks for listening to another great episode of Bridge the Gap.

‍