Sources:

•https://www.publicissapient.com/work/marriott-generative-ai-search

•https://homes-and-villas.marriott.com/en//search

•https://corporate.walmart.com/news/2024/10/09/walmart-reveals-plan-for-scaling-artificial-intelligence-generative-ai-augmented-reality-and-immersive-commerce-experiences

•https://a16z.com/state-of-consumer-ai-2025-product-hits-misses-and-whats-next/

•https://www.perplexity.ai/hub/blog/shop-like-a-pro

•https://www.amazon.com/Rufus/

Participants

Hosted by Isaac Sacolick, CEO of StarCIO

Special Guests

• Roman Dumiak

Digital Trailblazers

• Joanne Friedman
• Liz Martinez
• Joseph Puglisi
• Martin Davis
• John Patrick Luethe
• Heather May
• Derrick Butts

Summary

The content discusses several sources and a host, Isaac Sacolick, during an event featuring experts in digital innovation. Key topics include Marriott’s use of generative AI for search, Walmart’s plans for AI integration in commerce, and insights into consumer AI trends and future projections, indicating a growing interest in immersive technology.

Transcript

[00:00:00] Speaker A: Greetings, everyone. Welcome to this week’s Coffee with Digital Trailblazers, our 158th episode of speaking to digital transformation leaders around leadership, technology, AI practices, mindset, everything that goes into how we evolve our organizations. And today’s special topic. I’ve wanted to cover this for quite some time.

We’re going to be talking about AI first, user experience experiences, planning for the evolution of gender, customer journeys. And we’ve tried to cover this a little bit in different areas and different topics at the coffee hour.

This one is very specific to this question mark of are we ever going to get to the point where we can talk about.

Oh, boy.

We can talk about AI above and beyond.

[00:01:03] Speaker B: Our.

[00:01:04] Speaker A: Above and beyond our. Hold on a second.

What did I do here?

Let’s see.

No, it’s gone above and beyond our ability to do.

Above and beyond ability to do workshops. I’m sorry. To do efficiencies and to be able to run our enterprise and everything that’s behind the paywall and start getting out to areas where we have experiences that are customer facing, that we’re impacting our journey maps, that we’re really going out into areas like retail, areas like healthcare, areas in insurance, places where we can actually impact how our customers are using our tools and using our capabilities in revolutionary ways.

I had a slide here. I’m going to try to bring it up right now with some details I thought was going to show up in the whiteboard. And I’ve gotten into a little bit of a technical issue. So I’m going to try to do this right now in real time so that we can have a real conversation around this. Give me one second, folks. Roman, why don’t you introduce yourself and I will get the slide up in the meantime.

[00:02:28] Speaker C: Sure. My, my name is Roman Dumiak and I am adjunct faculty at DePaul University in their school of Computing. I’m also a. What’s called an executive in residence for their innovation development lab at DePaul.

[00:02:46] Speaker A: Thank you. Okay, here we go.

Can you see this, Roman?

[00:02:51] Speaker C: Barely.

Okay, much better now.

[00:02:54] Speaker A: Much better. Okay.

I spent a big chunk of my time yesterday looking for examples of AI and real customer experiences. And sadly enough, I didn’t really find that many of them.

One really good example, I found some work done by Publicis Sapient.

They did some work with Marriott Homes and Villas. This is your ability to go to Marri website and rent a home or a villa somewhere. And they are trying to solve the problem of how do I find somewhere to go when I don’t know where I want to go or I don’t know when I want to go. I know the type of experience I’m looking for. I know maybe a little bit of who I’m trying to travel with. I know some of my constraints. And so they built a large language model to be able to do that kind of search. You know, I’m looking for a beach vacation somewhere quiet.

I want a place that’s less than a mile off the beach. I want good food and all kinds of like criteria like that about the experience.

And it comes back with a whole bunch of suggested places that you can go rent. I’ve left you here a link that you can go see this in.

In the State of Consumer AI 2285, this has come from Andreessen and Horowitz. There’s a lot of good content in there, but they talked about places where you’re starting to see AI creep into retail.

Perplexity has Shop Like a Pro, which I’ve tried to use. I would say I would give it a C in terms of its capability.

It really wasn’t doing very much and I tried Amazon Rufus a few times and I’d actually give it a D. It just did not warrant the kind of response that I would probably use it as a first replacement for their browsing and search capabilities. And so I think we still have a lot of work to do in terms terms of making our AIs ready for the B2C experience. There’s certainly plenty of examples of happening of AI happening inside enterprise software. I did a blog post around 50 different AI agents that you can get on platforms like Salesforce, Workday, SAP Appian, lots of different places where you can use agents to improve workflow. I have a hypothesis that I’m going to be running by my group today that with every major disruptive technology that’s come out over the last two decades, in order for IT to scale, for organizations to build capabilities with, we had to really rethink our operating model to be able to do this. You’re seeing that on the web one on the right hand side, what triggered transformation and Web 1.0? This is going back to the late 90s.

We were still doing client server software development. We were still doing waterfall project management. We had to bring Agile and web development into our organizations. When cloud computing came out, we could order infrastructure, but we really didn’t get the value from the cloud until we invented DevOps. We started putting automation in place like pipelines and infrastructure as code. We challenged our assumptions around the IT’S culture around development and operations. We brought those worlds together even when mobile came out. Our first mobile experiences basically took our web ones, ported them onto mobile screens and it took a lot of effort around design thinking and the introduction of app exchanges before we started really seeing triggered transformations impacting customer experiences. I have some hypothesis what this is going to look like for generative AI, AI agents and agentic AI around what we need for the customer experiences to really start becoming part of our capabilities. And that’s our discussion today. I want to hear from our experts. We’re going to start with Roman, our special guest today. We’ll go around the horn to all of our normal speakers. I’ve got Derek, Joanne, John, Joe and Liz here today.

We’re going to start with this notion. We have 1.0. It required teams to think about web development and agile. We had to bring DevOps to make cloud experiences work.

We had to bring design thinking to make mobile first experiences.

Roman, what are some of the new disciplines or mindset shifts that digital marketing and IT leaders must adopt for really AI to take a foothold in the customer experience and the customer journeys? Welcome Roman.

[00:07:37] Speaker C: Oh, thanks for having me. You know what I wanted to do first though was maybe piggyback on your comment. I think getting to an AI AI first user experience is really going to require people who understand how to do that. It happens to be that colleges survey their recent graduates and from a career standpoint we know that people in technical fields have really been feeling it for the last year plus. But interestingly, people who are involved in human computer interaction or UX UI design seem to be doing okay.

Now we don’t necessarily from those surveys know exactly why that is, but the hypothesis is they’re doing well because in their career path they need both soft skills and critical thinking skills all the way from low level visual front end design into more advanced things like content specialist or UX architect is building a front end framework or even UX research.

Then the second thing is that the need for designers as product based companies start to implement AI in their tools is growing. We all know pretty much every software tool now says AI first or AI embedded or something like that.

But interestingly, robotics is now getting away or not away from, but enhancing beyond just automation and replacing tasks to having humans and robots working together. And that requires some, you know, expertise in how you’re going to make that happen.

And then last thirdly, we think that the skills that design people have are transferable to other roles, particularly in areas like product management and you mentioned marketing, for example. So, you know, currently the modality is really kind of text based chat, but people want to use other modalities, voice and gesture recognition and things like that. And there’s also a push to replace a lot of manual input and integration between systems with gen AI tools.

So that’s kind of what we see from our recent graduates, I guess is this whole field of human computer interaction seems to be fairly stable. I’m not saying it’s great, I’m just saying, you know, those are the people that seem to be doing well in this economy.

[00:10:10] Speaker A: Thank you for that perspective, Roman. I want to say hello to a bunch of people have said where they’re from on the Commons train. Hello to London, New York City, Maryland, Boston. What else we got here? Manchester, United Kingdom, New York, New Jersey, Pune. Awesome. We’ve got a real world crowd here. Joanne and Roman is talking about skill change, human computer interaction as a skill set.

Maybe we need voice as a modality before we can bring it to customer experience.

What do you think the gap is? Why haven’t we seen enough AI first user experiences just yet?

[00:10:53] Speaker B: Well, I think there’s two things. One is that, you know, in the past we were limited by form factor, right? Think about the fact that we went from PCs to laptops, from laptops to smartphones and others. And so a lot of the UX and UI were designed for the form factor. Now we have a human machine interface called AI that has various forms. One part of that is the generative AI, you know, type in your prompt or speak to it, use text to speech or speech to text, and you can kind of start from there. But in actuality, when you think about it, a user experience has to be tailored to the Persona, the role of the human being. There’s sentiment, there’s context, there’s perspective, there’s security in terms of their role and what they should be allowed and should not be allowed to see. And all of that leads you down a path.

Agentic AI is very well suited for this.

And I can tell you firsthand because ours is autonomously generated, it’s based on the factors that we put in to training the models to allow the user experience to be designed for that user’s role. And in the context that’s most relevant to them, whether that’s also in a modality of speech to text or text to speech in whatever language they want, etc, all of those things have to be kind of built into systems. So what we’re seeing from the agentic side, excuse me, is That a lot of these are headless systems and allow you to begin to design the UX to suit the user. Not a one size fits all that we’ve lived with for, you know, the last 80 years.

We don’t have to do that anymore. We can customize, we can micro personalize. There’s a lot of different factors that come in, but really it’s about governance and it’s about building trust. So I think you’re going to see a lot more of that kind of come to light over the next year. For a B2C environment, it’s not that much more difficult than within the constraints of a single organization.

It’s knowing what your user really wants by determining deterministic variable.

Interesting. Sounds convoluted, but it actually is. The fact.

[00:13:32] Speaker A: Well you know, here’s what I’m translating this to Joanne, is that, you know, we’ve just upped the skill level of different areas and the collaboration of different areas that are needed to make human machine interfaces work and we’re just not there yet. Maybe in software companies are there, you know, robotics in integrating robotics with AI that’s coming, but in terms of B2C we’re not. The companies that need to do this just don’t have the skill sets yet and the collaboration to make that happen, is that a fair way of interpreting it?

[00:14:11] Speaker B: Yes, but also I think that their focus has been on gen AI and not other forms of AI.

There are real differences and this is kind of where the rubber hits the road between agentic AI and generative AI because one is, you know, an ocean and the other is a bucket. So if we start looking at how you really define the customer journey and the process that you go through mirroring that into Agentix is relatively, I don’t want to say simple, but it’s not about their skill set, it’s about their approach and the mindset.

[00:14:47] Speaker A: So basically the technology is changing really fast and our companies are stuck in the mud is what I’m translating that to. And understandably, I mean these are big investments when you talk about user experience, customer facing capabilities. I just don’t expect even a company like Amazon to take Rufus and say, you know, what we’re doing, you know, going to wipe out 10, 15 years of how Amazon worked and just going to replace it with an AI and see it work. Let’s bring Derek in and Joe. Derek, your thoughts?

What do organizations need to start really thinking about AI first ux?

[00:15:27] Speaker D: Well, it’s. Some of the things Joanne mentioned are spot on. I mean the mindset definitely has to change moving across and you look at, you know, discipline across the leaders, they need to adopt this instead of going from build to ship. They need to look sense to understand, govern, to use and adapt, to fully utilize.

They need to better understand how this is going to work. As Joy mentioned, these the developers of artificial intelligence tools, they’re really trying to figure out what the end user going to want and users are trying to figure out what the developer is going to make my life easier. So when they look at this, I’m looking at it now from a resilience or risk point of view. How do I get this to be a thought process for proactive versus reactive. It needs to be a forethought as opposed to an afterthought. And the governance piece is going to be huge. The governance piece should not stifle innovation but it should adopt and move innovation forward. And some of the things looking at the tools that are out there just on the Gentex and the AI agents and stuff is really looking at the threat landscape and understanding. Although a lot of these manufacturers are embedding new tools to allow the ease of use and make it easier for the end user, that’s nice but you still have to look at how you’re protecting your ecosystem and your business culture. When they’re looking at the marketing and these other directives within their and I think looking at a holistic approach and designing the simplicity and the complexity of what you need to do up front is going to be key. But again it’s still an evolving process.

One of the things I’m looking at is the threat informed design for AI experiences is make it easier so that end user can now be trusting the systems that they’re using to have to worry about the risk associated with those. And that really goes back to the companies and businesses that are using these tools. Make that investment into tools like a Mitre Atlas tool actually looks to better understand what’s taking place in your ecosystem but actually look at those threats on the back end. So you end users not always trying to look over their shoulders to see what’s happening.

It’s just, it’s an evolving process and like I said, it’s going to take time to really get into it. But I think by taking these steps and working towards these systems to help an adopt, they can start developing trust from marketing and a design point of view so they can actually have more free flow of the utilization of these tools that are out there.

[00:17:37] Speaker A: You know Derek Kevin Wallace EID on the Common Channel. He was our special guest last week. He also echoes the need for governance and trust.

And you know, I’m going to translate that into two things that are difficult when you start thinking about customer experiences. You know, first is the impact of on brand. If you get it wrong, right? And we’ve seen a few companies try to get stuff out into production in B2C avenues and royally screwed up because once you put it out there, you know, it’s, you can’t take it back. You can’t. Well, you can take it back but you know, someone finds a hole and all of a sudden getting, you know, coupons or getting free things because they know how to game the AI to do that, you’re in real trouble. So there’s a brand governance issue here that marketers really have to be cognizant of and part of this and I’ll, you know, see if John and Joe have comments on this.

I don’t think our testing methodologies are strong enough yet. I don’t think we know how to test lms.

I don’t think we know how to test agents in a comprehensive way to say, you know what, this is ready for prime time. And I think that’s what’s slowing things down. And I think the third thing is this stuff is expensive.

It is, you know, it’s expensive to skillset, it’s expensive to build your models up. So the development process is expensive. And then, you know why, you know, why is Alexa from Amazon only just getting LM capabilities? Because this stuff is power hungry. It’s going to cost them a lot of money at scale to run an LLM behind something like an Alexa with voice, with comprehension and make sure they can get a return on that investment. So, Joe, a bunch of different reasons you want to echo one or do you have a new one you want to share with us?

[00:19:29] Speaker E: Well, I think 30 years on. Isaac, to your point about how well do we test things.

[00:19:37] Speaker A: For those of.

[00:19:37] Speaker E: Us who have been caught in I.V.S. hell, we still haven’t figured out how to think like people and not like a decision tree. I, I think Joanne made a very cogent point when she talks about mindset.

I agree with my good friend Kevin Wallace.

It’s about the mindset. We, we have to think about these things differently.

Roman said human computer interaction. Folks are doing well. Well, you know, they think about how people ask questions and how people talk to each other and, and they’re not like programmers, they’re like, you know, people.

So until we come at this with with a design for people. I think what Joanne said was very apt to, you know, we’re not designing for a cell phone, we’re not designing for a laptop. We’re designing for interaction with people. And that’s the kind of mindset and change in approach that has to be embedded in our staff.

One other thing I want to point to with respect to Joanne and Agentic versus Genai. Earlier this week I was in a discussion with some folks talking about training or hiring new managers. We’ve talked about how you need experts to govern the AI and to understand, for example, code generated. How good is it when we’ve lost the entry level positions? Well, the point was made that, yeah, we may not be hiring entry level people anymore, but now what we’re hiring are managers of AI, managers of agents who have to know how or learn how to manage the technology as if.

[00:21:16] Speaker A: It were their staff.

[00:21:18] Speaker E: And that’s another new skill set that I just don’t think we’ve mastered yet.

[00:21:25] Speaker A: We’re going to go to skill sets in just a second, so thanks for that. But I want to hear from John.

John, yeah, good to hear from you. You know, let’s put you in the hot seat, right? Your, your chief Digital officer and your CEO says six months. I want my UI to be, you know, agentic, whatever that means. You know, what are your, some of your concerns that are going to be at the forefront before you start transitioning to an AI first user experience?

[00:21:55] Speaker F: Well, I think we’re still figuring it out how to interact with AI. People have made the comments that we’ve got, the chat bots, the text, we got that part of it pretty figured out.

But anything that requires a physical device, devices really take time to build. And I even think back to a long time ago in Silicon Valley when somebody was talking about their startup and whenever they would mention a physical device, like everyone had shrugged on that one. And working at a company that used to build embedded devices a couple of companies ago, it’s just embedded devices, they take time.

And so figuring out how to stitch AI into these things is going to take time. Some ones that have been really, really successful I think of like the Nest thermostat. That was a thermostat that was extremely simple for people to use, but it had some pretty, pretty neat algorithms in the background to try to figure out how to heat the house and have it be the right temperature for people. But they made something that was so simple to install, so simple to use that it’s like it was widely adopted even Though it was a pretty advanced technology.

I’m in the healthcare space and we’re starting to see some really neat things where we’re deploying these, these, these devices, Sensei AI is one of them and they have their own cellular network so they’re really easy to install. You just drop them in people’s houses and they, they’re able to look for degradation in people’s health or when people fall and things like that. And so they’ve made it so the install is so easy and it uses AI to basically figure out what’s going on in the house and then it sends a report to the people who are monitoring it and basically it tells how people are declining over time. And so this is using a ton of AI, but they’ve figured out how to make it really easy to deploy and really easy for all the parties, the, the couple, the families that have these things that they’re typically in elderly people’s houses. But they made it so that there’s a human that calls them or talk to them. And so they’ve made that interface super easy. And those are being pretty successful. And then, you know, I’ve seen some pretty good situations where people have stitched AI chatbots into the web pages themselves and so that if you ever have a problem on the web page, there’s a chatbot’s actually built in and it can see everything that the production team can see on the back end.

And so it’s like that’s how I’m starting to see people stitching AI into interfaces.

That’s like the baby steps of it. But those are the things that work. It just, it has to be really simple, it has to be frictionless and it has to make people’s lives better.

[00:24:27] Speaker A: I think that’s why even on the enterprise software side you talk about, you know, agents taking over or partnering with specific roles. You know, so a hiring manager or a finance financial analyst or a content manager in marketing and what you’re really seeing is more task based agents out right now. So they’re focused on helping that role do one thing really well and then moving on to another agent doing another thing really well. And this whole notion of orchestration, you know, I’m going to wait for Joanne to comment on this.

I still think that’s a work in progress and in terms of what the technology is capable of.

But let’s shift gears. We’re going to talk about, you know, how to drive the innovation. It’s going to be a part of our last conversation. I want to Go back to our conversation around skills, particularly around product, around marketing. And it, you know, last week we had the conversation about how companies are just not hiring level one people.

And we have a lot of anxious grads or just near grads that want to make sure they are getting skills in employable areas.

And so that’s my question today. If we think that AI is going to go beyond software companies, that every retail company, every media company, every insurance company, every healthcare company, every bank is going to have to rethink their user experience to be able to leverage either a language model or become more agentic and everything in between, what do we think we need to train for? Go ahead, Joanne.

[00:26:16] Speaker B: I think that there’s a couple of things that we need to train for. One that is often overlooked is process design.

Because really what you’re doing is you’re mirroring the process, whether it’s the customer journey or the internal.

Okay.

Or the internal individual internal to the enterprise. I mean, we have to start looking more deeply at the process because whether you’re using a generative AI tool or agentic, it’s the process that has to change. It’s not about the modality anymore in term of, in terms of form factor or speech to text or text to speech. It’s about what is this process trying to accomplish and how to give the best value to the individual who is using it, particularly on the consumer side. I mean, AI bots do not know if I’m in a bad mood.

They don’t know what my sentiment is, they don’t know what my intent is. And what they’re doing is they’re taking my prompt, if it’s text or my speech, and saying, oh, we think she’s trying to do X or she’s. Her intention is to do Y. They’re not looking at who is she, what, you know, level of education does she have, what is her age, where is she coming from?

Those things are derived by parsing, using nlp, for example, natural language processing to a degree.

What am I trying to say to the AI? That interpretation from them then gets translated. But there’s a lot of nuance there, and AI has a hard time with that unless it’s using mathematics or a symbology as the way it’s doing that translation.

So if we go back to the process of the customer journey or the, you know, put something in a cart or buy it, those are the areas where they can make the best improvement. And the people that they should be hiring for are process engineers, not UX ui. Because The UX UI is from a visual form factor based perspective. There are not that many UX UI people that I have ever come across that actually deal in other modalities like speech.

So I think that’s where you’re seeing a shift. And process engineering is making a big comeback.

[00:28:57] Speaker A: Interesting, interesting. What do you think, Liz?

[00:29:02] Speaker G: I’m a big believer in process engineering, but I’m gonna have to disagree on a couple of things. One, the, the idea of doing cx, not the UI part, but the CX part, really has to do with understanding the customer.

What I’ve been seeing lately is that there are a lot of AI tools out there that are using the language. Like you can speak into the tool and it will derive tone and provide feedback. I was looking at a recent tool that is used to train call center reps where they can actually read, listen and transcribe all the, the incoming call, listen and transcribe all the information that the is being spoken by the call center rep and provide feedback on intention of the caller tone, provide feedback on what you can do to help, you know, address all of their concerns. See if we actually did address all of their concern.

And I think that this is exactly where we’re going.

And it has to do with the key skill of being able to listen.

So that when we’re creating these bots or AI tools, whatever you want to call them, that they are trained on not only listening to the language that’s spoken and making sure that they’re thorough, but even going beyond that and contextualize what the person is trying to do beyond the specific transaction.

[00:30:46] Speaker A: Interesting. Oh boy. Got a little bit of debate here, but I think, you know, I’m going to round out where Michael Volinger just commented, right? We need a, you know, layering AI on top of a legacy process or a point solution or a step in the process and expecting exponential results will miss 100% of the time. And I think why it’s relevant to say this is that when we’re starting to get into the customer experience, we have a lot of rethinking to do about what their needs are. And I think that’s your point, Liz. I’ll take my break here. We’re going to bring Roman back to talk about skill development, our upcoming coffee hours over the next four weeks. I announced them earlier this week.

This week is Data Privacy Week. I probably should have had this episode be our data privacy episode. But you know what? We’re going to talk about what’s really important, what we learned and why it matters. Data privacy.

Given all the announcements that came out this week, we’re going to do a recap next week and say what are we changing and our goals and our objectives around data Privacy. On the 13th we will be talking about transforming to skill and outcome based hiring. I think we’ll have to talk about this from the employer and the hiring manager and the employee side. That will be on the 13th. I’m hoping Heather May will be back here for us to talk about that important topic.

20th it’s national entrepreneurship Week. We’re talking about opportunities for digital trailblazers to go solo and create your great AI startup. It’s not what it used to be two or three years ago. You got to be smarter about your value prop going into it now. So I’ll talk about entrepreneurship on the 20th and on the 27th.

This is from Liz. She debated me whether or not QA is important.

We’re going to talk about DevOps in the AI era, restating QA’s mission and I’m going on record in saying if we want to bring AI, LLMs, agentic AI, AI agents, we need a better defined QA function to be able to pull that off. Because brands should be terrified about putting something out there that is more open ended, that is going to be battle tested by customers and not having a prescriptive way of running tests before we release and before we upgrade those types of models. Roman, you’re the dean of students.

[00:33:21] Speaker C: I am not the trustee.

I am so low on the totem pole.

[00:33:26] Speaker A: Okay, but you’re closest person here for us at the coffee with digital trailblazers.

You’re going to steer 10,000 students over the next two years over where they should focus in their career. And AI is changing everything. Let’s talk marketing. Let’s talk product. Let’s talk it.

You talked about human computer interface. What other areas of skill development you think are really important for students, for recent graduates and for companies to cultivate?

[00:33:57] Speaker C: Product management is probably the biggest one.

What we’re seeing is AI even today is augmenting lots of different roles and skills that are out there. Product management.

I know this session is focused on the design discipline and user experience, customer experience stuff by augmentation. What I’m really talking about in that area would be things like user research is getting augmented by AI. That used to be a very time consuming, resource heavy thing that they had to do, but now you can do things like analyze thousands of user interactions, have AI driven heat maps, do predictive analytics that highlight specific friction points for real life problems, not just customers, but entire processes.

The other one is, and I’m surprised we haven’t gotten to it yet is rapid prototyping and testing.

Vibe coding has gotten a lot of press lately but if you think about it, high fidelity prototypes are now something that, that lots of different people can produce using vive coding tools like Lovable or Vercel.

The other part of your question though was I wanted to mention the fact that if you are a person that doesn’t have people you can or a company that doesn’t rely on or doesn’t have people with design skills that are ready to go, what we’ve done in companies that I’ve worked with, that’s work worked very well is starting any new innovative use. We started with a cross functional team dedicated to rethinking either the process we want to change or how we do that. And usually that starts with some kind of training.

So you know, a simple starting point might be a cross functional team to help put together an AI literacy training curriculum or an AI agentic AI process view way of changing things. So the big company, I think I’ve heard this multiple times on your coffees is the big mistake companies are making is they assume if they just give out gen AI to their employees they’re going to become more productive and somehow that’s going to create business benefit.

And alone I don’t think that can happen.

So that’s, that’s my short end of it.

[00:36:22] Speaker A: That’s a pretty, pretty long for a short end. Roman, I’m going to add another one from you and I’m hearing a lot more of this right now.

I’m hearing the executive that is projecting a moonshot vision and rallying the board for a big investment and then going back to their teams and saying let’s go do this.

And I’m listening to it, I’m like, you know, there’s probably five or six rapid prototyping that I would probably do before I went and promised that to the board.

And so I think we have, I think we have both problems. I think on the bottom end we have, you know, throw the tools at the staff and see what they make out of it. And on the top down end we have big promises that haven’t gone through enough sort of feasibility steps to say hey, we’re, we’re inching our way towards something real here.

[00:37:20] Speaker D: Right.

[00:37:20] Speaker C: And I guess my suggestion is use take advantage of people like product management or InDesign. Those people have a lot of expertise that they can leverage Even today if you don’t have those people put together a cross functional team that that can start usually with something simple, you know, either AI literacy or AI in the help desk, things like that.

[00:37:45] Speaker A: That excellent. Let’s go back to John. John, we’re talking about skills.

Let’s also breathe lead into governance to drive safe and innovation. I’m really interested in when AI is actively shaping the journey, what customers see the offers they get, even the tone of responses. What governance do we need to ensure these experiences remain on brand compliant and fair and without killing innovation. So go in either direction. We’ve got 20 minutes left. This is a great conversation.

[00:38:19] Speaker F: I think AI is so good at informational retrieval and so one of the skills that’s really important are really like the library skill, the librarian skills. And so I think the skills that somebody learns as a librarian are how to organize information. And the better the information is organized, the better the AI can retrieve it.

And you’re even seeing that now on websites that you used to be able to ask Google like a question to the website and the website owners would be responding on it on the my Google local business but now they actually Google just points at the website and uses its language model to answer the questions based off what’s on your website. So the better the website you have the better FAQs are automatically generated by things and and then so I think if you want any of this stuff to work, the language models can change at any time. And so I think you have to build some sophisticated filtering in so that people aren’t using things inappropriate. And so you really have to think through what is an appropriate request and what is not an appropriate request and you have to put a bunch of safeguards in so that you know things are appropriate for ages and then you have to test it it when it’s released and then you have to continually test it because these AI models that are third parties can change at any time without telling you. And so when I think of why Alexa has delayed rolling out a language model, it’s because I have three kids that are using it on a daily basis almost the entire time when they’re home. They ask it all sorts of questions and it could really hurt the Amazon brand if it was saying things that are inappropriate. And so I think that the continuous monitoring and just really hardening it so it’s not doing inappropriate things is so important to for brand protection.

[00:40:07] Speaker A: Love that answer John. Let’s go to Derek. Derek, skills or safe innovation? Where do you want to go?

[00:40:13] Speaker D: So that being Mentioned earlier, I think the rescaling but also factoring those things of cross sectional or cross functional type operating from a resilience point of view, I mean some of the things I actually put into the chat here, but looking at AI product owners, looking at AI and user design and conversation and design agents, these are all things going to be required. AI prompt and policy engineers and what they do is not just one particular unit that’s focused on, it’s really looking at all the business units that are going to come underneath this. And even taken to the point of security, reliability engineers, policy engineers, all focus on AI is going to be key. But I think Dan mentioned in the chat also one of the things is reskilling personnel to understand it. So taking somebody who’s been an architect that’s worked with the systems and now helping them to become an AI first end user type architect to help them better understand it, taking somebody that’s worked from either the finance, the marketing or the communication side of the house, have them work with developers from AI point of view to understand what they really need to help in marketing and actually work with understanding how to push it out as an end unit or product.

Security and risk teams are always going to be something that you’re going to need. As John mentioned earlier, being able to monitor adversary AI threats and concerns are going to be key throughout all organizations at every single level, whether you work with your internal stakeholders or external. Because this is things now as AI is getting more complex, they’re actually now finding ways to get around things such as encryption. So we need to make sure we have the proper AI threat intelligence tools in place and somebody monitor those to better understand how is this going to impact my organization and train them in the way that they can continue to understand and learn how these things evolve over time.

[00:41:54] Speaker A: Thank you Derek. I’m glad you brought up Dan’s comments. He’s got a couple on the common stream about architects. And my response to this Dan, is that if you look inside enterprises that have architects which aren’t enough of them, they’re siloed because of the level of expertise that’s required to build up that skill set. So there’s data architects, there’s application architects, there’s solution architects, enterprise architects, security architects, Basically every discipline has its own architect. And now when you bring an AI solution to market now you need a hybrid of those skill sets, right? You can’t just be an application or data, you need to see how those two fit together. And then if you are concerned about compliance and Security, you can’t be absent of those skill sets. So I think this is a call for super specialist, super generalist architects that can play in a number of different skills to be effective when you’re building AI applications.

John, what do you think?

[00:43:02] Speaker F: Yeah, because like things get there 90%, it’s a Pareto rule. But if you want things to work at the 100% level, that’s where you need the specialist. And I heard a talk by the Microsoft guy that led testing for Windows and I’ve heard it a couple times and people get mad, you know, at Windows for, for having all these issues. Right. But what he ultimately said is, is when you have millions and you know, billions of people using something, you’re going to find the edge cases. And the only, the only way to find the edge cases are, is to get a lot of people using it. And, and if you want to avoid the edge cases, you got to get the absolute best people possible doing everything you can to, to, to not have those edge cases. And that’s, that’s where you need the experts.

[00:43:46] Speaker A: Interesting. We think we have enough off of a tool set or methodology to find and test edge cases.

[00:43:54] Speaker F: No, no, I mean when I think about all the automated testing out there, there’s so much for security and there’s, there’s so much for like methodology that’s been built up over the last, I don’t know, 100 years of like computers. Right. But I think we’re just starting to get the concept of testing AI out and there’s, there’s frameworks for it and things. But I think it’s starting with the requirements of what’s allowed, what’s not allowed. I think you have to really think through what’s appropriate for the specific situation and that needs business buy in and then from there it’s like you need the technical stuff to implement what the business people want and there needs to be a lot of more work out there.

[00:44:33] Speaker A: Excellent. I got another comment here from Rachel Radin. Talks about change management. I need to put that plug in here with Martin out of the seat, speaker seat today. I completely agree. I think we need more people who are change managers. I also think we need more anthropologists. I’m thrilled my daughter is studying this.

When we talk about human computer interfaces and we start thinking about it in large scale, being able to really understand the impact on people, I just think it’s an under appreciated skill set that will become more important.

Let’s go to Joanne. Hey Joanne.

[00:45:16] Speaker B: Hi there.

I think one of the things you Know, to, to build on what, what I said previously and what others have said, it’s not just about putting together the cross functional teams, it’s also representing the codependencies, interdependencies and cross references of data and processes. Right. We never, we always tend to, you know, end up with siloed systems because when they’re designed and when they’re built, they’re a particular function. But with AI we have the opportunity to be cross functional and to think about how business processes interact and interweave with each other. You know, think back a year and a half ago to Digital Tapestry. The idea is that things are related to other things, things. So to your point, generalists, enterprise architects, process engineering, all of that needs to come in when we’re talking about whether it’s a customer facing system or not, whether it’s internal or external, because on the customer side you can’t interpret everything that someone is going to say a hundred percent, you might get to 90 if you’re really, really lucky. But you need to think beyond what their current ask is and what’s the bigger picture. So as we put people together in organizations to look at AI, we need to think about all of the aspects of that as well because if we don’t, you’re never going to be 100% true to Brent.

[00:46:55] Speaker A: Interesting.

Let’s go to Liz. Are we talking about governance or we talk about skill development?

[00:47:01] Speaker G: Well, all the above, but I think that I wanted to build on some of these conversations we were having about prototyping and making sure and finding the edge case cases. I think that those, and change management, all of those sort of go together around this sort of. I guess someone in the, in the comment was talking about cxo. I think it was also Rachel, about how we need to think about how the end experience is actually going to be enhanced and validated and do it in a way that sends out some kind of solution, gets the feedback, the anthropological feedback, the understanding what’s actually happening and then builds on that and builds on that so that we can actually incorporate those edge cases.

That whole approach is sort of a dovetail of, of, you know, Agile meets CXO meets AI. Like there’s so much involved in there and then I mean, obviously there’s the internal architectures, but I mean to me it’s all about business strategy and the business case and expanding your footprint and making sure it’s valuable in the, in the marketplace.

[00:48:12] Speaker A: You know, Liz, I’m translating that as the new experience better and in what ways that can be scaled. If I take my testing of Amazon Rufus yesterday, I would say it’s not better. I wouldn’t scale it yet. Keep it in that tiny little button at the top end corner and I’ll try it again in another month. We had.

I forget who it was. It was my friend Stephen and we had our episode on AI for Social Good and he talked about, really language models that are the new front end to forms, right? And so I’m going to type in a paragraph form instead of going and clicking through 16 boxes on a UI, that could be a really easy form factor because the output is controlled and you can do data validation on that translation from language into how that form is being fixed and you can get people to validate it. Once you get your data in there, you know, I think we’re going to find some examples of that. I think Joanne wants to chime in on an example here. So, Joanne, chime in in an example. But I do have one final question for everybody too, and you’ll be the first.

You know, when are we going to see the uber or Airbnb AI agent moment in a B2C context?

Go ahead, Joanne.

[00:49:35] Speaker B: Let me answer the question first. Before the end of 20, you will see it. There are people actively working on it the other.

And. And it’s because there’s not just a need for it, it’s because the mindset is starting to shift that, you know, where we were so ingrained in form factor and we were so ingrained in the way UI was being designed, people are demanding it.

You’re not the only one who’s dissatisfied with Rufus. There’s so many. There’s so many companies that are trying to crack this nut and the question is the approach that they’re using to do it and whether they’re using the right approach or not.

And it’s not about LLMs, Frontier models, it’s about small language models.

Because it’s a much more constrained question that you’re asking. Regardless of whether you’re talking to the bottom, you’re typing something in, it’s giving you a response. Etc. The reason I say the end of 2026, I don’t know when I. When I speak to Gail, our product, and, and I say, hey, what’s going on? It answers me and it tells me when I look tired and it tells me, you know, all sorts of other things which are. Which is a conversation in and of itself. But the other point that I wanted to make with respect to this is to change management and governance they go hand in hand.

And if we’re not paying attention to what actually makes up governance, and that includes tribal knowledge, we’re not going to hit the mark. Because when you ask for a tip, when you’re filling in that one text box, instead of clicking to, you know, 100 times on a form, you’re giving the user the opportunity to either opine or give their knowledge. And that should be used to reinforce the learning of any model, whether it’s a frontier model or a small language model. And that’s what’s going to promote the change that we’re all looking for on the B2C side.

[00:51:42] Speaker A: Interesting. So. But you think it’s coming aggressively soon?

[00:51:45] Speaker B: Very soon. Very, very soon.

There will be when, when you ask a bot human and it will actually understand you want to speak to a person.

[00:51:59] Speaker A: Interesting. What do you think, Roman? When’s it coming? When we’re going to see the sort of that inflection B2C application that will get us thinking differently?

[00:52:08] Speaker C: Oh, I’m already seeing some of that today. I mean, when I call my plumbing company, I’m talking to a bot and they’re trying to engage me, by the way, they’re trying to upsell me after I asked for the plumber to come out.

So I think you’re seeing parts of it already.

I do want to go on to your last question, though, about governance a little bit before we run out of time. I think you and Derek both talked about upskilling architects. I think we also need to upskill our legal and compliance people so that they’re more focused on trust and safety and able to create those guardrails and human in the loop rules and audit playbooks that we’re going to need for governance.

A while back, I think it’s almost a year, you had a guest speaker, Sukanyan, and she wrote a primer that I used to develop something for governing frameworks. You wrote a book called Governing AI for Responsible Future and it’s a great.

[00:53:17] Speaker A: I’ve read, I’ve read that book. It is a very good book.

[00:53:20] Speaker C: And what, what I, what I gleaned from it and built out of that was really governance can be done in three dimensions. You’ve got, today with AI, you’ve got the breadth, scale, the whole lot of people who can build a whole lot more things. And how do we manage with what’s going on, you know, because you can’t do code reviews on everything. Then we’ve got the think of it as the Y scale, the depth. How do you Handle a single solution that needs to scale to an enterprise level, you know, from just one or two people to hundreds or thousands of users. And then on the third axis you’ve got, as some people mentioned, like an operational oversight. How do people maintain oversight of these agents?

Need new interfaces for high volume oversight, maybe dashboards, which, you know, also needs to include security, access, data, exfiltration of data, you know, make keep going down the list. So yeah, those are the things you got to worry about in governance. And I kind of did it based on some of the reading I did and creating that XYZ access of breadth, depth and operational oversight. And where do you want to draw the lines?

[00:54:34] Speaker A: Roman, thank you for joining us this week. Really good comments.

I see a separate comment here from Joe is, you know, also debating when we’re going to see the Alexa Siri moment where there’s clearly really good AI and voice responsiveness. I still don’t think we’re going to see it in 2026. I think we’ll see pockets of it. What do you think John? Do you have, do you have a favorite B2C?

[00:54:58] Speaker B: Yeah.

[00:55:00] Speaker A: One that you think is. Go ahead.

[00:55:03] Speaker F: I think the one that’s going to come in through the back door is the AI agents built into the web browsers because the web browsers are kind of like the middleman between one of the most used interfaces between us and the world and it’s used by so many things. And so I think we’re starting to see it already is that they’re stitching AI into the, the browsers and that’s going to have some serious governance implications because the stuff that you see on your browsers, financial business, everything like that. And then I think we’re actually starting to see AI specific browsers being built and that’s going to be really interesting.

[00:55:38] Speaker A: Do you have a favorite one you’re using now?

[00:55:40] Speaker F: I don’t use them because I just like, I don’t trust it. And so, and so especially so at work we’re dealing with very, very sensitive HIPAA data. And so I just, I don’t, I don’t, I have to like go through all my tools to make sure that they’re HIPAA approved. And, and so I’m, I’m waiting.

[00:55:56] Speaker A: I think you just introduced a topic we’re going to have to cover. We have not covered AI web browsers yet.

Every time I log into Perplexity, it’s bothering me to download comment and I’m like, I don’t have time to play around with this. Joanne do you have. You have. Have you played around with these yet?

[00:56:16] Speaker B: Yeah, I have, actually, and I’m liking a couple of them. I’m going to keep them out of the mix of which brands they are, but let’s just say that my Perplexity account and I are parting ways.

[00:56:31] Speaker A: Oh, you know, I’ve been feeling a little bit of that myself, but for different reasons.

You know, I think we’re going to see a lot of leapfrogging back and forth for the next few years, which I think is another reason why we’re a little afraid to start putting AI first uxs out, because it’s.

If we built it with one partnership, with one model and the cost change or the capability changes, we got a lot of rework to go. Do. I think brands are also cognizant? I think CIOs are cognizant of that.

It’s been a really interesting conversation. Any last thoughts? Joanne?

[00:57:06] Speaker B: Just very quickly, I think to John’s point, they are being built into the browsers.

I think the browser itself is getting radically different from what I’m seeing, because even those that are big makers of them are recognizing the need for multi modality.

Yeah, you need to be able to talk to it. You need to be able to have it. Have the heuristics that are needed as if you were speaking with another human.

[00:57:39] Speaker A: See, I don’t know, Joanne. You know, coming off my trip to Japan a few weeks ago, like, voice is not going to be a big win in Japan. And the simple reason is I can click on my phone on the train. I can’t use voice on the train. It’s like a.

There’s nobody speaking on trains there. It’s just one of those things. And it’s beautiful and it’s amazing, but you’re not going to see people in Japanese trains talking to an AI bot or an LLM or anything like that. I don’t think you’re going to see that in India as well because of the many dialects and languages that are spoken there. You know, I don’t know if that would be my first place to be looking, but I do think the form factor is going to change.

[00:58:22] Speaker B: Yeah, 100%. And, you know, there are different ways to skin the cat. I mean, it is a cultural thing. There’s no question.

I think, you know, we are not all that verbose on our trains either.

I think the issue is also, there’s also a need. Two needs, and one is accessibility.

Right. I mean, you. You’re gonna want to be able to have AI that understands international sign language. You’re gonna have it for any number of particular needs based on accessibility, but also generally speaking, that we’re all getting really tired of typing prompts.

It takes too long. It’s cumbersome. We need another way to be able to do this. You can’t do it on a factory floor as an example, but you could talk to something and that’s why mixed realities are so big.

[00:59:26] Speaker A: I think we’ve got a bullish forecast that we’re going to see big changes this year given all the skills and disciplines. We still need inside companies to change customer facing user interface. It’s been a really good conversation folks. Thanks for all the comments. I captured some of these in the whiteboard and this, this episode I will publish publicly. So thank you for that.

Jay Farrow closes this that’s all we need, more people talking to their phones out loud in public. I agree with you Jay. Folks, thank you for joining week. We’ll be back next week to talk about Data Privacy Week, what we learned and why it matters. That’s on the 6th. On the 13th, transforming to skill and outcome based hiring.

On the 20th we’ll talk about National Entrepreneurs Week, opportunities for digital trailblazers. And on the 27th we’ll be talking about DevOps in the AI era, restating QA’s mission. If you want to be a speaker, if you have ideas for topics, please do reach out to me. And folks, a lot of snow, a lot of cold in the US going on right now. Although I am thrilled at the start of this episode we saw so many people joining internationally from India, from uk, from all over.

Please invite your friends and let’s continue to make this a really vibrant conversation here at the coffee with Digital Trailblazers. Just remember, you can get to the next week’s episode at the URL starcio.com coffee that will redirect. And if you want to look at previous episodes go to drive.starcio.com Coffee I have the previous episodes posted there. Folks, have a great weekend and I hope to see all of you here next week for the coffee with Digital Trailblazers. Bye now.