Our 24 tech predictions for 2026 — and where we want to invest

At firstminute, a $500m early stage fund backed by 130+ unicorn founders, we back exceptional European founders building global technology businesses.

Last year our portfolio companies raised $2.7bn+, we saw our third unicorn (n8n) and we closed a dizzying 28 investments (12 new, 16 follow-ons). But 2026 is going to be even more exciting, we can feel it.

These are the 24 trends we're watching—and excited to invest behind—in the year ahead. Remember, if you are a founder building in any of these—or just anything else you think is exciting—do get in touch.

MODELS & INFRA

World models become a new frontier 

2026 will still be dominated by LLMs. But we will start to see the promise of the next breakthrough architecture: world models. Unlike language models that predict the next word, world models predict what happens next in physical or simulated environments. This unlocks applications in robotics (where machines need to understand physics and spatial relationships), entertainment (imagine game engines that generate realistic interactions on the fly), and agent systems that can reason about consequences before acting. Companies like AMI (Yann LeCun and Alex Lebrun) and World Labs (Fei-Fei Li) are racing to create the "GPT moment" for this technology—where it moves from research to widespread adoption.

Reinforcement learning comes for the enterprise 

Despite growing AI capabilities, most companies still can't automate much beyond codegen, search, and customer support. The persistent "last mile problem"—handling exceptions, edge cases, and complex multi-step workflows—has frustrated AI deployments and made using agents in enterprise slower than hoped. Enter frontier labs trying to crack this problem using reinforcement learning, a technique where agents learn to perform tasks by trial-and-error while receiving feedback in the form of reward signals. RL has existed for years—it's how AI mastered games like Go and Dota 2—but making it more useful outside games and in the more chaotic world of a company (where goals and the nature of success is more opaque than in a game) is a big prize to be won.

The internet redesigns itself for agents 

Companies today should be preparing for an internet where the primary users aren't people—they're AI agents. Already we see people asking ChatGPT rather than using Google, letting AI do the work of visiting a website for them. Visa and Mastercard are building payment rails for agents to complete purchases—they've already processed hundreds of transactions and expect millions more. If agents, not people, are visiting sites and making payments, what does this mean for every business on the planet with a website? How do travel sites design for agents that book flights, not humans who browse? How do B2B software vendors build interfaces agents can navigate? Companies that figure out "agent-native" experiences early—in e-commerce checkouts, SaaS workflows, booking engines—will have structural advantages. Those that dominate this infrastructure will be huge companies.

B2B

More AI-native services companies going full stack 

We continue to be excited about full stack AI companies. This means companies that, rather than, for example, selling AI software to legal and accountancy firms, are attempting to themselves become AI-native legal and accountancy firms. We think this is exciting for two main reasons. One, these companies can build in sectors where it's often painful to sell software (e.g insurance). Two, they can capture more of the value of AI efficiency if indeed they can make AI do much of the work that humans used to do. This is part of a wider trend of capitalising on AI systems that can increasingly eat into labour budgets rather than software budgets. Questions remain around GTM and speed to scale—this model needs killer teams who can build operational heavy business. Buy-and-build strategies offer another path: acquire the customer relationships and operational playbook, then inject AI to improve margins.

Systems of action relegate ERPs to the background 

For decades, enterprise software meant systems of record—databases that store your data. SAP and Oracle built empires as digital filing cabinets, owning the customer through data lock-in. But AI is flipping the power dynamic. Systems of action—AI-powered layers that sit on top of your ERP, CRM, and data warehouse—are becoming where employees actually work. They make sense of fragmented data and execute workflows, while SAP and Oracle get relegated to glorified storage. Whether the old guard will fight back by acquiring or building their own action layers, or whether a new generation of companies will own the interface layer entirely, remains to be seen. But we are interested in companies that sit on top of disparate data sources and be the place where functional leaders within companies spend their time. 

Agents becoming employees— with managers

Agents may not work perfectly in production in 2026, but huge progress will be made towards a world where agents are really more employees than software. This means agents that can take real actions—processing refunds, making purchases, updating CRMs. The shift will also a parallel economy: if agents are workers, someone needs to manage them. Enter agent orchestration platforms that route work, monitor execution, approve decisions, and handle exceptions. This enables the "do more with less" trend to accelerate: tiny teams hitting $50M+ ARR because product, support, and parts of sales are automated. Pricing naturally shifts from per-seat SaaS to per-outcome: tickets resolved, invoices processed, dollars recovered.

Someone builds the "Harvey for finance"

Legal got its AI-native workflow breakout with Harvey. Finance might be next, and the opportunity is arguably bigger. Finance teams spend absurd amounts of time on close processes, variance analysis, reconciliation, audit prep, and regulatory reporting—tasks that are standardised, high-stakes, and painfully manual. The ROI case writes itself: time saved, errors reduced, faster close cycles. Unlike legal (which required nuanced judgment), much of finance work is structured enough for AI to handle reliably while being tedious enough that CFOs will pay to automate it. Excel remains the interface where finance actually happens, so products might feel like "Excel autopilot" to begin with.

AI-native user research 

Customer interviews and usability testing have been manual, time-intensive, and expensive—limited to companies that can afford dedicated researchers. AI can now conduct hundreds of interviews simultaneously, analyze patterns across qualitative data, and surface insights in real-time. Companies like Listen Labs and Aaru are building the infrastructure for continuous, scalable user research. We can imagine a world where product teams can finally get feedback at the speed of deployment. This is part of the wider GTM stack that is being completely transformed with AI—from identifying leads to outbound sales to marketing copy and beyond.

Compliance tech cracks financial services 

Financial services compliance is a massive market (tens of billions annually) defined by laborious, manual workflows—AML checks, KYC processes, regulatory reporting. It's also notoriously hard to crack: high stakes, complex rules, entrenched vendors. But AI's ability to parse documents, flag anomalies, and generate audit trails is finally good enough. Expect 2026 to be the year compliance AI moves from pilot programs to production at major banks. Even though these are painful companies to build, the market opportunity is large.

Frontline workers finally get their tools 

HVAC technicians, plumbers, electricians, landscapers—these are $800 billion markets in the US alone, and they still often run on pen, paper, and WhatsApp. Tools like Avoca, ProBook, and Zimbly are building AI-native workflow software for field services: scheduling, dispatch, invoicing, customer communications. The market has been underserved for decades, and the TAM is mega.

ITSM gets disrupted 

IT service management is plagued by ticket backlogs, manual escalations, and endless back-and-forth. AI can now resolve common IT tickets instantly—password resets, software access, basic troubleshooting—without human intervention. Startups like Serval, Console, and Ai.work (a firstminute portfolio company) are challenging ServiceNow's dominance with AI-first platforms that learn from every ticket and get smarter over time. ServiceNow's $10+ billion market is ripe for reinvention.

AI models of the economy emerge 

Central banks and hedge funds have long built econometric models, but they're static and narrow. Now we're seeing the first AI models that can simulate entire economies—testing policy scenarios, predicting second-order effects, modeling supply chain shocks. It's early and noisy, but the implications are substantial: real-time economic forecasting, better policy tools, and a new class of institutional decision-making infrastructure. We are not totally sure what the big business to be built off this trend is, but it could be impactful for the world.

PROSUMER

Everyone becomes an entrepreneur 

The barriers to starting a business are collapsing. Lovable lets anyone build a front-end. Codewords can generate backend systems from natural language. AI marketing tools handle copy, creative, and targeting. We're entering the era of vibe entrepreneurship: people spinning up real businesses—ecommerce stores, SaaS tools, content platforms—in days, not months.. But so far we have not seen a single platform that unites this all together — where you can go prompt-to-business in a few clicks with a service that can test the idea, build the product and run the ads in a few hours. All the pieces are there — it just needs a team to string it all together.

Everyone automates their own job 

Tools like Codewords and Granola have shown that non-technical people can automate the annoying parts of their jobs—meeting notes, data entry, email triage—without writing code. People are building personal AI assistants that handle their busywork, and companies that enable this will see productivity gains that dwarf previous generations of software. 2026 will see more and more prosumer tools like these that employees use to make their lives just a bit better — and then edge their way into securing enterprise contracts. 

The personalisation of everything 

Media, commerce, and services are becoming increasingly individualised. Imagine a Pixar movie with a slightly different ending based on your preferences. An ecommerce site that reorganises itself for your shopping style. Healthcare that delivers personalised supplement routines and a doctor in your pocket (Counsel Health). Banking apps that adapt to your financial goals. Mass personalisation—impossible at scale before AI—is now economically viable. This is going to be one of the dominant consumer trends over the next 10 years. 

Wealth management for the rest of us 

Today, professional wealth management is effectively available only to people with £500k in investible assets—the economics don't work otherwise. But AI can automate the busywork (portfolio rebalancing, tax optimisation, financial planning), letting human advisors serve 10x more clients. Startups like Range and Clove are democratising access to sophisticated wealth management for people with £50k, £100k, £250k in assets. Jury’s out here on whether the best approach is to sell tooling to small wealth managers, or going direct with a prosumer offering. We suspect the later.

ROBOTICS & DEEPTECH

European robotics make or break

Europe's robotics ecosystem is maturing beyond research labs as talent from ETH, TUM, Imperial, and the DeepMind diaspora is spinning out companies focused on manufacturing, welding, defence, and logistics. Unlike in previous waves of robotics hype, customers are actually starting to buy—not just pilot. The convergence of labor shortages, rising wages, and reshoring pressure across European manufacturing creates real pull and geopolitical tailwinds around sovereignty and security add urgency. The challenge remains go-to-market: teams need both deep technical chops and industry operating experience to navigate long sales cycles and understand what customers will actually change in their workflows. We’re entering a window which may determine whether Europe credibly challenges the US and China, or remains a secondary player in the space.

Geopolitical deep tech remains a funding priority 

Governments are pouring capital into strategic technologies: AI sovereignty, quantum computing, bio-production, reshoring critical manufacturing, life sciences, and defense tech. For investors, insider access becomes differentiating: knowing which programs have real momentum, which consortiums matter, and how to navigate procurement processes.

FROs pivot to startups 

Focused Research Organisations—non-profit entities tackling large-scale scientific problems—were supposed to be the new model for patient, mission-driven research. In practice, many are discovering that the startup model (with its equity incentives, speed, and commercial discipline) works better. Expect a wave of "FRO refugees" spinning out companies in 2026, particularly in AI infrastructure and bio tooling.

European deep tech gets serious capital 

Several new €500M+ deep tech funds are launching in Europe, signalling a maturation of the ecosystem. These aren't generalist funds dabbling in hard tech—they're dedicated vehicles with technical expertise, patient capital, and conviction that Europe can compete in AI infrastructure, quantum, climate tech, and advanced manufacturing. The funding gap is closing.

HEALTHCARE

Workflow tooling for biopharma gets rebuilt

Biotech R&D teams are stuck in a fragmented maze of tools which lack any domain specificity: Excel for tracking, Jira for project management, PowerPoint for collaboration. The entire workflow stack—from process development to clinical trial infrastructure to post-clinical regulatory and quality—is ripe for AI-native rebuilding. Incumbents like Veeva and IQVIA feel like they're from the early 2000s because they are. The opportunity is massive but requires navigating biopharma's reluctance to buy software. The winning approach isn't pure SaaS—it's embedding within customer workflows through forward-deployed teams that prove ROI on discrete projects before expanding. Think Gong for clinical ops, or contract redlining tools for regulatory teams.

Hospitals adopt agentic operations platforms 

Hospitals have been buying point solutions—one tool for bed management, another for discharge planning, a third for patient transport. The result is fragmented systems that don't talk to each other, leaving administrative staff to manually coordinate handoffs. 2026 marks the shift to agentic platforms that orchestrate across tools and teams: AI that manages patient flow end-to-end, flags bottlenecks before they cascade, and coordinates discharge planning with available capacity. The impact is measurable: reduced lengths of stay, fewer readmissions, and real cost relief for cash-strapped hospital systems.

Pharma adopts AI through services, not software 

Pharma companies face pressure to move faster and modernise legacy software stacks, but they remain reluctant to buy software like modern enterprises do. The fastest-growing AI vendors in pharma will be those that disguise software as services—opex-friendly delivery models with forward-deployed teams that embed with customers, prove ROI on discrete projects, and expand once trust is earned. It's AI adoption through professional services until the software becomes indispensable.

Big Pharma's continued drive to be portfolio builders, not drug discoverers 

Early-stage biotechs now account for an increasing share of novel molecules in R&D pipelines, while Big Pharma focuses downstream on portfolio construction, late-stage development, and commercialisation. Discovery is capital-intensive and risky—specialised biotechs with focused platforms are simply doing it better. Pharma's actual strength is in deal-making, running large-scale clinical trials, navigating regulatory processes, and global distribution. Expect more in-licensing, more M&A of late-stage assets, and a bifurcated ecosystem where small biotechs innovate and large Pharma scales.