67. Top 10 Category Predictions for 2026

0:00Welcome to The Difference Engine, the show for tech founders, investors and innovators.

0:09what do we have to look forward to today in our first episode of the new year we'll be channeling our mystic mechs to make some bold category predictions for 2026

0:24what does the future hold let's look into our crystal ball well welcome to the difference engine it's that time of year again when we get our crystal balls and tell you what we think is going to be happening in the world of tech and category in 2026. One of the benefits of having this many miles on the clock and having been involved in tech so long is we are not afraid to take a risk and call it out. You'll not be surprised, though, to know that many of these new categories exist purely because of the rise of AI. And we believe this is the year when AI really starts to kick in and accelerates new category definitions.

1:03Okay, well, here we go. The top 10 category predictions for 2026.

1:13So coming in at number 10, you won't be surprised that we believe agentic platforms are going to create a new enterprise software category. So what we believe is going to happen is 2026 will formalize the shift from isolated AI models to multi-agent systems that coordinate tasks, negotiate constraints, and execute workflows autonomously. Vendors are going to begin offering agent orchestration layers, policy engines, verification and simulation sandboxes. And we love a good sandbox. So we think this becomes a new platform category. We call it agentic work orchestration. A-W-O. And I'm just wondering if A-W-O becomes AWOL if we have agentic work orchestration life cycles.

2:01So we did that.

2:05Right, at number nine, we think there's a radical change. And we are looking for the AI-first infrastructure, which we called the AI Compute OS or AI Compute Operating System. There's a high demand for model training, fine tuning and deployment. And this is driving a new systems layer that sits lower down in the stack, perhaps under Kubernetes or Docker or one of the alternative technologies out there. This is going to automate all of the manual IT ops processes that are involved in deploying, managing and scaling containerized applications. So a new compute operating system is all to play for.

2:43That's the category that's going to manage all of the GPUs, TPUs, NPUs, accelerators, memory, fabric, model routing, and power use. Who's going to win it? No idea. But it's about time enterprises started to treat computing the way that they treat cloud providers. And that changed a lot after virtualization, and it becomes the strategic control point for AI.

3:08All right. And number eight, The Chain. And it's not Fleetwood Mac. It's business units adopting AI chain of record systems. So why this? Well, as AI generated outputs permeate operations, companies require provenance, traceability, and lineage for decisions made by autonomous systems. So what do these framework logs? Well, it's this, you know, which models were used, which prompts or agents actually acted, how data was transformed, who or what approved the actions. So a new compliance tech category emerges, AI, chain of record. Right, and while we're talking about working on the chain gang, there's been a lot of debate about legislating AI, and there's going to be a hell of a lot more next year.

3:56That is without doubt. The Europeans clearly want more legislation, and the freewheeling break things and make things happen US does not. They do say the US innovates and the EU legislates. We're not falling for this. We think this is just US-centric dumbing down. And we need market forces. We also need legal interactions. So we think at the difference engine, there's going to be a halfway house as there was with the GDPR legislation. And as we see recently, all powerful US category leaders like Reddit is finding out in the Australian courts as it challenges their decision to make teenagers not have um social media anymore so guys not everything is going to go the way of the disruptors uh in 2026

4:51so lucky for some at number seven the rise of micro clouds for sovereign and edge ai and what we're seeing here a new architectural pattern forms an entirely new category micro clouds self-contained, AR-optimized stacks, deployable in a single rack, store, factory, or even region. Now, what do they do? They merge edge compute, model serving, data governance, and secure networking. Right. And all of this, we think, will be replicated in software. So you've all heard of large language models, quite possibly in the mainstream media or MSM. How about micro language models. DeepSeq has proven already that big and expensive isn't always beautiful or practical.

5:36And as any tech adoption proves, there are early versions, think about mainframes, loggables, or dial-up internet, and PalmPilots even. And then there's a steady state. And in a steady state, you often suffer a price war, which is exactly what we've seen going on. Smaller models such as Mistral 3 and others from Microsoft and Chinese Alibaba are coming. Now, this might even give way to micro language models or MLMs, if you like. And one reason is the classic good old fashioned reason in tech of constraints. Have you ever noticed how dogs have specialized skill sets and they do all that with less compute power than humans?

6:15Yes, not every AI case needs to ingest the entire Encyclopedia Britannica. And as we all carry reasonably secure supercomputers around in our pockets, maybe that's all we need and for more of this have a look at how this prediction was presaged in episode 50 of the difference engine back with our conversation with amber bodegaal

6:40okay so delivering their kicks at number six for those of us who sit on boards cognitive supply chains are going to become a priority now why is this because aiit expands from digitizing human based supply chains to augmenting them with predictive autonomous logic. And so a new category appears, cognitive supply chain platforms. And what characterizes these? Well, they combine real-time sensor networks, digital twins, multi-agent optimization engines, and demand forecasting and scenario simulation. The point is, these systems will pull together technologies already existing in supply chain management, which we know as SCM, and that's an existing category.

7:30But they will make a step change to continuously negotiate, plan and adapt with minimal human input. Although some input, no doubt. And I like this idea of supply chain, cognitive supply chain, because in our experience working with tech leaders like Max, Azoma, and others, the very best use cases, at least right now, is where AI and expert humans interact. And we've already seen folks such as our good friends at Sastra, who don't count their workforce just in carbon-based humans, they actually count the agents too. So blending humans and agents into one cognitive workforce, so to speak, is perhaps a more accessible way of considering the rise of this new category, cognitive supply chains.

8:22Entering the top five now, so we think 2026 will see the emergence of a recognisable new category, AI native application architecture. A little bit of a mouthful, I know, but it will become known as Anna. So apologies out there to all the Annas that are listening. And for those who don't know, Jono's surname, SimNet, was once used to describe the military simulation network, wasn't it? Yeah, yeah, it was. Anyway, let's get back to Anna. Yeah, right. Let's go back to Anna after that little diversion. So instead of writing business logic, developers are going to assemble some different things. And what we think they're going to assemble is agents, policies, goal functions, data contracts, and in fact, actually the guardrails, which we've talked a lot about in past episodes.

9:11What's really interesting, regardless of this new architecture we call it, Anna or not, is how notions of architecture, which obviously involves nice, neat classifications of structure and things like meta layers and meta language, rubs up against the somewhat chaotic world of vibe coding in which we're all going to increasingly live. So you've got products such as Lovable, Replit, Cursor, and this is giving a whole new cadre of innovators a way to build without overthinking things like data structures, user interfaces, or even, controversially, IPO ownership. To me, this is like how hip-hop DJs redefined musical structure with scratching and breakbeats.

9:53And last year, that's 2025, I saw this with my own eyes on a course where I saw a teenager build an amazing functioning app prototype in hours for a brand new, highly commercializable use case. Perhaps he needs an Anna. We should probably introduce him. Anyway, why does all of this matter? Well, I think it's because this is a paradigmatic shift, because software becomes more like governed emergent behavior. Think about that. Very, very dynamic, rather than the static code, which we've basically built for the entire history of the computing industry. And that just makes it just a little bit more organic.

10:37Ooh, nice.

10:42speaking of organic at four and a new forecast for us let's talk about elastic security mesh a new category that we think is going to predominate okay so cyber security has had to adapt to and use agentic and autonomous systems because that's what the bad boys are using So we think that in 2026, we're going to see security tools evolve into another new category. And we're going to call that the elastic security mesh. Now, I'm sorry if that sounds like some obscure form of underwear. It's not meant to. It's something that is formed of self-adapting security layers. Now, what do they do? Well, we think they're going to detect, model misuse and drift.

11:35They're going to auto-patch vulnerabilities. They're going to regulate against permissions. And they're going to provide dynamic micro-segmentation. So the point of this is, and we've seen that in some of the clients we've worked with, is that this is part of a journey where security becomes continuous. you know it's model aware and it's predictive rather than reactive so again very much the focus on vulnerabilities deciding what you want to do about them and going and dealing with them and again let's let's not focus specifically on the category name although elastic security mesh does sound like a goer to me there are thousands of cyber players here and they're all looking for some differentiation they should really think where this is going as a follow-on thought though So whether or not this new model of a security mesh will help depends very much on how quantum computing kicks in, because that's going to change the very definition of encrypted communications.

12:34And we're not predicting that QDay will happen in 2026, although we might need to be a little bit cautious on that. We do know it's coming.

12:47Are you still developing separate marketing plans for each of your different channels to market? Can you really sanction messages reaching the same customers from unconnected targets? Listen, it's now the AI age. Things need to change. Positive is a different sort of AI-powered agency. From our very beginning, we've integrated marketing communications from PR to events, social to paid ads. Working smarter means working integrated. Let us show you what a truly joined-up strategy could do for your tech business. contact positive today with an email to hello at positive marketing.com and now for some free thinking number three what if it's going to happen to number three well we think energy aware it fabric becomes foundational now why why this in 2026 well you know we are seeing a massive built out build out of compute ability particularly with the rise of AI and AI processing.

13:49And it's quite clear that compute demand is rising faster at the moment than energy availability. So infrastructure has not been matched by energy availability. And we know that particularly in this environment, that the sheer energy requirements of AI processing is not a good ESG look. So we believe in the coming year IT organizations will adopt a new layer category that optimizes workloads and that is based on a number of factors. We would think energy price, carbon footprint, thermal limits and the concurrency of AI training activity. So this should create the energy aware fabric that's the AF category and that we would define as a system approach controlling when where and how computing is executed based on sustainability and of course cost.

14:53Yeah and I think I sort of definitely agree that we're going to see something like energy aware IT fabric but for a slightly different reason. There's no doubt that energy awareness needs to rise again globally. And perhaps locally, in the West, we've constrained ourselves artificially because of net zero rules, which were conceived in a different era, with fewer wars, no tariffs, and in retrospect, just more general harmony. In today's environment, renewing the infrastructure is not a nice to have. And in times of conflict, it may become literally life or death. Obviously, we're hoping this does not happen in 2026.

15:37Last two, big two. AI-driven enterprise intelligent graphs, replacing traditional data operations. Now, this is another paradigmatic shift as companies unify operational data, model outputs, domain ontologies, and real-life context into, as we said, enterprise intelligent graphs. Now, this is so important because these become the strategic brain of the enterprise, enabling things which we will have to have going forward, which is autonomous reasoning, dynamic decision support, and holistic impact modeling. So why this is important is it marks the evolution from decades of attempts at data management to a different approach of knowledge orchestration.

16:32Yeah, and I really like this. I mean, you take data management to knowledge orchestration. It's a big step up. We're not doing the plumbing anymore. We're actually getting some value. Something's got to give. I mean, we've got dashboards and data lakes, and they used to do data management, but they're just too static and ponderous for this new AI world. We need AI at speed, and everyone who's used AI at scale, for instance, for deep research or coding, understands the limitation of current context windows. And we see workarounds like snapshots or running several LLMs in parallel, but this is a major issue, and we need to solve how we're using data and data operations for AI.

17:14This takes us to number one, and possibly listening to all these massive changes we're predicting, the one you've been waiting for, as you've been wondering where your careers and sales opportunities will be going in 2026. Well, we think that IT people, that's all of us, shift into a different role or different roles based around strategic digital stewardship, SDS. SDS. Only yesterday I took a call from somebody asking me about NeoClouds. So this is absolutely live, absolutely going on. And we've already seen that LLMs are resulting in brand new but critical roles. For instance, who would have predicted that good old customer service agents who have already suffered the indignity of being retitled as customer success professionals.

18:06There's a controversial title. And now being, quote unquote, upgraded. And we see a lot of what they call forward deployment engineers. This is going to continue. Forward deployment engineers literally make the difference between retaining loyal customers and growing AR, ARR, or being called out publicly, which is, in this social media world, death to future sales. Anyway, good news. As automation accelerates, IT's strategic responsibilities expand dramatically. Woo-hoo! And there could be a rash of new titles to contend with in SDS. Now, here's our guess at a few. So, how about Chief Agent Officer CAO?

18:54What are they charged with? well, essentially, governing autonomous systems. If that sounds reasonable, how about Director of Machine Policy and Governance, DMPG? That's exactly what it says on the tin. Sounds a bit like a government official, but yes. A DMPG, it does indeed, yeah. Here's one that I hope doesn't sound like a government title. How about Model Risk Steward, Mrs. MRS? Here's our last guess, right? Synthetic Workforce Architecture. which I particularly like. I know it sounds a bit dystopian. I really can't imagine a bunch of 17th century aristocratic landowners dismissing machine pioneers like Hargreaves, Compton and what?

19:40Just a bunch of bloody jumped up synthetic workport architects as they pass around their port after dinner. Anyway, if this new breed have to embrace architectural tradition by wearing ridiculous workwear jackets, red spectacles and saying the zeitgeist a lot. The good news for IT insiders is that IT, we're going to call it architects, become responsible for, and we think this is a term we're going to hear a lot in the next 12 months, digital labor, not just digital systems. Now that's a fundamental reframing of enterprise operating models and the roles that go with them. Yeah, so why don't we go a little bit surreal.

20:17Where does this all end up? Do we need to consider the rights of AI workers? Do they need a union? or maybe even an HR function. It might seem crazy to anthropomorphize AI. It's essentially, remember, just software code. But as tragic events have shown recently, some vulnerable members of society are using AI tools as more than guides, more than counselors, but more like friends. Just to be clear, AI is your teammates. AIs are your teammates. They're not your mates. Never forget that, people.

20:58To paraphrase a great TV detective, Columbo, which you may or may not be familiar with, just one more thing. So here's just one more thing about what we think is happening in 2026. Take it away, Paul. Right. So we think there's going to be a new, yet to be named, AI-friendly set of financial metrics. So there's a new type of revenue accounting that we need to represent in the AI era. And there's a lot of talk of a bubble at present based off of some facts, the extraordinary trillion dollar valuations, the accusation of round-robin dependencies between vendors, and that's seeing the exchange of equity and cash between those who provide LLMs and GPUs and TPUs and the AI infrastructure builders who build the data centers.

21:46and all of this enmity builds up in those of us not involved in AI and it's getting called out. Traditional accounting is now being replaced with a sort of futures-based logic and just like in the cloud computing era, the accountants need to catch up. Of course they do. Witness Oracle's recent use of something called RPOs. What's an RPO, Jono? It's a Remaining Performance Obligations. Yeah, does it sound a bit like BS to you? Does it sound like WeWork when they tried to rewrite the accounting rules? Oh, yes, that great real estate company masquerading as tech. Yeah, great office firm. As far as we can tell about RPOs, this is about future expected cash, or possibly debts, really hard to understand, from the newly announced and slightly unclear partnerships they have, this is Oracle has with NVIDIA, Meta and others.

22:38Slightly reminiscent of what happened when the idea of cloud revenues took hold. Well, it is. Yeah, it is. But that we seem to solve, right? Yeah, the accounting rules change because you could then properly recognize recurring revenues, which you'd sold and been paid for, but couldn't come onto the books. And when that kicked in, valuations normalized and everybody was happy. A whole of M &A valuations in software based on recurring revenues now. Exactly. Massive change. So our final prediction, listeners, is that there will be a new AI-friendly type of financial metric. It may or may not be RPOs yet to be seen.

23:21Remember, this process took years back in the cloud era, but we think at least it will be starting in the next year.

23:36So, what do we think we've learned from this? I mean, if you think we've gone a bit mad, then please give us some feedback. We'd love to hear from you. I mean, ultimately, after a few years building up, this year is going to bring a series of emerging themes together. It's going to be a big one for change. A new category creation will be based on a series of paradigmatic shifts in the way that we think about tech and our role in it and our relationship with it. And there will be a new language emerging that describes the technological principles behind the changes brought by the latest industrial revolution that is upon us.

24:14And make no mistake, we are in the relatively early stages of an industrial revolution. Now, these changes are so fundamental. Perhaps we should think of moving on from the current informal geological age, the Anthropocene. and that has been marked by significant human impact on the planet. And I think we need to move, with this change, to a new era. We call that the machine epicene. The machine epicene, where the arrival of digital labour forces us to face fundamental choices about how we treat both the planet we live on and the people that exist on it. Perhaps we should embrace new opportunities to act for the common good.

24:57Now, that should be our New Year's resolution. and machine app scene. You heard it first on the difference engine, folks.

25:08Thank you for listening. If you want to learn more about category design, head to becategorical.com. If you need help designing and dominating your category, then get in touch. The contact details are in the show notes.