There was a version of enterprise AI, not that long ago, defined almost entirely by what a model could generate. That version is fading. The version taking its place cares less about what a model can produce and more about whether it can be trusted with real systems, real data, and real customers. That shift is visible across the lineup heading to HumanX Amsterdam this year, an event covered by The San Francisco Tribune as part of its ongoing look at the HumanX Europe circuit, running September 22 through 24 at RAI Amsterdam.
A Market Growing Up
Growing up, in this case, means solving problems that don’t photograph well. Nobody puts a data access protocol on a conference banner. But that’s exactly the kind of work defining this moment, and it’s worth understanding why.
The Infrastructure Nobody Sees
Unfold is the clearest example. Still in stealth, the company is building a way to read closed, undocumented, or vendor-locked enterprise systems, without needing an API, documentation, or the vendor’s participation, all while operating inside the customer’s own environment with read-only access by default and full auditing. Its case for existing is blunt: business logic that stays locked away can’t be modernized, and it certainly can’t be reasoned over by AI.
Two other companies are working the same territory from different directions. Kensho Technologies, S&P Global’s innovation engine, connects large language models and AI agents to trusted financial and business data while structuring proprietary data for machine learning and generative AI workflows, a combination that matters enormously in finance. causaLens, meanwhile, is focused on what happens after access is solved: turning an agentic demo into something that survives production, through Digital Knowledge Workers built on pre-built Blueprints, a customizable Factory, and a governance layer it calls the System of Work, with causal reasoning and human-in-the-loop controls running throughout. And Trendium is making sure none of this happens in the dark, with its ContextGuard platform providing visibility and control over AI infrastructure and its AI Enablement Program helping engineering teams use AI coding agents responsibly, all built around visibility, control, enablement, and compliance.
The Voice Layer
Voice is where this shift becomes easiest to feel firsthand. AssemblyAI provides the infrastructure developers reach for first, with APIs covering transcription, contextual understanding, and real-time agentic workflows. Speechmatics goes further into the hard parts of the problem, low-latency recognition across more than 55 languages and multiple speakers, deployable in the cloud, on-premises, or on a device, serving everything from legal transcription to live captioning. And Otter AI has taken the boldest step of the three, moving beyond transcription into what it describes as a conversational knowledge engine, one that identifies decisions and action items and turns meeting history into something searchable and queryable, connected to tools like CRM systems.
Discovery, Commerce, and What’s Next
Two more companies round out the picture, each applying this same maturity to a field outside traditional enterprise software. Nuritas is using its proprietary Nuritas Magnifier platform to discover and validate peptides from nature, having already identified more than 8 million of them, moving discoveries from computational prediction through clinical validation in a fraction of the usual time, with direct applications in health and nutrition. PhotoRoom is doing something similar for e-commerce, building AI visual infrastructure that includes batch editing, automated quality assurance, and brand controls, with a firm commitment to keeping generated product images faithful to what’s actually being sold.
None of these nine companies is chasing a bigger headline. They’re solving the problems that decide whether AI actually works once it leaves the demo, which is precisely the conversation HumanX built its VentureConnect and SolutionBridge programs to have.