Is HMRC about to spot your clients' problems before you do?

[00:00:00] Hello and welcome to another episode of Digi-Tools In Accrual World, the go-to for accounting tech and all things accounting in our market today. We're very excited on today's episode because not only am I joined by the international man of mystery himself, Mr. 001 in my life, Mr. John Toon, but we're also joined by the wonderful Kendrick who is from Fishbowl and Fishbowl is one of our newest and latest and greatest supporters

[00:00:29] and is an inventory and manufacturing software for businesses that have outgrown spreadsheets and bolted on add-ons but aren't quite ready for a full ERP. So they've been working hard in this market but mainly in the US market for the last 25 years. They're also showing their muscle in the UK market with a desktop and cloud solution that works alongside QuickBooks and Xero. Kendrick, tell us more about you so that everyone can listen to your dulcet tones.

[00:00:56] I may not have the most charming accent kind of bounced around between Nevada and Utah all my life, but a pleasure to be here, Indi and John. Really appreciate the time. And for all those listening excited for a conversation today, as Indi said, Kendrick Hare, I'm the chief evangelist at Fishbowl where inventory and manufacturing, you know, nobody wakes up in the morning thinking I need inventory management manufacturing software. Typically things went sideways. And the thing I love most about Fishbowl is that it's designed to make it easy for businesses

[00:01:24] and for consultants, advisors, accountants to really get their arms around what's happening within the business and to see it scale and grow. Just yesterday, actually, I had a meeting with a company who they unfortunately shuttered their doors. You know, there was Shopify seller in the apparel industry and orders were just flowing through the door and they literally could not figure out how to get their inventory out in a proper way. And so that was two or three years ago and had the opportunity to meet with the founder at that time.

[00:01:53] And he's thinking about standing up the business again, but he knows he needs a better tool this round. And that's kind of where Fishbowl came through. I will say, as horrible as it may sound, I'm passionate about inventory management and good practice around manufacturing. It really does make the difference on that operational side. Margins are won or lost based on how you roll your stock. Yeah. And I think it's also great because on this episode or today's episode,

[00:02:17] we're going to cover some of the latest updates that include NetSuite and some of them launching in terms of the suite cloud developers. And so we'll come on to that. And we're also going to talk about some of the things to do with HMRC. And it would be great, Kendrick, to get your perspective on how that compares to the IRS in the US. And then there's a few other news stories that have come out from Sage. Digits and also a new launch for Campfire in London.

[00:02:46] So just catch up on this week. So we might as well get started. Are you struggling to attract the right talent? Are you needing to shift the busy work? Are you finding you don't have time to speak to clients? Trusted for over 20 years, the Advanced Track Outsourcing Model provides immediate access to accounting and tax experts who understand your local regulations and seamlessly integrate into your team.

[00:03:09] Digital transformation, robust security measures, top tier talent, no recruitment fees, no wait time. Focus on what truly matters, growing and improving your business and positively impacting the lives of your clients. Be more than just number crunchers. Be life changers. So I'm going to kick us off. And it was a big, clawed update this week. And I think this one has hit most people's feed. Anthropic has launched 10 ready-to-run agents for financial services this week.

[00:03:39] And the accounting pieces are fairly direct. So the sorts of things that a certain someone on this podcast has predicted, month-end clues, but also general ledger reconciliation, statement review, journal preparation and KYC screening. So they're very, you know, they're kind of specific in what they're trying to claim to do. They're aimed at finance operations. Anthropic says the agents can be run as plugins in clawed co-work and clawed code, or as managed agents with audited logs and all the sorts of things that you'd expect,

[00:04:08] like permissions, credential vaults and long-running sessions. So Claude is also saying that it's now working across Excel, PowerPoint and Word. And Outlook is also coming soon. So what does that mean? Does that put Claude ahead in, you know, the race in terms of what they're doing? I'm sort of guessing and speculating that yes, at the moment, Anthropic's moving a lot quicker than most of the other LLM providers in packaging finance work into ready-made agent templates. And maybe some of the others are still strong for broader analysis

[00:04:37] and, you know, things that might be more applicable to like, you know, native products, like things like Microsoft, for instance, and its finance agents in the advantages it has in Dynamics and Microsoft 365 Workflow. And God knows where Google is at the moment. Let's not discount them because I haven't looked at it. But everyone, should everyone switch to Claude tomorrow? I don't think so. I do think it's really impressive. But it's not the same as a proven month-end close engine,

[00:05:06] which is something that we've seen has been, you know, that we've seen a lot of in the UK market. John can probably name check a few of the specific usual suspects that we've been talking about. But reconciliations and the close are not just the sort of language problems. They need the data, the deterministic controls, some of the things around the thresholds, all the evidence, et cetera, et cetera. Perspective I have on this, I want to ask you about on this whole thing about like human in the loop

[00:05:34] and where do we really draw the line here? Because we're getting so much automation bias that is it really just a human in the loop or is it just the human is the one that's culpable and responsible? The signature is basically that's who you sue if everything goes wrong. But yeah, what do we think about this? I mean, I think this kind of human in the loop message that we sell ourselves is really just to make ourselves feel better about the fact that technology is about to replace us. You know, I haven't tested any of these new agents that Anthropic have launched.

[00:06:04] I mean, I guess fundamentally, you know, we've seen over the last probably 18 months to two years as lots of people, lots of commentators have used different models, whether it's Claude, whether it's ChatGPT, et cetera, to kind of demonstrate that things like bank reconciliation, things like dealing with messy, unmatched receipts or payments, helping with some of this month-end close kind of concept is doable, right? And it's doable in a way that's generally quite reliable. It's usually fairly accurate and actually most times quicker than the human being

[00:06:33] trying to process it themselves. And I guess we had a bit of fair warning of this from Anthropic as well because they had this big push, didn't they, to recruit, I think, was it 150 finance people to come into the organization to help them to build these kind of concepts and these agents and stuff. So for me, the launch of these agents is really just a wrapper around making some of the capabilities of these models more accessible. And we've also seen this fairly recently with ChatGPT in terms of then launching at a squivalent. And you mentioned Microsoft.

[00:07:02] Your Microsoft have gone down the route of bringing Claude's capabilities into the Microsoft chat model and not just rely wholly on OpenAI. So I think we'll just start to see more and more of this. What's difficult to understand, and it's partly because I haven't really played around with these, is where do they fit in the workflow in terms of things like Mayday and Nemo and there's a whole bunch of other kind of tools that help with lots of these things. Do they sit adjacent to them and effectively compete with them,

[00:07:29] or do they sit kind of like on top of them in terms of using them as a platform to kind of leverage off the back of them? I think that's probably what's unclear to us at the moment because these tools rely on APIs and MCPs to kind of get into the data and access these processes. And it may well be that it could be a combination of both. These could both compete with but also leverage these platforms at the same time. Yeah, so interesting to think about where things are headed next. And while it's exciting to see the agents coming out, I think we as humans have a way of always trying to get in the middle of it.

[00:07:59] So while they can do it better, faster, and more efficient, the combination of AI and HI in my mind is something that we're going to try to prolong as long as possible. We haven't seen AI develop consciousness. And so having folks in the middle of that I think is always something that at the end of the day there's going to be some oversight that people want to be involved and see how the data flows. Even if it's done for you tasks, there's a review process there.

[00:08:27] The exciting thing to me is to see everywhere that it can go and the connectivity that occurs with that. John, to your point, though, the connectivity is reliant upon that data set, upon those APIs, upon the MCP, and who knows what the next chapter will bring. Having that good data set is really what drives it all. One thing that I found really interesting about this data and underlying data was also in the same breath that they said that they have actually got collaborations

[00:08:53] with Third Bridge and GuidePoint, which for anyone that recalls of it, or anyone who gets these sorts of calls on repeat for the last decade, we've basically trained the AI by giving all of that knowledge away to these people that will call you up and give you 100 quid for something. I blame all of you. You know, that's so said. Oh, they're taking our jobs. Well, you kind of trained it yourself. Yeah, maybe, maybe. I think maybe what's quite interesting from the financial agent's perspective

[00:09:20] that Anthropic have launched, and I think going back to maybe we've trained them too well, is that they are talking about having some kind of verifiable audit trail to kind of validate what those models have done. You know, that's one of the big challenges that we're seeing from regulators all around the world in terms of sort of saying, look, you know, we've got this kind of black box technology available to us now, but fundamentally, we don't really understand how it's working, how it's coming to kind of some of these results or conclusions.

[00:09:49] You know, the regulators are now sort of saying, well, we want you to understand how that works, which is impossible, frankly, for the accounting industry to do. We're never going to be in a position to understand how these models work. And when, you know, ChatGPT is getting obsessed with goblins all of a sudden, just because of some random quirk of the code or the learning, how are we ever, you know, when the creators of these own models, in some respects, don't understand how they've come into kind of some of these conclusions and some of these outputs, how the regulators expect the accounting industry to understand that is completely unknown to me. It's nonsense, but that's the world that we live in at the moment.

[00:10:19] Hasn't that always been the case, though? The regulators are chasing things that maybe don't have a logical response associated to them? Yeah, you could probably argue that is the case. Yeah, for sure. I'm sure we could talk about this for ages, but let's move on to something else as well. And we'll maybe if we get a chance, Kendrick, we can talk about how daft the regulators are sometimes. But so we've got some news that indeed sort of referenced that Campfire have now launched in the UK, officially launched in the UK, I should say. And this came from John Glasgow, who's the CEO there.

[00:10:47] And CEO and CFO, he's doubling up. He's got two jobs for the price of one, which is good work. Campfire, for anyone that doesn't really know, very much an American product and it's kind of like formation and creation. Big, good ERP solution, really challenging the market now in terms of what they're doing. And they've had the clients, customers here in the UK for a while now, which is kind of inevitable because SaaS products are kind of universally accessible, right? So what they've worked on doing is building out some of the UK-specific requirements.

[00:11:17] So VAT is probably the biggest one because you're being able to kind of manage that effectively. And the product is going to save you a lot of headache here in the UK. And a lot of that has been driven by the fact that you've got US businesses that do have UK subsidiaries over here and they're kind of naturally gravitated to having a single platform solution, which is kind of what you expect. So yeah, so their first office was opened in London a few weeks ago. They're building out a team. It's good to see people having your boots on the ground. I think that really supports a good rollout.

[00:11:45] And I suspect you may well see them at quite a few of the big shows over the next few weeks, which unfortunately I'm missing account texts, but they may well be there. You are missing account texts. We are sad about that. And actually we had quite a few people already, you know, missing your presence. So Nick Levine's lunch, GBX, which was held by Advanced Track. And we're missing a big John-shaped hole in everything. But yeah, great news to hear. John, get us a cardboard cutout and we'll put you at the fishbowl booth for account texts. Everybody can come by and say hello.

[00:12:14] Yeah, everyone who knows John, right? He's a little bit there. He'd just be stolen and be placed all over the room. Or it might be the opposite, to be fair, Kendrick. People might run away if they saw it. They'd be like, oh no, we're not going anywhere near those guys. It'd be like the village John. That would be it, right? Kind of like a village bike, but just a John bike. Love it. Comparing over account texts and back to the U.S., you know, we don't have the bikes running around. We've got Where's Waldo to chase John. And it's kind of fun.

[00:12:43] Yeah, I was interested in one of the stories that came through. Really pretty exciting. HMRC is really focused on being the world's most AI-enabled tax authority. And such a stark contrast from what we see in the U.S. market with the IRS. Leaning in and rolling out 28,000 copilot licenses to existing staff and then plan to reach 50,000. It's a phenomenal mark. And in my mind, it just shows the difference in mindset of really leveraging those tools.

[00:13:12] And, you know, the last couple of stories we've hit on are all about efficiency and gains and trying to make it better. Kind of neat to see a regulated body like HMRC that's leaning into that and really driving for change and leveraging better tools. They've already delivered over the past £8 billion in benefit, helping close the gap.

[00:13:32] From an outsider's perspective looking in, we in the U.S. really started to take note on what was happening with the making tax digital push a few years back. You compare the just overall programmatic difference in how things are being approached. And, you know, to be fair, in the U.S., the IRS, they do leverage AI, but it's pretty limited. There are only certain places that they can employ it. And more for data analysis, fact check when an audit's occurring.

[00:14:01] So to see how the U.K. is really rolling that out and the model behind it, not just for their own purposes, but to help the taxpayers to embed within the tax authority itself. And, you know, I'm interested in where the future will go with rolling that out to other agencies and regulatory bodies, other use cases, and what that may look like for the global economy as it rolls forward.

[00:14:28] You know, the U.K. is clearly leading the way in trying to get that rolled out. And it may cause some waves back and forth, though, right? It may, but I don't know if we're leading the way. I'd say one thing I'm quite curious about is, and is probably just, again, is if HMRC become a lot better at spotting the gaps and the anomalies and all of the bits that, you know, they need to be picked up on,

[00:14:53] then, you know, we kind of saw with MTD it's going to speed up the cycle of how you have to report on the data more frequently. An accountant, you can't, as an accountant, you can't afford to wait to reconcile. You can't afford to then go, oh, yeah, yeah, we'll catch that at the, you know, year end. It's not like you can hold on. So, you know, I think especially if HMRC becomes a lot better at that, then what does it mean for the accountant?

[00:15:19] In particular, those ones that are, I don't know, managing landlords, sole traders, payroll clients, anyone that's sort of like, you know, exposed a little bit more with how their data is seen. And, you know, the risk as well is that HMRC's it before your accountant even sees it. I feel a little bit like, wow. Honestly, it does feel like Big Brother is watching. My mind goes to seeing Amazon sellers and, you know, the experience they have.

[00:15:46] And I know that's an odd comparison, but at the heart of it, you know, when Amazon's looking at trend data and trying to decide what they're going to stock internally, what does that do? Is there some competition that comes in through state-run opportunities or a private-public collaboration that ends up forcing major market shift? Not to mention the data security piece of it, right? You know, there's sensitive information that's being controlled there. I think it's always nice to have a third-party view on HMRC.

[00:16:12] Because I think when we're in the UK, we're always probably, you know, a little bit like anybody, right? Probably self-critical of the things that we have at our fingertips. And, you know, an HMRC, just like the IRS, has plenty of flaws, I'm sure. But, you know, actually to be sort of told that they are leading the way, which they're definitely trying to achieve, it's really fascinating to see. And I think this utilization of Copilot, you know, within the, you know, is in part probably down to the fact that there's been a bit of a brain drain, right?

[00:16:39] You know, we've had quite a lot of people, you know, fall out of HMRC as a consequence of a mixture of things. Some rationalization of people, you know, just natural kind of development and stuff. And, you know, some other things. It's good to see it. It poses the same risks in terms of, you know, just in the same way that the general public or accountants using AI, you know, we don't want revenue inspectors coming back to us and saying, you've not complied with this rule or this piece of legislation.

[00:17:07] And then find that it's been made up by OpenAI or by ChatGPT or whatever. So that's the risk that potentially we pose. But hopefully they've got a filter to kind of manage that risk. Yeah, actually, I've got a story on that a bit later, but I agree with you. I think what it's going to do is cause a ricochet, a ripple effect that everyone has to level up, especially with the technologies that they're using. And possibly if you're an accountant that isn't already considering how you can use similar types of technologies,

[00:17:34] it might accelerate that kind of learning curve for the tools that then do spot the risks and help you in your own anomaly detection. So that's good. I've got something from Sage this week, and I think we've got quite a few Sage updates. So Sage's platform strategy gets an AI boot here. So they've announced a new developer tools and commercial model across Sage Intact, Sage X3 and Sage Active.

[00:18:03] A few years ago, even on this podcast, I'd say that we were part of that first wave of people that were looking at Sage and were saying, they're too big. Where's the innovation going to come from? They've got a huge customer base, strong heritage, but they're very slow. I think the news articles at the time was that growth had to come through acquisition. But this update suggests something altogether different. So Sage is trying to make the platform itself more expandable, and it's giving developers a more unified way to build across Sage products.

[00:18:31] It's launched Sage Agent Builder and an AI gateway so partners can create AI-powered workflows plugged directly into Sage Copilot and Sage Marketplace. I think that's huge because, you know, the future of, I guess, this software and many others won't be one vendor that is building every feature. And we've seen that for years as edge cases, but it'll be core platforms plus specialist partners, perhaps even accounting firms that build an agent for specific industries, workflows,

[00:19:00] and different reasons that, you know, they want to solve for. So Sage is also introducing usage-based pricing and revenue sharing models, and that tells you how they're looking at this market as a huge opportunity for how they can actually partner, make money, and almost create new revenue streams for themselves and partners within their AI, and we're going to call it an ecosystem, and let's hope that it stays an ecosystem. So the key phrase that we've got to pay attention to is governed innovation.

[00:19:30] Governed innovation. So Sage knows finance, payroll, and operations cannot be a free-for-all. So alongside the AI toolkit, it's talking about a sandbox environment, certification, stronger identity, and Sage Verify. It's the advantages. It's not the flashiest vendor, but it's the trusted. It's the heritage. It's the installed base. It's the workflow debt, and it really turned a corner, especially even in the last few years of this podcast.

[00:19:59] We've seen them really turn what may have been seen as legacy weight into a genuine platform advantage. What do we think about that, John? Because you gave, in the early days, such a roasting to Sage, such a grilling, that I'm quite curious about how you see that now. And in terms of if Sage partners and starts building AI agents directly into its workflows, would you trust those tools more because they sit inside Sage?

[00:20:25] Or would you still treat them like any other third-party app that's in the market that we've seen come and go for the last however many years? So I think building agents into the platform is just a natural extension of the ecosystem that we see with other products, right? And Sage have had this themselves with Intact. When they acquired Intact, you know, had a huge ecosystem of third-party players that could connect into it. And obviously, you know, Xero and others have kind of like led on that as well.

[00:20:51] I think, you know, it's an interesting model to allow other people to kind of build that up, you know, from kind of anywhere. But like I say, probably a natural extension of this. And we saw that a little bit, you know, Silverfin had this, right? In the early days when they were, you know, when they were expanding was that you could build out kind of like a functionality for a specific area of specialism or a specific marketplace. And then effectively sell it on their platform and sort of trade it on their platform. So it's not new.

[00:21:19] I think, you know, if I was going to be critical of Sage, building agents within your own products, great. But actually, there are other players out there, you know, in the software market and the accounting and audit software market. They're actually not just looking at building agents for their own products, but building agents for the marketplace, which will then potentially plug into other third-party products. And for me, that's kind of like that's where we should be focusing. You know, we talked already about the fact that having an agent in a product, you know, is great if it's effective and it improves your workflow and efficiency.

[00:21:47] But actually, you know, the real challenge for accounting practices over the next few weeks, months, years is going to be like agent orchestration across products and platforms. And how do you manage that? You know, in the same way that we, you know, you want to manage your human capital, your human team. You know, you don't want someone who sits within one product and says, I only ever do this. You want them to be multifaceted and be able to work across multiple products and multiple processes.

[00:22:13] And so that's the kind of thinking that we need in this agent space. Well, I think too many software providers are too narrow-minded in what that tech stack looks like, even for a small business, right? Especially as you go midstream and upmarket, even an all-in-one solution isn't going to handle what's happening on the e-commerce front and CRM solutions and accounting and business operations all together. So, you know, to have that siloed approach can really be limiting. And it just creates barriers around.

[00:22:41] And there's so much conversation going on right now about data modes and where things sit. But the reality is that that data is becoming a commodity. And the value in so many ways is just having the right data at the right spot so that across the platform users can get what they want. And if you look at it from the lens of a business owner, they don't care where it lives. They care that the job gets done right and that it's efficient and that they're making margin and that, you know, the business is succeeding. And putting those walls up really creates a lot of friction, right?

[00:23:09] I was at a meeting with Steve Herr, who's the CEO of Sage, you know, last week. And he was talking about the fact that they had to bring in usage-based pricing, particularly on the Intact side, because people were absolutely hammering the API and using agents basically to log into Intact and undertake processes. And we're just burning through like compute cost and, you know, all sorts of other things, which was costing Sage, you know, basically tons of money. So they're having to recruit that, which makes sense.

[00:23:35] But I do think that every vendor that's looking at this needs to think of, you know, passing on a unit cost is really simple, right? You know, Kendrick will tell you in the inventory space, it's really easy to have a widget, you know, know that it costs you £50 or $50 and to sell it for £100 or $100 and you get actually a margin. But I think when we're talking about compute and utilization, it does not like a physical item and it's really difficult for an end customer to kind of understand, am I getting value from it for this? Will I be using it in the right way?

[00:24:05] And is that effective for me? You know, the Sage conversation is interesting, John. Sage Future, the company announced expanding their AI agents across so many verticals, finance, HR, payroll, operations, and all of that automation being embedded directly while it may be a little siloed. The key phrase they focused in on there is act on work and not just analyze it. You know, moving from that model of just showing you the data to being able to do things and create value, I think is really cool.

[00:24:32] And at least in the U.S. market, it's created a new level of excitement. So preparing payment reminders and moving tasks forward, flagging on operational issues and supporting payroll and workforce workflows so that teams can really manage exceptions and see what's happening inside that system.

[00:24:52] You know, the context here is interesting because Sage has such a deep UK rooted base, but they're focusing heavily on the global enterprise side, right? If I can be candid in a little bit of the same scenario as what Fishbowl has been through, where 25 year old company and this whole conversation around, you know, bolting on versus building out the AI component is certainly relevant. You have to reimagine how AI is rolled into things if you don't want just a cloud integration.

[00:25:22] And so it really fundamentally changes on the approach that they take. And I think Sage is really focusing on getting it right the first time. You know, in fact, they, you know, their CTO, Aaron Harris mentioned that almost right isn't good enough. And I think every accountant would agree wholeheartedly. Sage is calling it the class box approach so that it's explainable, audible and logged.

[00:25:43] You know, having that full audit trail is really what gives them peace of mind so that users can see what the data is, you know, where the logics come from, how assumptions are being made, and then ultimately driving to those recommendations. So there's a record of what's suggested and then, you know, still having a human in the loop so it can be approved and then followed up on. Indeed, to your point, you know, the traceability may lead back to liability and that remains to be seen. But, you know, certainly some good development there.

[00:26:12] I like that quote, traceability leads to liability. That's not good. I might steal that, Kendrick, for a presentation next week. Okay, well, I'll do one last little bit on Sage and their AI stuff because, you know, as Kendrick's kind of alluded to, you know, they're really leaning on this. So like many, you know, many vendors in the space. But they made a strategic acquisition of a product called Doyen AI, North America-based products, specifically kind of like looking at data transfer.

[00:26:40] So, you know, I guess for a lot of accountants that have been working in kind of like the small business space, you know, you'd be very familiar with a number of platforms that enable, you know, or facilitate this kind of idea of moving platforms, particularly on the accounting side, less so on the payroll and less on the HR side. But Doyen, you know, it's a pretty new business. It was only founded in 2024. So Sage had been pretty, pretty, you know, snappy on this and tended to picking it up and bringing that into the fold.

[00:27:07] But it is there to help Sage, you know, continue that kind of, you know, that growth objective for them in terms of making it easy to move. And so where Doyen specifically works is around financial data migration and validation helps with that kind of mapping of, you know, accounts and moving things like dimensions and account groupings across between products,

[00:27:29] which particularly, I guess, if, you know, we're talking about, you know, Ntact and X3, you know, that is kind of like where the big complexity comes with those products as a consequence of kind of like the shape and size of organizations they work with. Having said that, I do also think that potentially there's a really interesting opportunity here with Sage, you know, in terms of the practice tools. You know, if you look at kind of like working paper solutions, you know, one of the big things that they're looking to try to do is like, you know, when you're taking a GL, you know, whether it's a third party GL or it's your own, you know, Sage GL for example,

[00:27:58] you bring that into the working paper solution. There's a whole bunch of mapping and kind of logic checks that need to go there. And then when that comes through to the other side to financial statement prep, you know, which is very big in the UK, not so much of a thing in the US. And, you know, that is an ongoing challenge to kind of like have that flow between those products, you know, and see, you know, here's my originating data. Here's kind of like the accountant or bookkeeper kind of amended data. And then here's my financial regulatory data that's going to be published.

[00:28:24] And so I think there's potentially some great opportunity for them to utilize this product to do more of that as well. Very nice. I think that's a massive green week for us. Big green machine is Sage. I'm going to move on from Sage to NetSuite as well now. And since we have Kendrick on this episode, and so I'm just being a little bit selfish, we want to be able to talk and get another perspective on this.

[00:28:45] NetSuite has announced new suite cloud agent skills, which is essentially knowledge packages that help AI coding assistants understand how to build and customize inside NetSuite properly. It's a useful, perhaps, you know, a bit of a signal for where ERP and accounting software development is heading. So NetSuite isn't saying AI will magically build your whole ERP.

[00:29:05] It's saying if developers are already using coding agents, let's make sure those agents understand NetSuite's rules, permissions, UI patterns, security expectations and scripting standards properly. And so it's really about ERP customization. It's high risk work. And that means obviously like anything like a bad app or poor permission or a model that's broken, the suite script deployment doesn't just create a bug. It can affect things like, you know, all sorts of other things within the application.

[00:29:33] So finance, inventory approvals, reporting and controls. NetSuite is trying to make AI assisted development faster but safer. And the skills cover things like UI framework references, 684 permission codes, suite script, field IDs, document generation, security guidance and converting some of the older suite script 1.0 scripts to 2.1.

[00:29:57] The bigger picture is that the, I guess the, they're not just adding AI for the end users. They're adding it for developers and partners who can then continue to extend their platform. So really going along the same theme here, which is like, can we then customize as much as we possibly can using that, using third parties and shorten customization cycles from days to hours while reducing the rework and any security mistakes.

[00:30:24] And partners can build much more vertically and industry specific solutions a bit faster. And some, some thoughts around that. I'm really curious to see and hear how you think about that, Kendrick, because obviously as someone that is like in ERP world and is, is sort of raging mad about that. I guess development in, you know, in that sense can be dangerous if the AI produces plausible but insecure code or guardrails that aren't properly orchestrated.

[00:30:55] So what are your thoughts on that? Yeah, that's a really interesting and broad topic. And like we were touching on a few minutes ago, this idea that the developers of even these AI tools are potentially being held liable is crazy to think because of all that the agents can and will be doing. You know, first off, it's exciting from a development perspective to have those tools available and for NetSuite to be thinking more broadly about how and where things can be employed. It's not just that end user flow, but everything within the ecosystem.

[00:31:25] That to me is a brilliant approach. I think they're doing, they're doing some things right there, but there's not a defined set of rules on, on how AI interacts and where developers can leverage other platforms and what the data set looks like. I think, you know, as, as we look forward in the next 18 to 24 months, we're going to start to see a lot more conversation around how and where that fits in and then how it's monetized as well.

[00:31:46] Right. Is it the platform responsibility? Is it the end user who's subscribing to those things? Is it the vendor partner that's connecting in as a combination of those things? Is it usage based? Is it going to be ongoing? I think there's so many open questions and you see a new model come up. It seems like every other week if somebody's trying something, but I don't know how many of those are going to last yet. There's so much exploration. We're going to continue to see shifts.

[00:32:11] So building off of NetSuite, more MCP, we've got one around digits and interesting to see where digits is taking this. They really launched that MCP server and coupled that with the agentic ledger, meaning that firms and business owners can connect live digits data directly to tools like Claude, Chat, GPT and Cursor, other, other solutions. And that to me is interesting because for a little while, it almost seems like MCP was fading out as fast as it came in.

[00:32:41] And the data is there, but how and where it gets used, digits is really showing that value. There's so many times when accounting firms have had to bolt on extra apps and when the integration didn't work exactly like you wanted to, the accountant became the integration layer. And so much work going into exporting data, analyzing it, cleaning it up, reformatting it. Anytime there are multiple systems, now you're normalizing the data between however many solutions may be out there. Just a heap of work.

[00:33:09] But to see digits rolling out that MCP, it really removes that layer. And I think we're seeing this with a lot of solutions out there, but makes it so much easier. The caveat there with MCP, the AI tool is querying the ledger, but it's on a relatively basis. So it can't damage the books, the pitches, the digits is already processing transaction in real time. You know, in auto booking, 95% of the transaction, referring to the vendors, customers, categories, all those dimensions or objects are really contained and controlled.

[00:33:37] Rather than having messy text fields all over the place. It's just easier to deal with having the AI chat right alongside that accounting data really makes the outcomes driven back toward giving business owners that chance to see what they need when they need it so they make better decisions. Obviously, there's a whole implication around the app ecosystem with MCPs and what tools can work together, how data is passed back and forth, the governance around those. And where things are being plugged in, right?

[00:34:05] To me, the most interesting piece of it is when you look at where the data sits and who's leveraging that, what tools become most relevant and where those modes are at for the data structure to be there and to be able to follow those transactions. A lot of systems out there, I think maybe a little light to follow that story all the way through when you think through that cash and carry model. You know what? Before John does his John Login takes it, we're seeing a lot more of this as Digits has announced it now.

[00:34:33] But I'm pretty sure that Jorien is doing a similar thing. John, tell us. Tell us. Yeah, yeah. I mean, I agree with Kendrick. I mean, it's, you know, MCP kind of like was the thing that everyone wanted to talk about at the beginning of last year and then kind of like died a death and we've gone quiet. I think, I mean, your Digits sell themselves as kind of like the first AI-powered general ledger in the market. You know, you can ask, you can sort of question them whether that's generally genuinely true or not. And whether or not some of their AI is really groundbreaking.

[00:35:00] But, you know, fundamentally, I think what's really interesting from them and, you know, Kendrick touched on this is that, you know, they built this MCP server as a read-only server, right? So they recognize that the place where people want to sit and interrogate some of their data, you know, is going to be in these large language models.

[00:35:14] And, you know, we're seeing more and more of it as you kind of plug in more and more data sources that having that kind of centralized kind of like hub to be able to interrogate, you know, the context of maybe digits alongside maybe your CRM or maybe your, you know, maybe a product like Fishbowl, for example, brings together all of the kind of data in one place. And this is a big challenge for any organization. It's kind of how do you aggregate information? The fact that they've made it read-only, I think, is really interesting because it kind of places that control back in Digit's hands.

[00:35:41] And for me, it's telling me that they still think that the UI and UX of a product still has a value at the moment. And they're expecting people, human beings or products to log into that and to still use it. And that's irrespective of the fact that without too much technical capability, you can spin up an MCP over an API anyway. So you can access their API to read and write through an MCP server if you wanted to. So I think this is great.

[00:36:06] I think, you know, we're going to see more and more products really push a bit harder on building MCP servers to allow you to aggregate this and record a report on this data. So let's see where this goes. I'm going to end on something slightly lighter but quite revealing and to a point earlier that we were talking about relying on that data. The UK government content teams have found that Google's AI overviews are serving people outdated government information because they're scraping stale gov.uk pages that nobody has maintained for years.

[00:36:34] One example, people were searching the cost of setting up a charity and were showing old companies' house fees of £13 online or £40 by post. Interesting because obviously John and I had this conversation just the other day. The costs are actually a lot higher now, so it's £100 online or £124 by post. And when actually it was tested and Google's AI gave a completely different answer, they said first it was free and then they said it was somewhere between £13 and £183 plus.

[00:37:03] So that's not ideal. OK, so this is a perfect reminder that AI is like fairly early in its infancy of all of this. And the model can sound very confident when it does it, but it's a confident idiot because sometimes a summary can look super polished, but it's pulling from zombie government pages that should have been buried years ago. And actually is probably a good enough reason that we should all clear up our old websites. So I think this is like, again, relevant because, you know, we do have very joined up data sources.

[00:37:32] We have got companies housed. We are seeing a lot more in terms of MTD guidance that's being published, employment law changes. It's all down to consistent change. And if the AI then that you're prompting, you've got to check the source data, right? And again, getting that confident misinformation at scale is something you go back to your history lessons here. Always question the source. Check the sources. So I think the Department for Business and Trade has now started auditing old pages and redirecting them.

[00:38:00] And I think it's just a really useful lesson for this week that before we talk about autonomous agents, which we've spoken about this whole podcast and how they might replace us all. And let's make sure they're not quoting old data sources from 2017. Yeah. Ending on that happy note. Thank you for joining us on today's episode. And we hope you found it super useful. If you want to reach out to Kendrick and the team, then you are welcome to connect with them on LinkedIn.

[00:38:29] And we're going to tag him in this post and give him a look up. And hopefully you enjoyed his perspectives on everything that he's going to join us on and future podcasts to do with US perspective, to do with ERPs. And hopefully, you know, we can get some, again, a bit more intelligence into what we're doing. And as always, if you're not already following our Digital Disruptors page, then please do so. And look out for the awards that we're hosting pre-AccountXNorth. The nominations are now open.

[00:38:58] So you can, of course, nominate anyone that you would like to nominate for that and find out more about joining us at the awards event. It will be in September. Don't forget it. Put it in your calendar now. Coming back from the summer, have your nice beach bodies ready because you're going to be, you know, joining us at the event. So that's again, and we'll catch you on the next one.