As we near the halfway point of 2026, it's clear AI has moved from a position of hype to a more practical space. Our team is tracking the progress and sharing our conversations on AI in a new audio series. Catch the first episode below and hear what the team has to say on a variety of issues relating to AI in Insights.
Bob: Alright, thank you to everybody tuning in. Let’s start with a quick introduction. I'm Bob Graff. I'm Vice President in the Operations Group at MarketVision. And we've got Ben DeSeingalt, our internal counsel and resident AI expert with us today. It's been a little while since we provided an update for the industry on our point of view. Ben has been very active in this space with different industry associations over the last year, specifically with some things changing. So, I wanted to get his perspective and look forward to what is already an eventful 2026.
Ben: Yeah, I'll just add it's been a while since we've done a video, but we've been beating the pavement on conferences. So, people probably have seen me give some of this spiel somewhere, somewhere around the country in the last year.
Bob: It's been a good year for that, but things are happening quickly. So, I just put together a few questions. I think we'll just walk through those and get your point of view on where we are. And we'll share this out. Hopefully it provides some context to others and other people can digest it and see if it's in line with what they're thinking, what they're experiencing, and what they think next year will bring.
Ben: Yeah, that sounds great.
Bob: It's obviously been an interesting year. You and I were talking about the last video update may have been a year ago, and that's a lot of— that's a long time in AI. A lot has happened. We've had a lot of meetings, as you know, internally and with clients and partners and there’s a lot to understand. AI has clearly, you know, won the year. It's having its moment. So how would you assess where we are now, either relative to a year ago or just how things have evolved versus where you thought we would be?
Ben: I think we're moving along slower than AI people originally predicted, right? I think a year ago, we were all talking like this was a transition that could be accomplished in 12 or 24 months, and you don't hear that very much anymore, right? You hear more of a, well, a bigger organization, you probably need 3 to 5 years, small organization, maybe 1 to 2. But I think timelines are getting more realistic. I think that for a long time, the narrative has really outpaced reality. But now we've reached a weird point where the narrative continues to outpace reality, but reality is also outpacing what sort of the average potential user of these tools feels about them. You know, we're at a point now where if all development stopped, if OpenAI and Anthropic and, and Google and Meta — if all the major players just stopped and say what you have is it, it's still enough for a lot of change in the business arena, probably enough for transformational change.
So, I think that some of the kind of marketing shine on things has started to go away, but fortunately it's allowed people to be a little bit more realistic and less just driven by optimism and sort of hope for what it might be. We've moved squarely into a “what can I do today” phase, which is healthy, don't get me wrong. But it is an odd cycle that we don't really see most of the time with business technology or general purpose technology where we've got hype running so far ahead, but then average perception lagging so far behind, which it's a strange phenomenon. But the good news is that no one is stopping, right? There's a lot of talk about whether we're in a bubble, and I think that in terms of infrastructure spend and in terms of the capital expenditures of the, the Nvidias of the world, we probably are, right? You can't spend indefinitely. At a certain point, if everyone has access to these tools, you can't get any more money.
But I don't think we're in a technological bubble where the models continue to improve every year. There are people saying they've plateaued, we’re going to be stuck here. And then they don't. They incrementally or transformationally improve. We just saw Gemini, which is the main model that we use at MarketVision, dramatically improved both in terms of its general capability and its ability to generate images with Nano Banana. We're starting to see that some of the ideas that we've had too were wrong, right? The last 6 months ago to maybe a year before that, there was a lot of talk about, well, there's no more data to train the models, so obviously the models aren't going to get better. And if we train the models with synthetic data, they'll collapse. And none of that has really come true. So, I think in terms of the macro environment, it's starting to settle. The major players have pretty well-defined roles now. Notably, I would say the role of Anthropic is to come up with new infrastructural concepts and then share them with everybody. The first of those was MCP, Model Context Protocol, which is how any AI tool connects to other platforms. So, if you connect your Claude or your ChatGPT to Google Drive, that connection is something that Anthropic just gave everyone for free. And when they do this, everyone seems to adopt it. More recently, they released skills. We could essentially train your instance of Claude to do something that it ordinarily wouldn't be able to do by creating a little Python package that it can call on. And just last week, OpenAI said, well, we're going to adopt that too. So, everyone's leapfrogging each other, but they're also helping each other leapfrog each other, which is a very weird sort of collaborative type of competition that we don't see a lot of the time.
Bob: It feels a little bit like the Wild West, but it feels like any new technology but on steroids in terms of how fast it's moving. It feels like they're defining themselves or differentiating themselves in the marketplace.
Ben: It is, but at the same time it's accelerating day to day. What the non-technical experts, so that the AI people who are not engineers, what we're hearing has started to coalesce. I attend a lot of webinars, I read a lot of articles, I watch a lot of videos to make sure that what I'm saying and how I'm thinking about it, that nobody has come up with something that I need to be incorporating or that changes the way I take a position on something. And the narrative in that world has really started to solidify where I'm not hearing a lot of rogue ideas anymore. Everyone is kind of on the same page, and there are some disputes about how a specific industry should react or how they should strategically move forward. In the legal world, there's a big debate right now over whether you should use the legal-specific tools. Are the tool providers eventually going to try and compete with you and become firms? But, aside from that, the kind of what is possible, what capabilities exist, what capabilities are going to exist soon, what should you be doing and how should you be doing it has really started to solidify where there's not a lot of variation, which In research, low variation wouldn't be a good thing, but in this case, I think a consistent story from every expert is much healthier than what we had a year or two years ago.
Bob: Yeah, and it is interesting. I think the beginning of the year was we were still in the hype phase, a big-time hype phase. Now, we'll get a little more practical. We're all using it in different ways. We're getting better for certain things, looking at opportunities for others, but It's been an interesting 12 months for sure. We know in our groups that we meet with regularly that some people are making more progress than others, or their solutions seem to go towards a more obvious use case. In terms of the use cases for next year or just moving forward, what's leading at this point? We talk about integration of tools or experimentation of solutions, but which ones have really kind of taken hold for the insights industry?
Ben: I'm going to start with saying what I hope 2026 looks like before what I think it actually will. I'm hoping that what we see is a transition from this sort of my role is X, therefore my use cases are A, B, and C, to a my personal pain points are X, Y, and Z, therefore I'm going to use the tools that I have access to in the following ways. Because not everyone has the same bottlenecks in their personal effort. A really easy example is that I'm a very fast editor, but I'm a very slow writer. So, if I, if I'm given something to react to, I know immediately what I don't like about it and I can change it. But if I'm starting there just sort of sitting in contemplation at my desk thinking about what to say, I'm going to be spinning my wheels for quite a while. There are plenty of other people with the opposite problem, right? They're really fast writers, but then they sit and deliberate and struggle over the editing for a really long time. And we're now at a point where neither is really a weakness. Neither one really needs to slow you down because we have a tool that you can say, hey, here's what I'm trying to get across. I feel like I'm not quite doing it. Can you make some recommendations? And I think we all wish that we had the time to do that with colleagues on everything that's important. Hey, can you give me a few thoughts on this before I send it out. But we don't have time, in part because everyone's got a lot on their plate, and B, because whoever you would send it to for advice, they've also got a lot on their plate. So just that alone, I think these very basic problems are still tremendously unsolved.
A lot of tech companies are talking about the baseline being everyone saves 10-20% of time, everyone is 20% more efficient or more productive. And well, can you give me examples of companies that have pulled that off and where it's quantifiably true that they've pulled it off? And then suddenly we don't have a whole lot of examples. But I think that's still, maybe not a low-hanging fruit, but that's still the sort of first stop on the road. And I really think that we should be making that stop before proceeding to more complete overhauls of process. A lot of this conversation is people saying, oh well, legacy processes are broken, we need to evolve them.
Some of them might be, but not all of them, right? Not everything that existed pre-AI is bad. And we don't need to overhaul every process that we use to operate a business. So, parsing through that and figuring out where the value actually is, I'm hoping that not just our industry, but that the broader world of business is going to hit that milestone and move past it in 2026. For us, I think in the insights industry, there's a lot of fear about this technology. I think it feels like a much more existential threat than it really is because when we're talking about the basic things that, that these tools can do— analysis, communication, business writing— are top of the list. And well, that's what we bring to the table as researchers, but it's not the only thing we bring to the table as researchers. And I think coming to terms with what the critical expenses of time are. I always use the infinite interns analogy, so if you had infinite interns, what would you hand off to them and what would you keep for yourself? And I think a lot of people are still struggling with that because it's not necessarily a department-level decision. It's more of a personal level decision. And that is part of the reason that I think that the timelines were so wrong at the beginning, and everyone thought this was going to be a pretty quick process, and now that's obviously been proven untrue.
Industry-specific, I think that we have more companies than we need, where there are a lot of tool providers who provide a very specific tool, and that tool often has been built to some degree with the AI that is then operating the tool. But you know, an increasing part of the conversation is that if you can build it with AI, I can also build it with AI. And even more than that, if you can build it with AI, the actual developers can see exactly what you want. And then build it not with AI. So, I think we're open to this sort of, anything you build can be my prototype kind of a mentality that I'm not sure how we're going to resolve that. I'm not sure how that's going to shake out, but it could be that small companies thrive, right? And we have 3-person tech platforms who have a large handful of really dedicated clients and that keeps them going, which is— that's a fine outcome. Or we see a lot of consolidation and collapse and things merge together and try to become more of a many-limbed entity that is more broadly appealing. I don't really know which one. I don't think either one is bad or good, but I think we're going to see some growing pains.
Be sure to follow us and hear more from our AI conversation.
