A year is a lifetime: HumanX returns to a very different AI world
The show that planted an early flag in terms of leading the conversation around AI's customer impact is back
When HumanX debuted in Las Vegas last March, DeepSeek was a name that maybe a few hundred researchers would have recognised. Reasoning models were a novelty. Cursor had a devoted following but hadn’t yet become a verb among developers. And the idea that AI agents would be autonomously merging pull requests, managing corporate expenses, or finding and patching security flaws without a human in the loop would have sounded like a pitch deck fantasy – ambitious, sure, but comfortably distant.
Thirteen months later, every single one of those things is real and shipping. HumanX 2026 opens at the Moscone Center in San Francisco on April 6, and the sheer volume of ground that has shifted beneath the conference’s feet since its inaugural edition is absurd. The event has nearly doubled in size, but even that growth feels inadequately proportional to the pace of the industry it’s convening to discuss.
The move to San Francisco is more than a venue upgrade. Placing an AI conference at the geographic epicentre of the thing itself was the obvious play, but it also signals something about the maturation of HumanX as a gathering. The Vegas debut had the energy of a well-funded first draft – impressive speakers, genuine buzz, solid editorial programming. Moscone is where you go when you’re no longer proving you belong; it’s where you go when you’re setting the calendar for an entire sector.
And the sector desperately needs a coherent calendar right now, because the last year has been a blur. Consider just the broad strokes: DeepSeek’s R1 arrived in January 2025 (yes, only a year ago, and yet if feels way longer than that) and sent shockwaves through both Silicon Valley and public markets, demonstrating that a relatively small Chinese lab could produce frontier-class reasoning for a fraction of what Western labs were spending. The “DeepSeek moment” became shorthand for a kind of efficient innovation that challenged the assumption that you needed tens of billions in compute to compete. OpenAI shipped o3, o4-mini, and eventually GPT-5 over the course of the year. Anthropic released Claude 4 and iterated through many point updates across Opus, Sonnet, and Haiku variants at a pace that made keeping track of version numbers a part-time job. Google DeepMind’s reasoning models were winning gold at the International Math Olympiad. And underneath all of it, the real story – agentic AI moving from demo to deployment – was quietly rewriting how software gets built, how companies operate, and what “work” actually means for a growing number of knowledge workers.
That last shift is the one that makes this year’s HumanX speaker lineup feel particularly well-timed. The show’s roster spans the full stack of the current AI moment – from the researchers building the models to the founders deploying them in production to the policymakers trying to figure out what guardrails make sense when the ground won’t stop moving.
Fei-Fei Li – co-founder of World Labs and the person who, through ImageNet, arguably did more than anyone to make modern deep learning possible – will speak on what comes next. Matt Garman, now running AWS, and Mike Krieger, leading Anthropic Labs, represent two very different but deeply consequential visions of how AI infrastructure and products should evolve. Sarah Guo of Conviction, whose No Priors podcast has become essential listening for anyone tracking the space, will be there alongside Bret Taylor – co-founder of Sierra and chairman of OpenAI’s board – and Srinivas Narayanan, OpenAI’s CTO of B2B Applications. Vinod Khosla and Scaled Cognition’s Dan Klein will tackle the tension between super reliability and super intelligence, which is – if we’re being honest – the core question the entire industry is dancing around.
Then there are the founders who are translating all of this foundational capability into products people actually use every day, and three of them are worth calling out specifically – with the disclosure that both of these two are clients of SBS Comms, my current employer.
Michael Truell, the co-founder and CEO of Cursor, will be on the main stage discussing how AI is reshaping the act of writing code. Cursor has had a year that borders on the surreal. The company surpassed $2 billion in annualized revenue and hit a $29.3 billion valuation, but the numbers are almost secondary to the product story. Truell recently described a “third era” of AI-assisted development – one where developers are no longer writing code keystroke by keystroke, or even directing agents through synchronous prompts, but instead managing fleets of autonomous agents that operate like teammates. More than a third of the pull requests Cursor merges internally are now created by agents running independently in cloud VMs. Agent usage in Cursor has grown over 15x in the past year, and the company now has twice as many agent users as tab-autocomplete users – a complete inversion of where things stood just twelve months ago.
Eric Glyman, co-founder and CEO of Ramp, will talk about how AI is rewriting the rules of finance – a session title that might sound like standard conference fare until you look at what Ramp has actually shipped. The company’s agentic AI for expense management is now more than 99% accurate at tasks like determining whether a travel expense adheres to company policy – more accurate than humans working alone, and serving over a million end users. Ramp crossed $1 billion in annualised revenue and reached a $32 billion valuation late last year, and Fast Company just named it one of the most innovative companies of 2026. Glyman’s framing around “zero-click” experiences – stripping dozen-click legacy workflows down to a few clicks, and then eventually to none as the AI gets smarter – is one of the clearest articulations I’ve heard of what agentic AI actually looks like when it’s deployed at scale in a domain that isn’t software development. Plus, he recently dropped an essay on the shift that’s already emerging favouring companies who are leaning into AI spend, and that’s bound to be a buzzy topic of conversation at the show.
Beyond these three, the agenda is dense with sessions that reflect the breadth of where AI is actually landing: Demi Guo of Pika and Dean Leitersdorf of Decart will discuss the distinction between viewing and experiencing in AI video. Arvind Jain of Glean and Eric Yuan of Zoom will do a joint Q&A. Snowflake’s new CEO Sridhar Ramaswamy, Thomas Wolf of Hugging Face, and Anton Osika of Lovable all make appearances. There’s a whole track called “Control Room” dedicated to how AI transforms enterprise operations, and another called “Builders” that gets into the infrastructure layer.
The thing that strikes me most about this year’s HumanX, though, is how different the ambient mood of an AI conference feels now compared to even twelve months ago. Last year in Las Vegas, there was still a palpable undercurrent of “is this real?” – not about the technology itself, but about whether the enterprise adoption wave would actually materialise, whether the valuations were sustainable, whether AI would remain a tool mostly used by developers and early adopters. That question has been answered, decisively, in the affirmative. According to McKinsey's most recent global AI survey, 88% of organisations now use AI in at least one business function – up from 72% just a year earlier. Worldwide AI spending (across infrastructure, software, services, etc.) crossed $1.5 trillion in 2025 according to Gartner. The conversation has shifted from “should we adopt AI?” to “how do we not get left behind?”
Which means the conversations that matter most at HumanX this year won’t be the ones about capability – we all know the models are getting better, faster, and cheaper at a rate that makes Moore’s Law look leisurely. The conversations that matter will be the ones about deployment, integration, trust, and the deeply human question of what all of this velocity actually means for the people and organisations trying to absorb it. The conference’s tagline has always leaned into the “human” part of HumanX, and in 2026, that framing feels less like branding and more like the central challenge of the moment.
San Francisco is the right city for that conversation. And next week is very much the right time. I’ll be there, so drop me a note if you will, too, and let’s continue the conversation.
Disclosure: Michael Truell/Cursor, Eric Glyman/Ramp, and Glean are clients of SBS Comms, where I work. I was not compensated for this piece and all opinions are my own.