Shawn Cordner (00:13.294)

Hey everyone. Thanks for joining the Amplitude of Tech podcast. I’m Shawn Cordner, Chief Marketing Officer of Amplix. Today I have a great podcast for you. It is the first return guest, Scott Weiner from NeuEon. He’s their Chief Technology Officer and AI Transformation and Strategy leader. We had a conversation about his predictions that he made last year when he was on the podcast. And he also gave us his predictions for 2026. Some things are going to surprise you, some things might not. So I think you’re going to enjoy this one. Thanks for joining.

 

Shawn Cordner (00:48.654)

Let’s get into the nuts and bolts of your predictions from last year. You said that humans are curiously bad at predictions. Let’s see how you did. First prediction you made was agentic AI will redefine enterprise workflows and adoption will accelerate. From everything you’ve said so far, it sounds like you hit the nail on the head, but let me know, how did you do?

 

Yeah, I think so. think, I think I did. think, I don’t, I it’s tough because I know what I said. I don’t think when I said that I was trying to say that the world was going to be running on agents by, 2026. I think I said what I meant was that, agents would be the focus of the development that’s going on. And it really did. I think where I was off though is.

 

the emphasis has been much more on software development than other areas. Like software development’s really shined in agents and AI. In other areas, it started to kind of happen, but not nearly at the level that it did in software. So I think software kind of got ahead of everyone else, which is fair. But yeah, I think that’s, and I still think that that’s where we’re going. I think the big change is that organizations are now realizing that,

 

orchestration is a big part of this and they’re going to have to put a lot more emphasis on orchestration than they did in the past. Things that we ran into that I don’t think were fully anticipated, but I think we saw some of this was, you know, a lack of memory and context is part of the problem. Until we have more context and memory, it’s hard to have agents do long term stuff, which is what we’re hoping for. Organizations themselves just have friction. You know, there’s just an avoidance of wanting to use these tools.

 

Um, some of the things we just talked about, you know, there’s the fear out there. Um, I did see a lot more focus on demos over value than I had hoped for. I’d hoped more companies would get serious about it. And a lot of them did these pilot projects. And that’s why you saw some of those weird stats, like, you 95 % of pilots fail. Well, yeah, but a lot of them were never intended to be ROI. So that’s not really a fair thing, but that was a, you know, part of the statement, but infrastructure is the wrong thing. Also, I think the metrics have been wrong. think we’ve, um,

 

Scott Weiner (02:59.138)

Companies were measuring the wrong things when they focused on AI early on. think they, they overestimated what it was going to do for them. So while I thought agents were going to come through and, and did, I think that the focus on automation was probably the wrong place to focus. It should have been more on collaboration than automation. And the main reason is that these are non-deterministic systems. These are probabilistic systems that we’re building because we’re building around large language models mostly. And.

 

They’re just not good at deterministic stuff. If you need it deterministic, build a workflow. There’s great workflow tools out there. Use AI to do something that should be just a workflow. So I think they conflated the problems with what it’s good at. And I think they’re getting smarter about it. So I think it’s coming along.

 

Now the next one I think you just touched on, if you want to expand on it, please do. But if not, we can keep moving. But it was the shift from experimentation to operation.

 

Yeah, that’s exactly, I kind of talked to it already. Yeah, exactly. think that that didn’t happen evenly. So where that did happen really, really fast was in finance. They seemed to really, the financial organizations seemed to, maybe they were just because they were ready for it, but they really seemed to get behind this idea of how to operationalize this much faster. They had the governance in place.

 

They started putting the infrastructure in place. mean, it’s not, it’s not totally even within that industry, but that industry seemed to take it much faster than a lot of others did. So we’ll see more of that in 2026 for sure.

 

Shawn Cordner (04:34.638)

What’s driving that do you think in the finance industry?

 

well, the desire, I, I, I, again, I think a lot of it is just that they were more organizationally ready for it. They have better data. They have better systems for compliance and for regulatory purposes. think they’re just more ready for. One of the things we found from a couple of the studies that came out last year is that organizations amplify their successes and failures with AI. do not change their, their trajectory. So if you’re already well running.

 

AI can help you. If you’re not well running, AI is going to make everything more obviously bad is what’s going to happen. You’re have more problems. It complicates your life if you don’t have good things in place.

 

That’s such a great insight because I’m hoping to get a friend of mine that is head of compliance for digital assets at Citi to be on the podcast next week. For those listening, my dog decided to join the podcast here for a second and she’s knocking things off my desk. So I’m hoping to have him on there. We talk quite a bit about the challenges of

 

Thank you.

 

Shawn Cordner (05:45.474)

deploying AI in heavily regulated industries, but you’re right, they do have a robust compliance infrastructure and a culture around compliance and regulation and cybersecurity, right? So maybe in some ways it actually is easier in those industries.

 

I so. I think it is.

 

Next prediction was multimodal AI.

 

Yeah, that definitely took off like a rocket. It was amazing how fast that happened with this idea of, you know, AI that can look at things, know, images and video and sound and all that. That actually came faster than I thought I was going to. That was surprising how much that’s taken off. Now there are models coming out where the models are actually from the ground up trained, not just that they’re combined.

 

models, but actually trained from the ground up to be able to look at all of those things together. So it’s getting even more efficient and effective. what that why that’s important is when you think about like robotics and the ability to have AI that is in our world and experiences our world, that’s how they’re really going to learn in the future, right? They’re going to be experiencing things the way we experience them. That’ll make them much more useful to us. And multimodal was part of the formula that needed to be kind of done. So yeah, that was exciting.

 

Shawn Cordner (06:58.098)

And the next one was, hang on a sec. Sorry. We’ll clip that, mark that Brendan. You guys hear me okay? Yes. Okay. All right. So the next one is a work train workforce transformation and upskilling. It seemed like you thought that enterprises were going to be investing in their workforce and getting them trained and ready to use AI competently and safely.

 

Yeah. that’s a tough one for me. I don’t have enough visibility into all the industries to tell you if that’s across the board, but that definitely, there’s definitely been an uptick. There’s no doubt. forget the percentages now, but it’s definitely gone up significantly. I’ve been disappointed at the lack of, basic training that has gone on at the org. You’re investing all this money in AI now. They all, they all are, but the investment in training and skilling of their employees has been

 

in my limited view, minimalistic. It’s been insufficient. And I’m biased towards this for lots of reasons. One is that I do that kind of work, so I care. But it’s frustrating to watch people using AI and still kind of doing rudimentary prompting and not really understanding what they could be getting out of their experience with AI. And it’s unnecessary. We have all the information. We know how to do this well. It’s not there yet. So it’s definitely happening, but not as fast as I would have helped.

 

hope.

 

Would you recommend those listening that are in that position to make that investment in 2026?

 

Scott Weiner (08:30.302)

I think we’re going from, they’re going to be doing it to you have no choice. think that you’re going to find that companies are going to start falling behind that don’t do this. think we’re at that point where, the difference between what someone who knows what they’re doing with AI and someone who doesn’t is, is going to be orders of magnitude. It’s going to be such a difference in productivity between someone who understands how to use it. And now what I, and the only caution I have in saying that is I also can’t see yet the tooling.

 

If the tools get much, much, much better, then it may be able to hide a lot of the complexity from people so that they don’t have to do that. So for instance, I talked about early on these harnesses. Well, it may be that that kind of sophistication is built into the tooling so that you don’t have to think about it so much. So it’s possible it gets buried eventually to the point where you don’t see it. I just don’t see that happening in 2026. I don’t think we’re there yet. So.

 

I think if you want to be effective with it, I don’t, almost any industry, if you’re using AI for anything, getting some education on how to do context engineering and prompt engineering and all that is kind of a table stakes. You kind of have to do it.

 

And that’s an important distinction to make as you’re talking about investing in the skill of using AI and not necessarily awareness around the risks of AI, right?

 

Well, so the risks part, right. Any company that isn’t creating governance around AI is actually just asking to go out of business. I mean, at this point. I mean, and because not only do you have every single person in your company is now using AI in some way, whether you want them to or not, but the idea that they don’t understand how to use the information safely or what’s being trained on or not being trained on and you know, all that stuff.

 

Scott Weiner (10:16.174)

There are horror stories. You’re either going to invest in training or you’re going to invest in paying for lawsuits. It’s like one of the others going to happen. So, you know, pick, pick your poison. You can’t, there is no middle of road with governance. And the problem is companies have been slow. And what do we say before? That’s thing about AI is the speed at which it’s coming around. And so we don’t have time to sit back and wait. We have to invest in this. And, I really feel bad for any company that doesn’t, I think they’re going to.

 

I think they’re really going to pay the price at some point. I really feel now that it’s more of an when, not an if. so how long can you go without health insurance is kind of look at it.

 

some point, if not already, it’s going to be a requirement for cybersecurity or cyber insurance, right?

 

Yeah. I, it already is in some areas. Yeah. It will be the, the insurance companies are going to start requiring it. Absolutely. here’s the problem though. All, all compliance frameworks, even the best ones out there are static in the way they’re designed. You know, they’re, they’re, they’re, they’re check the box type of compliance processes and AI is evolving so fast that those standards will not keep up.

 

And so you need to have an agile kind of process to evaluate what does it mean to be compliant. And so I think that the spirit of it is more important than the letter of it. Like, are you really thinking about things in the right? Like for instance, being able to explain why the AI did what it did, being able to have observability, being able to have layers of protection. For instance, there was just an article came out, another person died because of, I think it was ChatGPT.

 

Scott Weiner (12:01.738)

naming one, if I’m not a hundred percent sure, think it was ChatGPT. But it happened because they asked it for medical advice or medicine advice. And ChatGPT or whatever it was said, no, I won’t give it to you. And they just kept poking at it. And eventually it gave them advice and they followed it and it died. I don’t know the whole story. So I don’t want to go too far into that one. But the point is like, that could happen on your chat bot on your sites. That’s supposed to give people like advice on how to use your, your equipment, you know.

 

And they, they ask it about medical advice and you don’t have the right protections in place. And you just gave them advice and they go kill themselves. And so you’re responsible. Um, generally what lawyers tell me is that I’m not a lawyer is that, um, the, the sphere of responsibility has to do with the sphere of control. So if you have control over the thing and you don’t protect it, then you are responsible, not people down downstream from you.

 

So this is where companies can’t just say things like, well, I just trusted my AI bot that I bought from this other company. Like your site, you’re saying it’s you, you’re responsible. You you can’t get away from that unless you have some legal thing. So I think, yeah, I think this is a big area. We have to really watch for this. This is going to be, we’re going to see a lot of unfortunate things happen in this area, but hopefully.

 

people get smart about it. don’t know what to do because I, I spent a lot of my time talking to companies about governance and I feel like they listen to me and they say, oh yeah, yeah, you’re right. But they don’t necessarily do it. So it’s one thing to kind of say you’re right. another to do it. And like, I don’t, not trying to be right. I’m trying to give you something to do. So hopefully they do it. Please do it.

 

Please do it. And the next was industry specific models. And think you touched on this before. It sounds like this is coming fruition.

 

Scott Weiner (13:53.868)

Yeah, yeah, it is. It’s interesting because what I hadn’t seen coming as much was how much optimization is happening, how fast it’s happening. So the models at both the high end and the low end, so the small language models and the large language models are all optimizing. What we’re also kind of finding is, you one of the problems we have right now is that we’ve run out of interesting data. So

 

Where’s the interesting data? It’s specialized, it’s within organizations, it’s capturing and leveraging the data that you have within your organization, which is great because that becomes a moat as well. Right? So that’s not a bad thing necessarily, but you have to know how to capitalize on that. So some companies have already figured that out.

 

Okay, and your last prediction that we were tracking at least is large world models for training.

 

Yeah. It definitely was work, research work that was done last year. A lot more talk about it now. There are companies that have spun out that that’s what they’re focused on. Some big ones, some big names are doing that. It’s still early. It was, it was not something that took off necessarily in terms of commercialization last year, but it’s definitely something that we’re going to see more and more of. And I think you’re going to hear more of it this year too. We need it because those, those,

 

Those world models are really about trying to give AI more perception of how to learn in our world. Like we’re going to need that for things like, you know, self-driving devices and things like that. We need to be able to have things that can learn in the real world and you can’t just let them go wild in the real world. That’s not the way they’re going to learn. So remember how these things work for people who are listening is that, you know, AI is basically a trial and error machine. You know, it tries something fails tries again million times. Well,

 

Scott Weiner (15:38.798)

You can’t keep crashing into people and things to figure out how to drive. You got to have some other way to learn. So world models are going to be really, really important for that type of thing. The other thing I talked about was, I don’t know if it’s on your list, is synthetic data. I don’t know if that’s on the list that we talked about. So yeah, I did talk about this last year too, but synthetic data is going to become even more important. And it’s going to be much easier now because the models are getting so good, they can create fake data that looks like real data. I was just talking to a client this morning.

 

actually they’re trying to build an educational system and what they want to do is they want to be able to capture from students how they did at the beginning of the year, how they did each quarter and how they did at the end of the year using their methodology and their tools so that they can prove their system works, right? It’s funny, I just got an email from them just when we were talking, it’s weird. It just popped up. so anyway, so the idea is that

 

they’re saying, we gotta go collect all this data now from all these schools to figure this out. And I said, no, you don’t. I said, go get data from some people at the beginning of the year. Tell me what you want the end of the year to look like. What’s the best case scenario for the student? And then we can have the AI simulate all the data in between. And we can create thousands of variations of this. And then you can use that for your prototype, your MVP training data. And you can train the whole system on data that looks like the real data.

 

because we can use really, really expensive high-end models to build this data and then use a really small cheap model to run the system, right? And that’s kind of how they’re going to do it. And will they get real data eventually? Of course, when they go live, they’ll get real data, but this gets them, you know, maybe it’s 80 % of the way there with very little investment.

 

Data obviously is critical to AI, but it does seem like it becomes a crutch sometimes for why someone is not starting an AI project and not having the data is one of the excuses that you hear sometimes. So that’s interesting that there is a solution for that.

 

Scott Weiner (17:42.734)

Yeah, what I said. That data is really important for that. That topic is the ability to simulate it close enough. In fact, you know, that’s how some of these big open source models like DeepSeq, which when DeepSeq came out at the beginning of the year, it created a huge panic because all of a sudden this really inexpensive model made for a fraction of what the large models cost looked like it was as good or close to as good. The reality is they took.

 

They took one of these bigger models and they said, hey, give me give me fake data and we’re to use that to train our model with basically give us data and we’ll train our model with your data. And so without the big model, couldn’t do the little model, right? It’s not that little, but you know, right. So I wouldn’t call it cheating, but I would say that it’s not the same thing. and you can only get so far with simulated data because at some point it’s, needs real world anomalies. can’t predict everything in the, in the training data for a lot of situations.

 

For instance, if you were building a healthcare system, there’s probably enough data out there for me to extrapolate, you know, people with certain, you know, characteristics have certain illnesses and so on, and I could do some prediction models and so on. But at some point I need real data. can’t, you know, I can’t do that forever.

 

but it could be enough for a pilot.

 

Absolutely, to your point, right? So the idea that data should stop us from trying, absolutely no reason not to be able to kick off a pilot without having all the data. And in fact, if you have bad data within your organization, meaning it’s not well suited to AI yet, this may be a way for you to kind of visualize and think about what the data needs to look like in the future by trying some of this out.

 

Shawn Cordner (19:26.048)

Interesting. Okay. So that’s enough of your past predictions. I’m going to give you a solid B Scott only because you missed the mark on a couple, but I’m going to give you an A plus for showing up here and facing your predictions head on. a lot of people always. So I appreciate that. Next, I’m going to put you on the spot for some predictions for next year. But before I just let you turn loose on your predictions, had a couple of things that I wanted to ask specifically about. One of the things that

 

just happened over the weekend actually was AWS rather sneakily increased the cost of GPUs 15%. And that is going to potentially increase the cost for AI workloads for a lot of enterprises in a, that’s a pretty significant increase. So what’s remarkable about that, aside from the significance and the magnitude of the increase is

 

AWS has pretty much trained the market to believe that compute costs are going to continue to go down. And now all of a sudden, we’re seeing costs go up. And that was something that most people probably didn’t predict. So I wonder if others may follow suit. And what I’m asking is in 2026, do you think something like that could potentially have a cooling effect?

 

Yes, it could. It’s hard to say. Let’s put it this way. Last year, the GDP of our country grew because of investment in AI infrastructure. Like that was the main driver of our GDP. That is an investment in the future. That is not an investment in today. So it’s also possible that those prices start to come down when some of these infrastructure investments start to pay off. that’s also a possibility.

 

I don’t know enough about the economics of it right now, how it really is going to play out. I do know that there’s been a lot going on in the chip market lately and because of that, you know, AWS is reacting to something. And so I don’t know if they wanted to raise the prices or if they had no choice. I’m not sure. So if if others follow suit, it’s probably because they’re feeling the same cost pinch. That’s very possible. I got to tell you, that’s one of my biggest worries is that inference costs start to rise because.

 

Scott Weiner (21:45.006)

If they do, yeah, it’ll have a cooling effect, but if companies are already kind of invested and locked into these business models, they’re going to have to live with them. Uh, interesting thing happened to me this weekend. This is how personalized this particular problem a little bit. Um, I’m on a, uh, a monthly plan for my, my own AI, right? But I also track costs at the API level. So even though I’m not using tokens, uh, I track them. And so every time I send a prompt in, let’s just use that as an example.

 

I get a cost analysis automatically that’s coming out of the system for me. So this weekend alone, I burned through $400 worth of tokens. This is my personal account, right? Just in the weekend, but I’m paying like a hundred dollars a month. At some point that economics doesn’t work for the company that I’m using, right? And so what’s going to happen? Think right now, I’m getting so used to it. I love it. I’m having a great time. I’m building all kinds of stuff and I’m pretty efficient at what I’m doing, but I’m not.

 

I’m not being super efficient because I’ve got this all you can eat plan, you know, yes, they can at any point cut me off or say you’ve hit your limit for the month or the week or whatever. I don’t hit the limit. So I’m happy. Right. At some point when they start saying you’re hitting the limit every hour, I’m going to have to switch to an API cost and say, I’m going to have to not do it all. can eat anymore. Well, if that happens, I don’t know if I can afford to do what I’m doing at some point. That was my first reaction, but I just want to put this in perspective.

 

Let’s say it actually did cost me $200 a day to run if it did. So I’m looking at like, you know, $4,000 $5,000 a month in costs. That is nothing compared to what it would cost me to pay someone to do what I’m doing with AI. Today.

 

The question is, are you getting $4,000 $5,000 a month in value of what you’re doing with AI today?

 

Scott Weiner (23:37.614)

Right. Right. I am. I can’t guarantee that everyone is. I’m getting, I think I’m getting $20,000 worth of value, honestly. I think I’m doing like a whole development team worth of people doing stuff for me right now. I don’t know that everyone can do that. I don’t know if everyone has the skillset yet or the infrastructure to do that and so on. Again, this is at a personal level. Business could do this. Sure. Sure. And so I think that that’s going to be the other part of this is that we’re looking at the economics a little

 

backwards because I think the consumer, yeah, they’re not going to pay $400 a day for this, but will the business be able to do that and justify the cost? think in a lot of cases they will. What might be cool to some of these pilots that have no value attached to them. That might cool. That makes a lot of sense.

 

Yeah. Yeah. But like we said, even at the business level, there’s starting to be more scrutiny on the return on that investment that’s being made on it. So the cost will have a cooling effect. mean,

 

That’s economics, right? Yeah.

 

But maybe that’s a good thing, right? I don’t know. You know, that MIT research study that you came out about the 95 % of pilots failing, my first reaction to that was, so what? We’re learning, right? And that’s what the real value of all those pilots are is getting through the learning curve faster. And those businesses that are making those investments right now are going to be ahead of the curve from the competitors that are not making those.

 

Scott Weiner (25:03.404)

I mean, I will say, I don’t know that the way that model was, that study was put together was great either in terms for this purpose, for the arguments that are made about how, look at all this. They looked at a very narrow segment in a very specific way. I think the bottom line is you can, you can ask any business, do you think AI is helping you? That’s actually implemented something in real. They’ll tell you it’s helping. They will.

 

I can’t speak to ROI. can speak to productivity and all of that. So like I said before, if you’re creating more capacity, but you don’t have a plan to use that capacity, then yeah, okay, AI helped you. That doesn’t mean you’re going to see ROI. So you got to have more than just a, you have to have an ROI plan if you want ROI as an output. If you just want to make life better for your employees so they don’t have to work as hard, great. I want to work for you too. I mean, that’s, that’s wonderful.

 

That’s an important point that I think a lot of technology leaders need to hear is that we’re in an age now where, you know, AI as a technology is not something you can look at in a vacuum, right? It’s, part of a larger ecosystem and you have to look at it holistically and you have to think like a strategist rather than a technologist to make sure that these initiatives are having the business impact and that business impact is not going to happen by having the AI initiative.

 

implemented in itself. There’s other pieces of the business that need to be aligned behind that and for you to be able to get the real value.

 

Absolutely.

 

Shawn Cordner (26:38.018)

Okay. My next one is trust. think 2026, from my perspective, you’ve talked about governance quite a bit. In the past, you’ve talked about explainability and observability. And so I think that 2026 is potentially going to be all about being able to trust the AI that you’re implementing. What do you think?

 

and observability.

 

Scott Weiner (27:03.566)

Yeah, I may have written an article about this already this year. I am very big proponent of the idea that the companies that can develop trust in their AI are going to win. That there’s going to be, there already is a backlash against AI and trust, trusting of AI. You know, we all heard about hallucinations last year. That’s all I ever heard was hallucinations. Now the good news is hallucinations are getting better. That’s almost

 

a bad thing because that means we’re going to start to trust the AI before we should. So it kind of cart before the horse a little bit. What I’m talking when I say trust is that we put systems in place that guarantee that what we’re seeing is real and not all those systems are necessarily AI systems. There are other systems you put in place around it. A quick personal example. So I’ve been going through some of my own AI development and I’m building out a fairly large AI agent network that’s solving some problems for me.

 

And the first version of it was a bunch of prompts. Version A was just a bunch of prompts. And then the second version was prompts with tooling. So the prompts then have the ability to do things for me in the real world. Go do research, write reports, create fancy graphics and so on. And then the third version of it was scaling. So now this one from prompts for me to do work.

 

to putting it on a web server and giving other people access to it. So now you could log in and say, you know, build me a report or, or do this, for me. And, and that was better. But then what happened was I realized the process of putting it out in scale meant that I needed to have some state management because now you have different people doing things at different times. And these, the, the system is not, it’s not a prompt. It’s a.

 

It’s a series of things that have to happen in a sequence. Okay. Like you do research and then you have to synthesize the research and then you have to organize it. And then you have to have a kind of back and forth with the creator to say, what do you think and all that? Well, when you’re sitting at a, at a, uh, on your website, right. And you’re just typing in to chat, GPT or whatever the prompt, you know how to do that. Right. You say, that’s not what I wanted. I wanted this and you have it. Right. Now that’s all hidden from you because that’s all behind the scenes. All that, that part.

 

Scott Weiner (29:21.526)

And so what you’re interacting with is what? And so I came up with a user interface for you to kind of interact with it. The problem became that if I just flow what you do into the AI and then back, you have no visibility into what’s really going on in the world. Like, what’s it thinking? What’s it doing? You know, and I could spit out all of its content, its thinking to you and all that, but because it’s, it’s moving real stuff around, it’s, it’s saving files and it’s opening things and it’s researching.

 

You need more than just the, the, the it’s thinking thing, right? So what I ended up doing is tying it into a database that, that, can actually lock down things. So an agent can say, I’m working on this task and nobody else can. And it could create an atomic lock on that thing using the database as the atomic lock. And so now I could have multiple agents running on the same project and not worrying about them overriding each other or trying to do the same thing at the same time. It’s solved a lot of problems. So I’m trying to say is I went through an evolution.

 

Well, let me fast forward this evolution a couple of weeks. This is a couple of weeks in work. This isn’t years of work. Where I’m at now is about 60 % of it is actually just raw Python code that is a framework, a rail, and the AI is just doing the creative parts of the process. So everything else is automated tooling and it’s saving every single event to a database. And so I have every started this and did this.

 

and so on, and the AI has been trained to reply in a very specific format. And when it doesn’t reply in that format, the code that’s sitting around it basically retries or does other things, or eventually comes to me and sends me a Slack message and says, you got a problem, go fix it. Right. But the point is that I’ve had to put a lot of rails around it and now I trust it. It took me weeks of iteration to figure out the best balance between

 

How do I keep the thing creative and flying and efficient and effective? And I can trust everything it’s doing is really what it says it’s doing. And I’m there. Now there’s other things I’ve done. Like there’s, techniques that call re-ranking and I have, I have evaluators that kind of evaluate the output and I have other things going on in the city. So it’s a more complex than this, but, the bottom line is that a lot of the solution was not AI. And so that’s, that’s the other piece of this. And all I’m saying is that infrastructure is everything. And that’s what companies are going to have to be focused on is the orchestration.

 

Scott Weiner (31:46.144)

of the AI is going to be what makes this work. Got it.

 

next thing I wanted to ask you to comment on is, this most recent CES, seems like everything was about AI and the equipment. What do you think about that for 2026?

 

I think Johnny Ives went to chat to open AI and I think that they’re getting close to announcing something next year. And I think 2027 they said, and I think everyone’s like, we better get our thing out before they get theirs out. I think that’s what’s going on. And I don’t know. I don’t personally know if the demand is there. Now I’m a, I’m a fan of this stuff. I have a lot of toys. have a lot of tools, AI tools, and I love them and I play with them a lot.

 

but I feel like I’m playing with them for the most part. Very few of them do I find are actually solving real problems that I really need. They’re not answering the problems really. They show me hints of where we’re going, and I think that’s really exciting. I can tell you the number one tool that I have right now that’s an AI tool that I use is this one. It’s a Plod Note. I don’t if you can see it. It’s probably the closest thing I have to something really useful from an AI perspective.

 

And the only reason I say it that way is it’s essentially a recording tool that also creates transcriptions for me, right? I could do that with my phone. So when I say almost, I’m like, you know, it’s more convenient than a phone and it’s better in a bunch of different ways. But, but you know, it’s a step up from my phone. It’s not like it’s this evolution, a revolutionary thing that I can’t live without. I am not hearing about anything coming out that I think feels like revolutionary, you know.

 

Scott Weiner (33:28.386)

Like I got to have that, you know, that’s just so new and so exciting. I think the exciting thing is the robots, except they’re not there. They’re not ready. You know, I’d love to have the robot that goes and makes my dinner. You know, I love, I love the idea of it, but, but we’re not there.

 

I know man, I saw that Will Smith movie, it doesn’t turn out well.

 

Hey, did you see there was one in China? Was it not the R one, the X one? I can’t remember which robot it was. But it was like a, it was like cutting up vegetables and then it just went crazy all of a sudden out of the blue and just started slashing.

 

No, I didn’t see that. to be honest with you, Scott, I used to be a chef and I worked in a lot of kitchens and that happens more than you think with real people too.

 

Nice.

 

Shawn Cordner (34:13.71)

What kind of chef were you? I’m a classically trained chef. I started cooking when I was 15 years old and got a full scholarship to a culinary school and I cooked until I was about 22 and I realized that working 100 hours a week for $7.25 an hour wasn’t going to get me where I wanted to go.

 

Nice. Well, that’s fun. Yeah, I was in India once and I had a couple days to kill and so we were staying at the Ritz and they had a really nice restaurant there and they said, hey, if you want to come down, we’ll teach you how to cook. And so I spent several hours with them and I got a master chef certificate and a big hat. Don’t think it’s quite the same thing, but there’s a lot of

 

The hat’s everything though. Okay, so that was the end of what I wanted you to specifically comment on. So Greenfield here, what do you see coming in 2026?

 

so the year of agents may have been last year, meaning that people started talking about thinking about focusing on agents. And I think that was true. I think if you look at it through that lens, this year is the year of agentic infrastructure. think this is, that’s kind of what I was trying to say is that we’re going to find a lot more emphasis on how to orchestration and measurement, measurement, at a P and L level of course, but also just measurement of productivity within, within the, how the AI is functioning. people are going to start really focusing on the economics of AI.

 

we, we’ve been talking about that, but that’s what I think you’re going to see a lot more of, data quality is going to continue to be both a problem, but also solutions are going to start to come up because it is the number one bottleneck that organizations have to being successful is data quality. And the other one is talent development. think so getting away from AI as a thing, think AI skills are going to become really in demand. They already are, but they’re going to continue to grow in demand. And I think you’re going to see a lot more organizations investing in AI talent.

 

Scott Weiner (36:11.522)

I’m hoping that also translate into re retraining their workforces. That’s my hope. I hope that they don’t make the mistake. I feel it’s a mistake of saying, we can, we can just get new people. I just, I really hope they don’t do things like that. I think companies are going to move from kind of a bottom up experimentation to a top down strategic approach. If that makes sense. So you’re to see more senior leadership focused AI investments. And I’m already having those conversations with, with executives. I’m already literally that’s been a change is.

 

In the past, it’s been a lot more like the director level saying, we’ve got this project I want to try, can you help us? And now I’m hearing a lot more from CEOs and CEOs and CFOs even kind of talking about, you know, I really need to get us aligned with what’s going on out there. So a lot of the challenges that we have, hopefully we’ll start to, you know, kind of get dealt with. Obviously there’s a massive impact on economic growth.

 

I don’t, I’m not a financial guy. don’t, I don’t know how big it is. It feels, it feels like this is going to be a big year for AI again. I don’t think this is going to be, I don’t know it’s a repeat of next year or so on. You talked about cooling. That is tied up in politics too. So part of it is, I don’t know how much like the terrorists are having an impact too, and, other things going around on on the globe. mean, I, that’s a tough one. There may be a cooling, it’s possible, but.

 

I don’t see any cooling right now on the innovation side of things. So, and the commercialization maybe there will be. But everyone I’m talking to is telling me, you know, I got to get our thing up and running. We got to get our, we got to optimize and organize and, and solve these problems and AI looks like the solution. So let’s figure it out. One of the things I’m interested in this year is something I don’t know I mentioned it to you yet is the idea of a

 

context graphs. Did we talk about that yet? don’t remember. Okay. So I just, do think that you’re going to see a lot more focus on context and, and the idea of understanding an organizational layer of what kind of what’s going on. Why do we make the decisions? Where’s that source of truth? And I think that’s going to be a big deal. I think you’re going to see a lot more of that. I’m starting to play with that somewhat. I find that

 

Scott Weiner (38:33.58)

I’m not that good at it yet. So I have some inefficiency problems. I’m trying to understand how to make it more efficient. But generally speaking, the idea that an organization can allow agents to understand how we make our decisions is going to allow it to be a lot more of a collaborator in our organization. You’ve heard of AI being treated like an employee. You should onboard your AI, your agents, you onboard people, give it context, give it, you know, what’s our mission, what’s our goals and all of that. That’s part of it too.

 

Let me see, is there anything else? Well, lots of stuff around cybersecurity. I really hope, I’m hoping, so this is not a prediction, this is hope. I really hope this is the year where people get serious about AI governance. That they really hear the risk of not getting people up and get those governance dashboards together. You know, I built a demo that I show people that I built with agents. It’s a governance dashboard and it basically showed all the different things you should think about in governance. And that’s the reason for it.

 

We can do a demo of it sometime, it, it always, every time I bring up, reminds me of all the things companies need to be thinking about that, that seemed to be on the back burner. anything else? me think. those are, those are some big ones. Look, there’s a lot of stuff that’s going to be coming. There’s, I, you know, the, the, world of synthetic data, the world of world models, that stuff’s coming. Definitely going to have more of that. Absolutely. and.

 

I think that the, thing I’m not sure of yet is several companies have signed up for different payment systems for AI to automate agent payment processing. But as you said, trust is a big, part of this. So what I’m not sure of yet is are they going to solve the trust problem so that people trust AI going out and doing the shopping for them? I was showing someone a demo the other day where I brought up a, an agent and I said, Hey,

 

go on Amazon and buy me a toothbrush, buy me the best toothbrush, is what I said. And while we were talking, the AI agent brought up a web browser and clicked on things and went out on the web browser and ordered it. I was like, whoa, whoa, whoa, stop. I just explaining it. I didn’t really want you to buy the toothbrush. And obviously it was an electric toothbrush, but it was expensive one. But that’s not how it’s gonna work. You’re not going to watch it go out on the web and click on buttons, right? That’s today.

 

Scott Weiner (40:56.96)

It’s going to just do it. It’s going to be instantaneous. It’s like, you know, negotiate the best price for my flight. And it’s going to go out and say, I got you the best price on the flight. Here it is. are people going to trust that? You know, I don’t, I mean, I know I won’t because I understand how this stuff works. And so I have a lot of questions, but the average person, I don’t know if they, if this isn’t what you do all the time, are you gonna, are you going to push that button and says, go negotiate the best price for me? And it tells you it did, and you just trust it. Is that how it works? I don’t know.

 

So that’s where I’m stuck.

 

even trust myself to make the right buying decisions. I’m definitely not. Something I want to bring up to you just briefly. I was listening to a podcast with Walter Isaacson. know who he is? The biographer and he’s done Ben Franklin and

 

Well, he did Steve Jobs when I was working there, so.

 

jobs. Elon Musk, he’s done, you know, Jennifer Doudna, lot of the big, names, people that have made a huge impact in our world. So anyway, he was he was talking about his newest book, and talking about the founders and how the founders believed in the commons. And the idea there is that people that can’t afford land, you know, back in Boston Commons, right? You’re in Boston, aren’t you?

 

Scott Weiner (42:15.438)

area. Yeah, the area.

 

So Boston Commons, the purpose of Boston Commons originally was for there to be public land and people that didn’t own land, they would have a place to go, a place to congregate. Even further back, that land was used for like, not in Boston, but the idea was it would be used for people that let their animals graze if they didn’t have their own land for grazing, things like that, right? So that idea eventually transformed and Franklin was a huge proponent of this and he was someone that really advanced this idea. It was this idea that people

 

need services in this country that can’t afford those services. so, the collective pays for those services. so that started with things like public libraries and with, night watchmen, which became police and, and fire departments and things like that. And fast forward, high school was something that came from the idea of the public commons, right? And it was because we were going through industrialization and people needed.

 

to have a higher level of education in order to be able to work in these manufacturing settings and they need to have basic skills. And so he posed this question that he wonders if maybe the next evolution of the commons is, is college because we’re going to be in this kind of AI world and people are going to need a higher level of sophistication and technology, acumen just to be able to function in this world, not to.

 

not to be a data scientist or to be an AI engineer or developer or anything like that, just the literally function in this environment. What do you think about the upskilling of society in order to be able to kind of deal with this evolution that we’re seeing?

 

Scott Weiner (43:56.878)

That’s a great question. So I actually talked to a lot of educators about this. as a, I, I taught an MBA class last semester in AI for, so it was for a business, business group. taught AI and a lot of what you just talked about is what the class was really, it was really teaching them kind of how to think about AI and what it is and all that. And I, I know while was teaching it, I really was like, everyone should learn this classic. Everyone should have this class. I wish I could give this to everybody and I can.

 

Maybe I will at some point, but that’s really where I’m going with this is that do we need schools to do that? I mean, we have, we have mechanisms to teach now. In fact, there are co courses. I’m was like Stanford put their whole AI course online. You can just go take it. Anyone can take it. Like there’s so much free information now. I think it’s, it’s less about the education system and more about our values. And do we value having everyone up skilled or not? And cause cause people, the reason people don’t do it is cause they don’t know they need to do it.

 

Right? majority of people, like I’ve talked to many people who say, well, I don’t really need that in my job, but you do. And you may need it because your job needs it and you may need it because you may not have a job if you don’t have it. So, but you don’t know that. And so you don’t go in and reach out and so on. So one is accessibility, knowing where it’s at, where is it available? And the other is motivation. Do you understand that you need this? And so could we easily as a society promote the need for it?

 

and articulate it in a way that people go, really should learn this. And then can we make it easy? teach at our local community library on a regular, not a regular basis, a frequent basis, I try to do it. I teach AI to anyone who wants to listen. And so I basically come in and all walks of life. I’ve got 13 year olds there and I’ve got 97 year old. Actually one of them, was 97, I think, the oldest person I’ve had. She was awesome, by the way. She was really phenomenal.

 

And I teach them AI. teach them different stuff. I teach them how to build little agents. I teach them about what a large language models are. I experiment with them and anyone can learn this stuff. It’s not, like you said, it’s not the data scientist. We’re not talking about that. That’s specialization. We’re talking about just general use and how to take advantage of the stuff. And I’m telling you, and I know you know this, but I’m just for the audience, that

 

Scott Weiner (46:20.364)

The difference between someone who understands how this works and why it works the way it does and all of that, and someone who doesn’t could be a hundredfold in their capability. I mean, it’s just, if you hired someone to work in your company and they were a hundred times better than the person sitting next to them, what does that mean? How would you treat them? How would you think about them? And you know, and so on. And anybody can do it. So why wouldn’t we? So yeah, I’m a big advocate of that.

 

So whether it’s free college or it’s, I’m not even sure if the academic system as we have it today is gonna be what we have in the future. I’m just not sure. It seems really antiquated to me when I see what AIs keep at bill. There’s a school system in Texas called Alpha School. They’re in other places too, but that’s where I learned about them. And anyone who’s really interested in this probably should look at what they’ve done. I don’t know how extensible what they have done is, but it’s fascinating because they’re trying to create customized education for kids.

 

And their results from what I’ve seen, I I don’t follow this enough to know how real they are, but the results are incredible. I mean, stories like one after the other of amazing things they’ve accomplished. Now they provide a lot of dedicated time and training and specialization. go off on trips. go kind it’s not an inexpensive program, but at the heart of it is they’ve got this idea of AI being able to create custom education based on your learning capability, your learning approach and so on. I think that’s where we’re going. I think.

 

We should and by the way when you look at how well we can create media now, right? You know Netflix has talked about this but the idea that you could be watching a movie one day and kind of almost choose your own adventure or have yourself in the in the adventure You know that kind of stuff. Well, what could we do with education if we could do that kind of stuff, right? The one I’m not thrilled about is they’re gonna start putting Placement ads within the movies, right? There’s no reason not to do that because they can so

 

after a long time, man, just probably not as elegantly as AI is gonna be.

 

Scott Weiner (48:15.34)

Yeah, but, specific to you. Right. Right. That that’s the, that’s the thing. So yeah, that, that I’m not thrilled about the one that has not come up, by the way, I don’t know if I guess I didn’t say it in our predictions that I’m, still looking for. I still want is these dynamic user interfaces. So it does exist out there. There are places, but it’s not commonplace. I’m dying to have this idea of you go to use an app and it kind of adjusts the whole user interface to how you want to work.

 

And it just kind of flexibly kind of figures out and learns with you and so on. So the learning aspect of it is, I think, very exciting. We could do a lot with that.

 

really interesting use case for marketing for that, right? Digital marketing and website experience, user experience. Um, you know, under we know so much about the preferences of the users that are coming to our websites and you can get an infinite amount of data on users. You can know who they are. You can know what their preferences are. You know what their interests are. I know what their behaviors are. If you could dynamically present the UI.

 

to be optimized for their preferences. mean, do it, and doing it for years in some way by having, know, variable ad placement, right? Based off of your preferences. But imagine that the whole website experience, the whole user experience could be redesigned on the fly based on who you are and what your preferences are. And that’s, that’s wild, but it’s also scary. mean, marketing is scary. What we’re able to do and how we can influence people.

 

It’s a little different on the B2B side, but on the B2C side, it’s, man, they just, know so much about you and they just, manipulate you in so many different ways. And who was it? Adam Smith had talked about the invisible hand of the markets, right? This is the invisible hand of marketing and it’s, it’s wild. You don’t even know what’s happening.

 

Scott Weiner (50:09.632)

No, you really don’t. the thing is, that’s the trust component of all this too. That’s buried in all of this is can we trust it? I don’t know. I’m having trouble with the trust thing still. We’re not there yet. I just, again, I can’t say this enough. It’s the speed of what’s happening. It’s not the change itself that’s scary to me. It’s the speed and the fact that people are treating it like it’s not going fast. And it really is. so…

 

I literally on a daily basis am learning a new tool or technology every single day, every single day. I don’t have a day that goes by that I’m not. So I can’t imagine being someone who’s standing still on this, how you can even have a sense of what’s going on around you. I don’t feel bad for them. I feel bad for me.

 

You’re the frog in the boiling water if you’re not engaged with this now.

 

Yeah, yeah, yeah, yeah. So anyway, it’s a lot of fun. It’s gonna be a great year. We got a lot going on. I will be interested to see if this cooling down thing has a real effect. Let’s definitely talk about that next year, whether it did or didn’t.

 

We’ll see. Before we go, Scott, just let me know how people can find you and how can they find your book?

 

Scott Weiner (51:18.126)

Oh, the book is on our website. So it’s newion.com N-E-U-E-O-N. It’s under our insights area. It’s called the pilot purgatory and part one and two are out. Part three is coming out next week. I’m putting out one chapter a week. It’s seven chapters. And yeah, I love feedback. If you do check it out, and talk to me. You can also find me at LinkedIn. I’m S-Y-N-E-R-O-N.

 

and I love to talk to people about AI anytime you know that, and, it’s always great talking to you about this stuff too.

 

Great talking to you. My takeaway is don’t trust AI and, uh, but use it. Yeah. Don’t trust it. Got it. That’s the mantra for 2026. Uh, thanks again for your time and expertise. Scott. Always a pleasure to talk to you. Thanks. You too.

 

But use it.

 

Scott Weiner (52:10.232)

YouTube, take care.