AI Adoption Is Surging but Finance Pros Are Falling Behind Due to Bad Data and Poor Strategy in 2026
In this first episode of Future Finance for 2026, hosts Paul Barnhurst and Glenn Hopper take time to reflect on how AI actually showed up in finance over the past year, and what that means going forward. Without a guest, the conversation focuses on real experiences, observations, and lessons from working directly with finance teams, CFOs, and operators who are navigating AI adoption day to day.
Paul and Glenn discuss how individuals have become far more comfortable using AI tools in their own work, while companies as a whole have moved much more slowly. Topics include ongoing data quality problems, hesitation around governance and security, and why many organizations still struggle to integrate AI into core systems and workflows. They also share their thoughts on notable developments from 2025, including OpenAI’s shift toward consumer use, Microsoft Copilot’s mixed results, Google Gemini’s rapid improvement, and Nvidia’s continued growth.
In this episode, you will discover:
How AI adoption differs between individuals and organizations
Why poor data quality still limits many finance teams
What recent changes from OpenAI, Microsoft, Google, and Nvidia suggest
Where Microsoft Copilot works well today and where it falls short
Why automation and basic app-building are becoming more important
Paul and Glenn share concrete examples from Excel, Outlook, reporting, and close processes. They also emphasize that banning AI use is no longer realistic and that clear guidelines matter more than strict restrictions.
Join hosts Glenn and Paul as they unravel the complexities of AI in finance.
Follow Glenn:
LinkedIn: https://www.linkedin.com/in/gbhopperiii
Follow Paul:
LinkedIn - https://www.linkedin.com/in/thefpandaguy
Follow QFlow.AI:
Website - https://bit.ly/4i1Ekjg
Future Finance is sponsored by QFlow.ai, the strategic finance platform solving the toughest part of planning and analysis: B2B revenue. Align sales, marketing, and finance, speed up decision-making, and lock in accountability with QFlow.ai.
Stay tuned for a deeper understanding of how AI is shaping the future of finance and what it means for businesses and individuals alike.
In Today’s Episode:
[02:47] – How Individuals vs. Companies Used AI in 2025
[06:12] – OpenAI, Monetization, and Market Signals
[09:27] – Google Gemini’s Turnaround
[15:56] – Big Models vs. Specialized Tools
[21:21] – Consumer AI and Platform Control
[26:26] – Real Copilot Use Cases in Excel
[35:28] – What Finance Professionals Should Focus on in 2026
[39:31] – What Finance Leaders Need to Do Now
[41:45] – Automation, Vibe Coding, and What’s Next
[43:07] – Final Thoughts and Closing
Full Show Transcript:
Host: Paul Barnhurst (00:00): Welcome to another episode of Future Finance. This is our first episode of 2026. I'm Paul Barnhurst, and I'm your co-host. Along with me for the Ride is Glenn Hopper. Glenn, how are you doing? First off, happy New Year
Co-Host: Glenn Hopper (00:57):
Then. So that's the good news, new Year, not a lot of new possibilities this year. I'm trying to stay optimistic. I've had some kind of cough crud for weeks now, and I'm kind of losing my mind. So apologies in advance to our listeners as I hack and weave, I'll try to mute so I don't hack and weave into their ears
Host: Paul Barnhurst (01:16):
Now. Is it losing your mind or losing your mind more?
Co-Host: Glenn Hopper (01:19):
Okay, fair point. Yeah, I'm certainly not getting any more sane. We'll say
Host: Paul Barnhurst (01:26):
I gave that boat up 20 years ago. I just owned the crazy train. Alright, our audience is like, wow, these guys are unique. No, so cannot believe it's 2026. So why don't we start, we want to just talk about the AI landscape. We don't have a guest for this episode. We know people have really enjoyed these when we talk about it and how CFOs and finance leaders should be thinking about ai. So why don't we start with a look back. I'd love to get your thoughts Glenn did 2025 ago, kind of as you expected for ai. Were there any big themes or surprises for you?
Co-Host: Glenn Hopper (02:03):
The interesting part, let's look at it on two levels. One, individuals in their use of AI in two companies and sort of official adoption. I think individuals, I'd love to see, I haven't seen the latest on what open AI's monthly users are, Geminis or Anthropics, but that's gone up. Maybe open AI because they were so big, maybe they've levelled out a little bit. But certainly with Gemini three, Google has levelled up and anthropic in the coding world. But here's what I think is interesting, and I'm seeing this in my trainings and in my consulting work as well, 20, 25, many, many more people, and I should have done a little bit of more current research, but many more people have started using AI and started to understand, oh, you can do more with generative AI than just summarise meeting notes or write emails. I think people have found, oh wow, you can actually work with code, work with data, work with numbers, and get a lot more efficient at what you used to do.
(03:14):
What has surprised me is that at the beginning of the year, at the beginning of 2025, we knew people had budgets to implement AI and to build projects and integrate it into systems. And this is anecdotal, but I've talked to a lot of people and at the company level, I don't think people are adopting it. So what you've got is individuals improving their productivity and I don't just mean efficiency, like they're working three hours a day instead of eight and then going and playing golf. I mean the quality of work that they're doing is better because you can go down a lot of rabbit holes that don't take anywhere near the time that it used to. So you're asking new and different questions, looking at data in different ways, and everybody's got this kind of personal thought partner. And on the company side and system side, I have seen a lot of applications where AP and AR are using AI a lot.
(04:13):
And certainly probably the biggest is fp and a narrative generation. I know in close and reconciliations we're seeing it, but a lot of times that's because the software providers themselves are incorporating AI into it. But I think as I talk to companies, they've still got the same issue that they did when we were talking about machine learning, we were talking about reporting. We deal with it in fp and a all the time. It's a data quality crisis. And I don't care if you're an SMB, A mid-cap company or an enterprise level company, the majority of companies don't have their data in order and that's a problem for ai. And then I think maybe integrating AI into existing systems and workflows is a little bit more complex than people realised. There's a governance issue around how many companies even have formal AI policies and are those even being adhered to?
(05:09):
There's still a trust deficit. I'd say the trust deficit is more maybe at the senior level than I think the worker folks who are actually having to do stuff have increased their trust. And the interesting thing though with all that is there's kind of this skill shortage because you can look all over LinkedIn and all over the internet and find all these people who call themselves AI experts. And I contend that a lot of these AI experts are, it would've been the equivalent of 20 years ago saying, I'm really good at Googling. Just because you're a good prompt writer doesn't mean you're an AI expert. So there is a skill shortage about knowing how to deploy these. And I don't know, it's an interesting time. I think in a lot of ways 2025 went the way way I thought it would, but I still think a lot of companies are behind on deploying AI kind of across the spectrum in a unified way where all employees are using it the same, if that makes sense. No,
Host: Paul Barnhurst (06:12):
It does. So I'll go a little different, I'll list three things I think were, I'd say somewhat surprising to me and I'm going to go more events I think, or they were signals I should say. So one was Sam Altman and chat, GPT saying reversing their view on porn was a sign they needed to make more money showing just how difficult it really is to scale a business because we all know, and I'm not going to get into the morals of this at all, porn sales, it's an industry you can make good margins at. And I thought that was really interesting. How quickly they reversed was just as a sign that shows how much cash they need to fund that business
Co-Host: Glenn Hopper (06:52):
And this is what them doing ai, I'm doing air quotes here, companions or what is they'll,
Host: Paul Barnhurst (06:58):
I don't know what it all includes. I think they're going to do companions allowing you to create certain images within limits and all those type of things. But you get into how do you control that? That's a whole other podcast that side of it. But I think that was a sign to me of just how many new avenues they need of revenue to survive. And that's crazy to think when you think how fast they're going to go, not that I'm saying they're going to go bankrupt because somebody would buy 'em before that ever happened. There's too much value there. Now is that value a trillion dollar IPO? I don't think so. I don't think you think so, right? We probably all think it's a little overvalued. Nobody's iPod for a trillion dollars and I don't see it happening in the next couple years. So I think there'll be a reset there.
(07:44):
So that was a sign to me. The second is I've seen a couple reports now. Co-pilot has been woefully short of its expectations on revenue, way less people using it. And even though there are some challenges with copilot, and I know we'll get into some of those, it still surprises me that they're that far short. I definitely find areas that it can provide benefit and the fact that it has the built-in advantage of integrating directly into office, you would've thought they figured it out better. Now is some of that on Microsoft? A hundred percent is some of that on just misreading the market? Sure. If you ask most people what are the best LM, if I ask you, has anyone put copilot first on their list of the best AI tools?
Co-Host: Glenn Hopper (08:30):
Never have I heard a single person say that.
Host: Paul Barnhurst (08:33):
There may be some people say there's things they like and they use it, but nobody puts it as their first tool. So that's one that's, I thought Microsoft would've figured it out a little better right now. Now they have a history of taking a while and then once they really figure it out, crushing everybody. So maybe 20, 26 that year. I mean, I'm not down on Microsoft thinking they're going to lose the battle per se, but disappointed in where they're at and what they've accomplished so far, I think don't, it's been well short of expectations. And then the third one was how quickly from, remember what we were saying about Google at the beginning of 2025 and Gemini and 24 and Bard and we had that list of the article. We read the 10 things that AI had said, and all of 'em were Googles that just horrible things like put glue on your pizza, eat rocks.
(09:27):
The benefits of human sacrifice was on the list. I mean just stuff where you're like, did a fifth grader check your tool before you released this To seeing many, many people saying Gemini is the best LLM out there right now. A lot of people really like it. I think it's great for marketing copy. There are other areas I don't think it's as good, but I use it if I try to write marketing copy, I use Gemini. And so to see how fast it's come in addition to all the other things they've done. So I think those are probably three big ones. And then obviously the other, I'll add one fourth one is just the growth of Nvidia, just watching them go from a big company to some argue that the biggest company in the world almost, right? I think that's been a big thing for 25. Any thoughts on any of those? Those are kind of my,
Co-Host: Glenn Hopper (10:16):
Yeah,
Host: Paul Barnhurst (10:17):
I probably different approach than you.
Co-Host: Glenn Hopper (10:19):
Yeah. So interestingly open AI's sort of migration towards a consumer tool. I posted about this or I wrote about it maybe in my substack, you
Host: Paul Barnhurst (10:32):
Posted on Lincoln about it too that you predicted they're going to seed the business and be the consumer tool.
Co-Host: Glenn Hopper (10:38):
Because if you think about the complaints that people have when they went from GPT-4 to five and they missed this sycophant sort of companion sort of way that it talked to 'em and they're going after think about 7 billion, however many people there are on the planet, they're going after that. Whereas I think that Google and certainly philanthropic, philanthropic may be more on the coding side, but even how well they work with,
Host: Paul Barnhurst (11:08):
Well their financial services, which I think we'll talk a little bit about.
Co-Host: Glenn Hopper (11:12):
But I think that OpenAI making that shift, it's funny because I might've guessed earlier on that would be the realm of meta and xai GR because of the nature of the data that they're using. But I think the models for both of those have been underwhelming. So I think chat GPT OpenAI may just end up being that people, their personal BFF digital companion and it's amazing that Google has come so far and that Microsoft hasn't, but I think if you think about Microsoft, I mean they're not trying at a large scale to build these frontier models. All their eggs were in the chat GPT basket, but then they put so many guardrails on it. I think that's the issue is Microsoft is being so careful they have the
Host: Paul Barnhurst (12:10):
Security require.
Co-Host: Glenn Hopper (12:11):
Yeah. So they don't want maybe the early problems they had around what were their early models, but with the hallucination there. So they did things like cap the number of back and forth that you can have in a conversation. And the fact, truthfully, you mentioned Claude Financial services and I'd love to hear your thoughts on copilot, but the fact that Claude for financial services Excel plugin was so much better than copilot when I did a bake off between the two several weeks ago. But Google now Google has the same thing, obviously not as embedded in the business as Microsoft is, but between Google sheets and the whole Google workspace and all that are in a pretty good position. So it'll be interesting to see what happens there. And the other thing on Nvidia I guess that I would say is we had another deep seek moment over the holidays with whatever latest model they've released. They're finding ways less compute, fewer parameters, ways to make the models smarter. And so there was an argument that this big capital buildup that we were going in the wrong direction because more compute isn't going to be the solution. I would argue that those two things are not mutually exclusive.
Host: Paul Barnhurst (13:29):
I agree with that. I was just going to,
Co-Host: Glenn Hopper (13:30):
It's kind of like Jay Bond's paradox, right?
Host: Paul Barnhurst (13:32):
It's not one or the other.
Co-Host: Glenn Hopper (13:34):
Yeah. Yeah. And I did, I wrote recently, this was for my book that's coming out in 2026, big thing AI ready CFO will be out later in the year, but I was writing in my book that even if we never went beyond LLMs as they are today, there are going to be engineering enhancements and changes that make the AI better and better and I think the new model version of Deeps and there's a paper that came out with it I haven't had a chance to read yet, but there's going to be ways to have improvements in training to have improvements in whether it's a mixture of experts or whatever and training, there's going to be engineering improvements where we are going to see incrementally better. And I don't mean tiny steps, I mean I still think there are a lot of big jumps before we even figure out how we're going to get so much more compute and so much more data and all that.
(14:28):
We're just in the nascent stages of this. So this is like, I mean I just keep going back to we're still in the prodigy and a OL era of AI right now and even if we don't have another massive breakthrough and a GI ends up being further off, the AI that we have is just going to get better and better and I suspect even by the end of 2026 we're not going to be at a GI, but we're going to see significant improvements in these models and more on agents and more on vibe coding and every, I know there's some other predictions around that, but I've been rambling for a few minutes so I'll stop there. Yeah,
Host: Paul Barnhurst (15:05):
So it sounds like you're saying you think there just, there's a lot of room to be made on the model side before we even have to ramp up the computer, the hardware, we're going to need that as well. It's not an either or and we'll make progress on both. We'll have periods where we kind of hit a wall I think, which is normal, but I don't think 2026 we're going to necessarily hit the wall yet. I think we'll continue to see improvements, but what's your take? So I think we'll get back to finance people here in a minute, and I know you on my take on copilot, but what's your take on this whole just huge LLMs versus specialised models, training models? Are we heading more and more to specialised in 26 or do you think it's a continuation of trying to find 5 billion more things to train the model on?
Co-Host: Glenn Hopper (15:56):
Yeah, so that's a weird, really the push, if you were to ask the leaders of any of the frontier model companies, they would say no, we're not leaning towards specialisation. We want to go even broader. We want to go from large language models to world models. So that's the part and that's Jan Koon left meta and his new company that I think he's trying to raise 500 million and God knows what the valuation is at that, but he has said for a couple of years now, LLMs are a dead end as far as the path to a GI goes, and instead of just models and now the models have evolved. They have text and audio and video and are being trained across the spectrum, but I think what Jan Lac is saying is to have a true world map model, you need that tactile feedback you need to have. So robotics comes into that, so picking up something and knowing how much it weighs, knowing intuitively how much pressure to put when you're picking up an egg versus a 45 pound weight, so they're going to push to that. When you talk about specialised models,
Host: Paul Barnhurst (17:09):
Do you think that's a lot of X AI's focus though too? I mean it feels like Musk is really all about the robots and the robotic side, so you need a lot of that information to make that work.
Co-Host: Glenn Hopper (17:20):
He would, but every demo we've seen from Musk is they'll demonstrate their robots and then find out after the demonstration, well they were actually being controlled by a person, a remote control or whatever the case is. So not so Musk is great at defining the art of the possible, but I don't know where they actually are with it, but absolutely he sees Tesla now as a robotics company.
Host: Paul Barnhurst (17:47):
Yeah, that's what I was going to say. As much as it is an automotive company.
Co-Host: Glenn Hopper (17:51):
Yeah, yeah. Now I do see one of my predictions for 2026 will be I could a hundred percent see XAI and Tesla merging.
Host: Paul Barnhurst (17:59):
Oh, that wouldn't surprise me at all about the solar company and although XAI is under a lot of pressure, I don't know if you saw the articles over the weekend, they allowing Twitter to be flooded with pictures of people. People are uploading pictures of somebody and asking 'em to turn it into a bikini and then they get a flesh coloured bikini and you can guess where this goes and smaller and it's processing many of those requests.
Co-Host: Glenn Hopper (18:24):
I didn't see that.
Host: Paul Barnhurst (18:25):
Not sure to what extent, but pretty far. And so they're taking a pretty hard publicity hit over that and rightfully so.
Co-Host: Glenn Hopper (18:34):
I guess on the small models, what I would say is I wonder if people are going to be who the people are going to be raising money to train their own small models or if it's going to be wrappers on top of pick your model and with system prompts and engineering, making them guiding them to be more experts in a certain area.
Host: Paul Barnhurst (19:02):
We've lot of that today, right? It's a question of does that continue? I mean so many companies, they're rappers where they've at least they've trained it, they've done some stuff on top of it.
Co-Host: Glenn Hopper (19:11):
I think it does continue because, and it might even get more useful and better because if I'm selecting options, what if there's a wrapper in my finance function, I need help with the monthly close and I've identified 10 or 15 different tasks that I want to look at automating and I've been trying to sort of build my own agent workflows and ways to do it, but there's some company out there that says, you know what, you can pick your model, you can use Gemini, you can use Claude, you can use Chat GPT, and we've got system prompts and methodologies that you don't have to be a coder, you don't have to be a super prompt engineer. This is just going to do it for you. It's going to plug right into your ERP and whatever. I do see value in that because how many companies have the expertise other than it's kind of crazy now and I would love to look back at the Bloomberg GPT and see if they've gotten any value out of that, the amount that they spent to train that model, which is specialised just on Bloomberg's data, how that performs compared to if you, and I understand you can't have infinite rag retrieval augmented generation, so there's something to the amount of data that Bloomberg had in training that, but I just don't know who can compete with the amount of data they have and to give it the sort of underlying IQ that it would need to be specialised.
(20:41):
I see wrappers more than new specialised models being trained. Okay. I do know
Host: Paul Barnhurst (20:47):
Campfire released its own, they called it large accounting model, so they released their own, I dunno how much they trained it. I know the detail. I dunno if it's using some other models. They've called it their own in the system if it's really just a wrapper, but it sounded like they had trained their own, so I thought that was interesting.
Co-Host: Glenn Hopper (21:03):
We should have them back on to talk about that because that is very
Host: Paul Barnhurst (21:06):
Interesting. So I'll have to do a little more research on it and we could definitely have 'em back on. It wouldn't be a problem at all. I know they would come on, we can get someone to help kind of develop that, but that's the way they build it at their conference a couple months ago in San Francisco is the world's first large accounting model I think is what they called it.
Co-Host: Glenn Hopper (21:21):
Really interesting. I'd love to know the specifics on how they got enough data to train one and I imagine it might be a lot of synthetic data truthfully, because you can, it's kind of like bootstrapping a statistical model, right?
Host: Paul Barnhurst (21:36):
Real quick, going back to chat, PT being a consumer and then I think we switch this a little toward finance. I'll share some of my thoughts on copilot, but I was going to say, when you mentioned the consumer, I think the path where they can really own the consumer space is who is going to be the LLM integrated into the operating system chat? PG signs something with Apple. Can they really own the Apple operating system? They're not going to own the Android because that's going to be Gemini. That's one of their big paths because if you own the tablet, if you really own the AI for iOS, that's huge and I think they have to own that as part of their consumer path, don't you think? If that's the path they end up going because they can't own Android, at least in the short term, I don't see any scenario where they could own Android even in the long run.
(22:26):
Ever feel like you go to market teams and finance speak different languages? This misalignment is a breeding ground for failure in pairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet QFlow.ai, the strategic finance platform purpose-built to solve the toughest part of planning and analysis of B2B revenue. Q flow quickly integrates key data from your go-to-market stack and accounting platform, then handles all the data prep and normalization. Under the hood, it automatically assembles your go-to-market stats, makes segmented scenario planning a breeze, and closes the planning loop. Create air-tight alignment, improve decision latency, and ensure accountability across the team.
Co-Host: Glenn Hopper (23:34):
Here's the interesting and this is Apple. You talked about Microsoft being delayed. Apple hasn't been a first mover since the iPhone. They wait and wait and wait. And so if you think about Apple ended up, remember when Apple Maps was just terrible and they basically just said, you know what? Google Maps is the default app here and then before Safari, whatever the default browser was that they had, but Apple right now is sitting on, I don't know if you took their total cash and marketable securities over a hundred billion dollars
Host: Paul Barnhurst (24:16):
If you look at their money on hand. And what's crazy is you do know they've bought back hundreds of billions over the last decade of stock. That's where they're still spending all that cash. Otherwise they'd be sitting on six, 700 billion I think.
Co-Host: Glenn Hopper (24:29):
Yeah, so obviously, and Apple's embarrassingly far behind on ai, but they're letting everyone else go out in this knife fight when hallucinations aren't sorted out and all the problems that they've had and they don't have the egg on their face that Microsoft got that Google got that even open AI got at times and they're just sitting back sitting on their pile of cash and at some point they're going to be able to swoop in and say, okay, now we can use ai. This is how we're going to deploy our money and I can see them taken off on the consumer side, they're more than likely to decide we'll just build their own once all those issues are worked out and dominate that. Or one of these companies, I guess Philanthropics looking at an IPO in 2026 open AI probably as well, but open AI may be a victim of too high evaluation and the pre-liquidity events.
(25:29):
So I don't know. It'll be interesting to say what happens there, but they could pick somebody up with, and I don't even mean, I don't mean the Acqui or Microsoft did with, I can't even remember the name of the company anymore, but I mean a legitimate buy someone or they could partner with someone and they talked about doing that with OpenAI, but to your point to be true consumer grade, they're not going to get into Android and Chrome and all that, but I think there's still, they've got their own browser. OpenAI does. I don't know. I could see them having a path on their own of getting to whatever the next generation equivalent is of the social media that came around and they could be the new meta whatever, new AI driven platform where you don't even have real friends. You just have a bunch of digital companions in flesh coloured bikinis, I guess keeping your company online.
Host: Paul Barnhurst (26:23):
Oh man, thanks for that image.
(26:26):
Although I did bring it up on myself, so I think we'll bring it back to what this means in 2026 for finance. But before we do that, you had asked a little bit, I know we talked a little bit beforehand my take on copilot because I know you haven't really liked copilot in Excel. I will first say I have not. I know you have to have the pro and I think it's beta below that or you have to have a certain on Claude. So I have not been able to try the Claude Financial services. I have tried lots of AI tools that are built on top of Claude and different things and some of them definitely do things better than Agent, but some of the things I've been impressed with this weekend I was playing with copilot a fair amount working on a course, we're doing some training for a company on copilot, so I had to get myself up to speed watching other people's videos, coming up with ideas and building things and there were a couple of things I thought were cool and where I think it's pretty good.
(27:14):
Complex formulas really good at helping you write those. I wanted to just show that I could write a formula that this thing had a lap big old long with all your variables declared. They had four lambdas inside of it and it was like this long, it took a little bit of time. It took me about half an hour to get it right, but it worked perfectly where it was taking a prepaid expense over 36 months and as one formula, having the headers five columns, the entire amortisation table all is one Excel formula. I wouldn't have written that on my own. I understand what it's doing now that I've seen it all. Oh yeah, that makes sense. And so I got there with a few little tweaks. It did one section of the code wrong and then I had to have it be iterative a couple of times.
(28:00):
I had to figure out one of 'em, it did a formula wrong. Another area the math was wrong and I had to just add a plus one and then I had to build, but 99% of it was done, so that was impressive to me. I thought I did a really good job there when I asked it to explain the formula, the answer wasn't very good. I would add a chunk it out, but I realised I could get there to really use AI to help me understand all the different pieces of it, but it would take multiple prompts when I try to do it as one, it's like, well, here, I'll just explain those two functions. No, there's a lot more going on here than telling me the general terms of those two functions. One thing I thought was cool that someone shared is, so I asked chat GT in Excel to summarise a list of all my emails for the last week with sender subject and action items and it was pretty impressive.
(28:51):
It was a good list. I hadn't thought of doing something like that, just bring it right back into Excel. So I like that. It's good. Copilot function is really good at categorising data. I had a list of 150 names where there should only be 15 names. You spell Delta 10 different ways. It's Delta Airlines at Delta Airlines Inc. And so I just had the master list and all I did is I wrote the copilot, take this messy list, bump it up against the master list and return the right one for each of them. It was a little wonky for a moment, it added an extra row in. So those are some of the copilot funky things where I was like, it's a little buggy, but once I played with it for a few minutes, I got it to work and a lot quicker than I'd be doing that manually or having to match all 10.
(29:36):
You know how we used to do it, you have a row of 10 of the correct name with 10 of the wrong name next to it and you do a lookup, right? We've all done that, but that's something I could had other tools do. It's not like that's unique to copilot, but I do like the integration. I do think it's getting a lot better. I do like the app skills where it can start to write things now again, but a little buggy, it's now on the browser. I still on the desktop, but I find the browser better because the desktop sometimes it doesn't work. I don't like that you have to be connected to OneDrive or SharePoint to use it. The OneDrive thing is a little annoying. I think everybody would prefer to be able to have file on your desktop and have it analyse it, but the reality is if it's what your company has, my take is you can be more effective with it than without it.
(30:25):
It has this data, cleaning it now has the formula completion. That's pretty good. Again, it's a little buggy. It works on one of my computers, not on the other. I've had conversations trying to get it fixed. Haven't been successful yet, but I think people can get use out of it. And if you're a Microsoft shop, just the integration in copilot, like when I need an image at being able to ask right in PowerPoint or different things. That being said, I do think it lags behind the other models in some ways. I think that's true of Excel's agent and I think, I know Microsoft's trying to figure out how deep do we go versus how broad do we go with agent because I think the challenge or copilot in Excel, the challenge gets into they get $30 for copilot across everything. If they go really deep with agent and they do a copilot for finance, which they had started to do, they had that preview, but they haven't gone deep there. They go really deep in fp and a or financial modelling. Are they going to charge more? Is it going to become usage based? They have to figure out what's the revenue model. They have the resources to do whatever they want. This isn't a cash issue, this isn't a resource issue. It's a business issue of trying to figure out what makes business sense to best maximise the value of the copilot asset. In my mind,
Co-Host: Glenn Hopper (31:45):
And I don't know, I'm not a marketing guy, but it seems like if I were in Microsoft shoes would my play here would be get it as functional and as amazing as possible. Get it into everybody's hands, get 'em hooked on it like a drug dealer and then tell 'em, okay, you like that. Now here's the cost for it. Just get it. Because that's what the advantage they have is that they're on everyone's desktop and I sound like a broken record with this, but a much maligned clippy. But that's the dream. And if they could, and I understand you have to have OneDrive and that's where we're saving your documents. But all these companies, if all your documents are in SharePoint or OneDrive and you have the layer and you can look at your team's messages and your outlook and interact with all your data across there, if they really find a way to go across the whole Microsoft ecosystem or if you're in Dynamics and just everything that you have is just automatically integrated into it, that's amazing. But that's such a huge ask. It seems like even if they took them out one at a time, go after PowerPoint first, go after word first, go after Excel first, whatever it is, nail that and get people thinking, oh, copilots great. And get them excited for whenever it's rolled into the next.
Host: Paul Barnhurst (33:10):
I mean, I think the two places that if they made it world class outlook where it automatically gave you your action items, gave you your summaries versus having to ask for it had a whole entire section built in and excel. I think if you had world class in those two areas, you'd have tonnes of interest. Yes, words used, yes. PowerPoints used. The PowerPoint would probably be next. I'd put word probably fourth on that of the core products of the office suite. It'd be interesting to watch.
Co-Host: Glenn Hopper (33:41):
Meanwhile, OpenAI has the plugins and so it's been a couple of months now. I turned over a client that I'd had for two and a half years. Everything was saved between SharePoint and all the emails on OneDrive. Integration with SharePoint wasn't that great, but all the files are there, so whatever. But two and a half years of emails, I did a deep research project with the Outlook plugin to chat GPT and said, go through the build a chronology of this client and talk about everything that was done, the year end projects, the issues we had and incorporate the SALs in there and everything. But in 25 or 30 minutes, a deep research report, not in Outlook, but in chat, GPT put together an amazing turnover report that was way more detailed than anything I would've done. Or if I did have to do it or if an admin or somebody else had to do it, it would've taken them days and this was 30 minutes the best sort of client turnover report that I'd ever seen.
(34:45):
So I mean that's what Microsoft envisions. If you go find all your chats, all your emails, all your files on it and being able to just interact across that, but we're not there and a lot of it is a limitation of the technology in how they have to interact or how their memory can be limited. You just couldn't go across that full universe, but you can kind of see the future that they're aiming for. It's just I think we're all just, and it's good news honestly, to hear that copilot is catching up and it is useful because there's so far to go with where they need to be for this end state. And I've given up on trying to predict when they're going to get there. But incremental progress I guess is good.
Host: Paul Barnhurst (35:28):
And I mean, like I said, I definitely think they've made progress. They still have a ways to go. So that's my take there. So let's bring this back. We've rambled for 30 plus minutes now. We've shared our 25. So things from 25 to 26, what does this all mean for the finance professional? What should the guy sitting at his desk be doing? Guy, girl, person? What should the CFO be doing? What's your thoughts? I mean this last year, especially when we first started podcast was like just experiment. I think we've moved a little bit past or I think for most people we've moved past the just experiment stage. Very few people, especially in finance, have not at least tried lms, have not used it for something in their work. Let's start with the average finance worker, corporate finance person. Let's ignore the leadership for the minute. Any advice on what you think they should be doing for 26?
Co-Host: Glenn Hopper (36:25):
They've seen firsthand rather how AI can be used in their jobs and they're figuring things out. I think for individuals it's going to be okay, I get it. I can use an LLM and it can help me in my work, but every time they go back to it, even if you build your own custom GPT or have a project where you have all your project information, it's still every time you do it, you're kind of reinventing the wheel. You're just coming back with that blank slate of a new chat or you have one really long chat that is getting confused because you have too much information and it's blowing out of the context window. I think for individuals who figured out how to use this well, they need to take an extra step and learn to use some of these automation tools like Zapier, make N eight N and all these and start, okay, these are the tasks that I do every month when I go through and do my close checklist or this is my quarterly report, whatever it is, start automating that and then figure out ways that you can enhance.
(37:30):
And then I think it's more this year, 2025 was proof of concept 2026 is don't wait on the software vendors to build it in. Don't wait on a one size fits all solution. Start building those workflows. And for my advice for the finance leadership would be get your head out of the sand. I was going to say get your head out of somewhere else, but get your head out of the sand and start finding ways. Encourage your people to use, understand that this is a shift. This is the introduction of spreadsheets, this is the introduction of the cloud, this is of the web, whatever. It's too late. You don't want to be a laggard because the businesses who do in 2026 kind of get over that fear and chasm and start adopting early are going to run laps around the laggards. This time the technology is just moving so fast.
Host: Paul Barnhurst (38:23):
Good advice there. I mean I think you made a good point of learning to automate when you said that. I'm like, I need to do more of that. Okay. How do you go beyond just using the chat or just using the tools to doing some automation, creating some customised GPTs or agents? I think it's time to probably move to that next level. The other thing I advise people is learn some. If you've just been doing it all on your own, maybe go out and read. What are some of the best practise methodologies for prompting? Like you, I'm not a huge fan of canned prompts. There are situations where they make sense and you can use them, but what happens when they don't work or they don't give you what you want or you got a unique use case. So I think really investing some time to maybe learn a little bit more or what are some of the things that really help you because wording can make such a difference in what you get back being specific and really learning how to build out that context. So I think those are two areas, like you said, what we'll wrap up here in another I think two, three minutes, right? About right about there. What about for the leadership? What advice would you give them?
Co-Host: Glenn Hopper (39:31):
Yeah, I think they get your head out of the sand. If you don't have a policy or if your policy is too strict or too draconian. Individuals are figuring this out. The people that have to do the work, if you're telling them they can't use it or they can only use copilot or whatever, then they're breaking the rules. They're going to go do this because if someone that they talk to on Reddit or some other forum is doing the exact thing they're doing in one 10th of the time, then they're not going to be the chumps that are doing things the old way and they're going to start figuring it out. And if you don't have policies and procedures and guardrails in place, they may end up doing something really stupid they may end up using like Manus is a great model. I don't use it for any sort of client data or anything, but it works really well. And if I didn't understand about what's happening with the data for Chinese owned model, then actually Manus had some kind of partnership they just announced with Meadow, which that's going to be interesting. But anyway, you don't want to be the rock and roll as a fad person. Like this is here, this is coming. It's making a significant impact. Now it's time for you to act and figure out how to deploy. It
Host: Paul Barnhurst (40:42):
Kind of wrapped up. And I agree with you by the way. I think for leadership it's have a governance policy take a stand, and that stand needs to be we accept that it's here, we're going to embrace it in whatever that means for your company. That may not mean a full embracement. There's different levels, there's different security things you have to think about. Every company's going to be unique, but you can't just say don't use it. That ship has sailed. And so I think that, and then for individuals, go to that next level. Either go deeper on the prompts, start to use N eight N and automate some of your workflow. Figure out how to do that. Always keep in mind security talk it as we always say. But I think the thing is it's just going to continue, right? I think 2026, we're going to see more and I don't expect any huge incremental. We're not getting a GI in 26. It's not going to be revolutionary, but I think we're going to continue to see it evolve. So any last thoughts before we call this an episode and stop boring our audience?
Co-Host: Glenn Hopper (41:45):
Yeah, the last one I would throw out, I guess think about vibe coding, think about things that you do. So I talked about those automations and those are going to get, there's going to be a blurring of vibe coding and building these automations. Any repetitive tasks that in the past, if you're like, man, if I could just write code, I could automate this tasks that I have to do every day or every week or every month. And I know there's CISOs out there and data and rightfully so worrying about the sort of citizen development happening. But again, that's why leadership needs to come up with a way to make this work. Because even if you can't write code, you can now vibe code little mini apps. You could have an army of these mini apps that for the price of a cursor or rept subscription every month manual things that you used to do, you just build an app, dump 'em in there and automate 'em. So automation's in vibe coding and companies get on board and individuals, I don't know, I kind of built my career on being a citizen developer. So I think that this is a time where more people than ever could do that. So people need to be thinking the same way and they can't be in the same sort of Luddite mode and we need to figure out a way to make this a reality because that's where the real productivity and efficiency and better work quality is going to come from.
Host: Paul Barnhurst (43:04):
I agree. Thank you for that. And you reminded me I need to do more on that next level of taking my skills even further. Right now it's on the vibe working side of Excel, but the automation is an area I need to tackle. So drop us a note. You can reach Glen or I on LinkedIn. You can find us out there. We'd love to hear from you what you'd like to hear in future episodes. We hope you have a great start to 2026 and we're excited to bring you many great guests throughout the year. So thank you for listening to us and until next time, so long go forth and prosper. Is that right?
Co-Host: Glenn Hopper (43:38):
Do it. Yeah, that's our new close. Alright,
Host: Paul Barnhurst (43:41):
Sweet. Go forth and prosper. Alright, thanks, Glenn. Get feeling better, and we'll see you soon.
Co-Host: Glenn Hopper (43:45):
Thanks Paul.
Host: Paul Barnhurst (43:47):
All right. Thanks for listening to the Future Finance Show. And thanks to our sponsor, QFlow.ai. If you enjoyed this episode, please leave a rating and review on your podcast platform of choice, and may your robot overlords be with you.