Big Tech Is Hungry, but the Real World Isn't on the Menu (Yet)

A week in San Francisco, a five-year SAP migration, and the gap in between

I just arrived back in London after a full-on week in San Francisco, and let’s just say the jet lag was the least disorienting part. The moment you land in SF, you see it: this city breathes technology. Every billboard screams about the latest AI product launch. The coffee shops double as pitch rooms - overhear any conversation and someone is founding something, funding something, or pivoting something. I hopped into a Waymo, no driver, no small talk, just a calm robotic voice and a smooth ride through the city, and it barely raised an eyebrow from the locals. Autonomous cars are practically mundane there. Tech campuses gleam with the promise of trillion-dollar futures. Even the food delivery robots trundling along the pavement seemed unremarkable to everyone but me.

And yet, for all that dazzle, the contradictions are stark. You can ride a driverless car down one block and pass a row of tents on the next. The wealth and the hardship exist side by side in a way that’s impossible to ignore and hard to square. Needless to say, I left San Francisco equal parts enthralled and unsettled - and quietly banking on autonomous vehicles reaching London before anyone in my life finally forces me to get a driving licence. I’m just long on autonomous self-driving technology, what can I say.

Enter Clawdbot

The tech highlight of that week was the release of Clawdbot, an open-source AI agent that quickly became the talk of the town (and the internet). Everyone around me was experimenting with it, and the demos I saw were genuinely mind-blowing. This wasn’t just another chatbot spitting out text - Clawdbot could actually take actions autonomously. Hook it up right, and it can spin up mini-agent helpers and perform real tasks on your behalf: sending emails, managing your calendar appointments, even writing code. In one demo, I watched it scour Reddit forums and produce a detailed research report summarising what users were saying - complete with neat bullet points and relevant links. The output was so polished and professional that it read like something a human analyst prepared. It also handled scheduling like a (slightly overeager) personal assistant, adding events to calendars with suggested travel buffer times. In short, I was massively impressed (and a bit freaked out) by how much initiative this “autonomous AI intern” could show. Clawdbot’s rapid emergence and capabilities had everyone in SF chattering about an agentive AI future where bots do all the things.

So by the time I flew back to London, my head was in the cloud(s) - buzzing with excitement over this AI wave and the groundbreaking potential I’d just witnessed. But I’ll admit, underneath that excitement I felt a pang of panic. As a venture capitalist, I started asking myself: What does this mean for the world? For the companies we invest in? If AI agents can build products, automate workflows, and basically “eat the world,” where does that leave the rest of us trying to build businesses? My week in the tech mecca had me wondering if the startup ideas we’re funding today will get steamrolled by Big Tech and their ever-more-sophisticated AI models.

Two Doses of Reality

And then the real world did what the real world does. In my case, reality arrived in two sharp doses.

Dose one: my fiance, who works in private equity and has a front-row seat to the less flashy side of business. No sooner had I returned home and begun gushing about autonomous this-and-that, he blurted out in exasperation: “Five years. Five bloody years to implement SAP, and they've just delayed again”. Now, profanity aside, that one-liner was exactly the bucket of cold water I needed. It instantly burst my AI bubble and snapped me back to a crucial insight: we all become products of our environment. I had just spent a week steeped in Silicon Valley’s cutting edge, where every company is racing to deploy AI agents, but my fiance’s comment reminded me that most of the business world moves at the pace of molasses in comparison.

Think about it - while I’m marvelling at Clawdbot drafting a perfect report in seconds, many large enterprises are struggling for years to roll out something as unglamorous as an ERP system upgrade. His example wasn’t an exaggeration: plenty of huge companies really do take half a decade (or more) to fully implement systems like SAP. In fact, nearly 60% of big firms’ SAP software migration projects blow past their budgets and timelines. We’ve also seen reported how almost ten years after SAP’s latest platform launched - around 39% of SAP’s major customers had still not even started migrating to it. These are Fortune 500-type organisations with massive resources, yet their digital transformation is crawling.

Dose two came a few days later at a dinner, where I found myself seated next to the CEO of a large industrial enterprise. Naturally, I couldn’t help myself - I launched into a spiel about agentic AI, autonomous workflows, the whole frontier-model future. He listened politely, nodded along, and then said something that stopped me in my tracks: “Bella, this is great, but I’m not a tech company. Sure, I’ll use tech and it’s important, but I have no interest in building and maintaining my own custom software. It’s a distraction to us. We produce automobile parts.” That landed hard. And he’s not alone. IBM’s 2025 CEO Study found that 57% of CEOs actively prefer to outsource non-core activities like technology rather than build it themselves. Most businesses don’t want to be tech companies. They want tech that works out of the box so they can get on with making automobile parts, running hospitals, or selling insurance.

Hearing both of these reality checks - one from the world of enterprise IT, one from the C-suite of the real economy - was a moment of genuine self-awareness. It reminded me that my tech-optimistic viewpoint had completely left out the everyday challenges and slowness that define so much of the real economy.

I had to laugh at myself a little. One week in San Francisco and I was ready to herald the AI revolution as if the whole world had instantly changed. The truth is, I live and work in a bubble - the startup/VC bubble, where we hype each other up and genuinely believe the future is already here. It’s an intoxicating bubble, but it is a bubble. My fiance’s frustration came from outside that bubble, where businesses fret over basic IT projects and incremental improvements. The CEO’s bluntness came from even further outside it - from a world where tech is a tool, not an identity. His world (and arguably most of the world) couldn’t be more disconnected from the tech utopia I’d been basking in. As Marc Andreessen famously wrote, “software is eating the world” - but sometimes it’s easy to forget that not everyone has had a taste yet.

Stepping back, I realise our tech bubble’s reality distortion field blinds us to how far behind much of the real world really is. Case in point: I grew up in Japan, a country often perceived as ultra-modern. Bullet trains, robotics, high-speed everything — but try opening a bank account. You’ll likely need a physical personal seal (hanko) to stamp forms, because many banks and government offices still insist on ink signatures for official business. In 2025. The government has been pushing to cut hanko use by 90% in public services, but progress has been painfully slow. It’s a surreal mix of cutting-edge and antiquated.

And it’s not just Japan. Look at Europe, look at the United States - beyond the tech hubs and headline-grabbing pilot projects, a huge chunk of our economies still runs on paper, pens, and 20th-century tech. Government agencies, hospitals, banks, you name it - many are only inching toward modernisation. A recent survey across U.S. and European organisations found that roughly 20-40% of companies still rely on a mix of paper-based and digital workflows for their core processes. In other words, fully digital, automated operations are the exception, not the rule.

Want a painfully ironic example? Healthcare, a field that should be at the forefront of innovation, is one of the worst adopters of technology. The majority of pharmaceutical firms are still only partially digital, with the vast majority of industry digitisation programmes failing entirely. The vast majority still shuffle lots of paper along with their new digital tools. Many U.S. doctors’ offices and hospitals still communicate like it’s 1985 - fax machines are unbelievably common. (Yes, those same screechy fax machines that my AOL dial-up used to sound like.) About 70% of healthcare providers in the U.S. continue to use fax machines to exchange medical information. If that isn’t a reality check, I don’t know what is. We have AI models that can beat humans at complex strategy games and generate code, yet your local clinic might ask you to fax over a form because email is deemed too modern.

The broader point here is that technological change takes time to permeate society - a lot more time than our Silicon Valley timelines would have us believe. There are islands of the future (like San Francisco’s streets full of driverless cars) surrounded by an ocean of status quo (like the bureaucracy that still demands paper forms and rubber stamps). As investors and founders, we can’t afford to ignore that reality. It’s easy to drink our own Kool-Aid and think the whole world is on the cutting edge; in truth, much of the world is decades behind the frontier tech we’re excited about.

All this leads me to a more tempered outlook on the big question: Are Big Tech and their AI “frontier models” really going to eat the world? Is every venture-backable company doomed because Google’s or OpenAI’s super AI will swoop into every industry and take over? After my little journey from tech fantasy back to real-world reality, my answer is: Maybe one day, but not anytime soon.

The frontier model providers are certainly trying to build the all-powerful AIs that can do everything. And yes, in the echo chamber of tech insiders, it can feel like an inevitability that these giants will steamroll across the economy. But when you spend time in the actual economy, you see how many hurdles remain. Industries don’t change overnight. People, organisations, and regulations are slow to adapt. Legacy technology and entrenched habits are like sticky mud that even the brightest AI headlights can’t immediately blast through. And as that CEO reminded me over dinner, most companies don’t want to be tech companies - they want tech to serve them, not consume them.

From a venture capital perspective, this realisation is oddly comforting. It means there’s still plenty of room for startups to solve the unsexy, practical problems holding back the real world. If a corporation takes five years to implement new software, that’s not an AI issue - that’s an opportunity for someone to innovate in enterprise change management or create tools that simplify adoption. If hospitals are still faxing documents, there’s a lot of runway for healthtech companies to modernise data sharing (in fact, there are startups tackling exactly this). If most CEOs want to outsource their technology rather than build it themselves, that’s a screaming market opportunity for companies that make powerful tools simple enough for non-tech businesses to actually use. In other words, the sky isn’t falling for entrepreneurs just because frontier AI is advancing. The cutting-edge AI will find its place, but there’s a long, long tail of needs that require bridging the gap between old systems and new technology. Venture-backed companies that understand real customers’ slow pace and pain points can absolutely survive and thrive in the shadow of Big Tech.

That said, I’m not naive - there will be overlap. I can see a future where frontier models don’t just compete with vertical software but become the orchestration layer sitting on top of it, owning the UI and the user relationship. That’s a real threat to any startup that’s essentially a thin wrapper around an AI model. But the companies with genuine moats: proprietary data, deep domain expertise, regulatory know-how, outcome-based pricing, are much harder to disintermediate. The frontier models may own the top of the stack, but someone still has to build the messy, specialised layers underneath.

So, will Big Tech and their frontier models eat the world? Someday they might feast, but right now the world is serving a very slow buffet. The next time I get carried away by the hype (and trust me, I will - I love this stuff), I’ll remember to balance it with a dose of real-world context. Living in the bubble is fun, but stepping out of it is enlightening. Most of humanity isn’t beta-testing the latest AI agent or riding in a robotaxi to work. They’re filling out forms in triplicate, waiting on hold, and cursing at ancient software that still doesn’t work right. And that’s actually okay - It means our startups have time to find product-market fit in niches the giants don’t see or can’t be bothered with (yet). It means the future might be incredibly cool, but it’s not evenly distributed - and therein lies the journey (and the fun) of building technology that truly, eventually, eats the world.

Disclaimer: All views are my own