After two transformative years diving deep into GenAI – including 250+ projects and 20 product launches across our Hg portfolio – one thing remains crystal clear: the GenAI flywheel isn't just spinning, it's accelerating.

Fresh from our latest Silicon Valley reconnaissance - meeting with leading AI investors, model builders and start-ups - here are our latest observations:

It pays to be bold & brave

While sceptics talk of the ‘trough of disillusionment’, in reality leaders are already banking real returns: a consistent 2% margin uplift is evident also in our Hg portfolio, with GenAI product extensions delivering even stronger growth trajectories.

The key differentiator? It's not just technical prowess – it's leadership commitment and the operational transformation muscle to convert promising AI pilots into P&L impact. And just as we saw in the previous SaaS transition, the leaders are pulling ahead early.

Foundations models no ‘moore’ magic (at least until the next act)?

After early dizzying, exponential foundation model progress, talk of a plateau in the valley is building – so will more linear steps will be the norm? Well, only until the next set of technical breakthroughs - for instance, near-infinite memory is rumoured for as early as 2025.

… but for the user lots of (agentic) magic still to come

Still, many see a clear path for years of value creation impact, as the wider GenAI scaffolding (e.g. the ML Ops layer) is still maturing, and the AI app layer is getting to grips with agentic frameworks to take their value add from co-pilots (‘do it alongside me’) to agents (‘do it for me’). Firms like Cognition AI’s Devin or Anthropic’s Computer use, provide a window into the agentic ‘magic’ still to come.

Already in these early innings, we’re seeing the potential for ~2x revenue per customer from these agents (Bessemer has some smart thinking here) matching the uplift the SaaS transition gave us only after years of optimisation.

Foundation model act 2: from research labs to product strategy

The oligopoly between OpenAI, Anthropic, Google and Meta seems to be firming up, with stratospheric R&D costs a challenge for the rest (or at least ‘most’…Grok may yet catch up). Maybe in response to more linear model progress, they are now eyeing up the product stack above them - and we’re excited for what’s to come. This is creating waves: horizontal tools that represent ‘thinner wrappers’ on top of their models are now pushing deeper into applications (e.g. Perplexity enterprise search).

SaaS Incumbents have a head start

Many are pursuing the massive TAM expansion that promises a paradigm shift from ‘technology making labour more efficient’ to ‘displacing entire portions of labour’. Vertical SaaS incumbents rightly have strengths to leverage: they’re embedded with customers, already ‘plumbed in’ to deliver certain workflows, and their domain expertise and proprietary data puts them in a prime position to drive GenAI value-add. But start-ups are also coming, and fast: smart tactics like customer partnerships overcome domain and data gaps, and to get distribution (see Harvey & PWC). Leading incumbents are ‘refounding’ themselves to overcome their innovators dilemma and to capitalize on their head-start (see Intercom).

Architecting GenAI-first: new buildings not extensions

A new GenAI tech stack is emerging, with GenAI agentic products being built both separate to, and alongside, existing SaaS product stacks. It also requires a new approach, focusing on flexible workflows, constant learning from data, human feedback, and ethical checks, with an SDLC built for experimentation and improvement. Therefore the sharpest investors are looking past ‘GenAI extensions on top of SaaS foundations’ and, instead, seeking out teams capable of driving ‘GenAI first’.

This enables fast innovation with dedicated GenAI squads (which is the inspiration behind our Hg GenAI incubator). Still, a company’s tech foundations and talent need to evolve and become GenAI first.

Many waves, not one revolution. But the more you own, the faster you can go

The GenAI revolution will be multi-speed: GenAI tech firms, ultimately, are reliant on their end-customers transforming their operations and, just as with SaaS, we will see adoption ripple out from tech-first businesses (e.g. Klarna). But savvy investors are seeing a short-cut from owning the to-be-disrupted labour pools, seeking out service heavy verticals, to disrupt from within.

The upshot? Be bold, and brave!

Our Hg companies demonstrate that GenAI technology available today already offers material value-add on product and operations. But it’s a hard transformation. And like with the SaaS transition, we’re encouraging our firms to ‘fire bullets, then cannonballs’ – experimenting fast, then scaling up successes. Those leaning in are already pulling ahead, and will only accelerate further as this GenAI ecosystem matures. Just as we saw from the previous platform shift.

Join the journey

At Hg we’re investing in our ‘GenAI flywheel’ – through exceptional Hg AI experts, R&D talent as part of our Hg incubator (Hg Catalyst), to tight partnerships with pioneering AI businesses to support our Vertical SaaS portfolio. If you want to join our journey as we drive this next platform shift, get in touch: AI@hgcapital.com!

Read more observations from our last visit to Silicon Valley six months ago, where we connected with foundation model providers, VC firms and the developer ecosystem rapidly growing around GenAI.