Future of the Industries
I was recently at the NXT BLD+DEV conference in Westminster for construction tech startups. I’ve attended this conference annually for the last 4 years and this was its 10th anniversary year.
Until now, it had always been an exciting date in my calendar, buzzing with optimism for the future of our space.
Construction was largely left behind by the SaaS revolution. The daily reality is that most of us remain tied to 1990s desktop software on Windows PCs. NXT BLD is a global role-call of rebellious construction-tech upstarts and the last decade has seen these firms organically develop in both maturity and momentum.
This year, that optimistic atmosphere had changed. You’ve already seen the punchline coming; founders are facing insecurity stemming from the unstoppable rise of AI – what it means for us and our sector. If nothing else, it certainly feels like an existential threat. Most of these firms are B2B (I’ll get more into what that means later).
All year, I’ve been thinking deeply about the kinds of impact AI might have on different parts of the workforce. What follows is my best guess as to where all of this is going.
Contents:
- Tech sector business models
- Professional services firms
- Physical labour
- The hollowed-out company
- The headless Internet
Tech sector business models
Business-to-business (B2B)
By size and revenue, the largest part of the tech sector exists as B2B enterprises. Often called SaaS firms, they sell licenses to their software and platforms, which are often just part of the cost of doing business for their customers.
I predict that these kinds of firms will be hit the hardest; some are already calling it the SaaSpocalypse.
At every commercial juncture, people on all sides can be seen sizing each other up, quietly wondering if they can automate away everyone else and capture that value for themselves. I see this as being spurred on by an implicit job insecurity; “If I can become the predator, I’m not the prey”.
It’s not surprising, really. The world of work is full of companies trying to find any commercial edge that they can. Our moats are dissolving faster than we can ever hope to build them. Software licenses constitute a large commercial outlay for an organisation; vibe coding potentially lets orgs reduce their cost, complexity and commercial reliance on tech firms. It’s a no-brainer!
This makes SaaS platforms vulnerable. They seem extremely uncompetitive compared to vibe-coding your own technology now.
- Platforms host competing companies together. They share feature parity instead of having access to some special secret-sauce that gives them a true competitive edge. The direction that platform’s development moves in is always a compromise between you and all of your competitors’ needs.
- All customers are tied to the platform’s roadmap and development speed, which is always slower than it would be if features were just developed in-house. The feedback loop can be much smaller, faster and more secure.
- Customers are locked into annual price rises over which they have little-to-no control. You’re just another inhabitant of someone else’s walled garden.
- Customers need to share their valuable IP regarding products, materials, processes, patent information etc with an external company. This requires a lot of trust in a SaaS company, and faith in their data security practices.
In reality, customers never really want all features of the full platform. Platforms are broad and complex beasts. It’s commonly said that your customer only buys your software for 3 or 4 features – it’s just that each customer wants a different set of features.
It’s not great on the other side, either: SaaS companies spend most of their time, money and effort on the construction and maintenance of the platform itself – data security, financial regulations and internationalisation, performance, reliability, legal requirements. Very little of the effort expended is in leveraging how the firm’s business domain expertise can bring value to their customers.
As a potential SaaS customer, your cost-benefit analysis is clear: hire your own developers and build what you need in-house. It will be cheap, probably fairly fast, exactly the way you want it, with a low ongoing maintenance cost. Not to mention, leading such a project will do wonders for your career and professional reputation! It’s a rare win-win-win scenario.
As for the “BIM 2.0” startups I mentioned in the intro (at NXT BLD / DEV), the vast majority of these are B2B SaaS companies. It’s very hard to know what happens to these next. Some will be acquired and get boiled down into features within a wider software portfolio. But the days of firms like Spacemaker being acquired for $240m seem long behind us now. For many of us, that’s a fairly bitter pill to swallow.1
Business-to-customer (B2C)
Firms that offer their software directly to regular consumers are different, and I believe they’re less likely to be hit as hard.
Firstly, your average customer or consumer is generally less technical than someone performing work within a business. They have a lower tolerance for technical nonsense and they’re probably using your app on a phone or a tablet. There is a non-trivial amount of hassle involved to get an app approved and made available on an app store (or even just loading it onto your own personal device).
Furthermore, just because they can, it doesn’t mean that everyone will develop their own apps. If you vibe-code the new Facebook or Spotify on your phone, nobody will care or use it. The value of those platforms lies elsewhere entirely.
I expect B2C will have a smoother ride. The tools are now at their disposal to finally handle UI, UX and customer support as they would ideally like to. Generative image, video and music could do wonders for your marketing if it’s high-enough quality to break through the slop barrier.
Consulting
The consulting space is by its nature pretty flexible. This may come to be of real benefit to those individuals with the skills and the experience to act as tech mercenaries in the years to come.
In an age where many business leaders are unable to keep up with the rapid pace of change in AI (frankly, who can?), being able to offload these new challenges to a safe pair of hands seems well worth it.
There’s all the benefits of vibe-coding your own solution (keeps your IP in house, you get a custom service) but with none of the full-time hiring nightmares. Will consultants cost you more? They will. But if what they’re building you is adding enormous value and helping to build your moat, then it can easily be worth the extra uplift. You may even be able to afford a higher level of skill than you’d get if you dropped £100K on a new employee’s salary.
Google Disco and where tech is heading
People often discredit “vibe coding” as being unfit for serious projects. I’ve heard the sentiment confidently declared all over – “it generates shabby software that can look great but is poorly-structured”. Codex, Cursor or Claude2 will soon lead you [a fool] into a swamp of unmaintainable chaos that you’ll come to regret. If only you’d toiled away in the mines of an Agile SDLC.
I think this is remarkably short-sighted and perhaps even denial as a coping mechanism. The term “vibe-coding” is only one year old. Automated software generation has moved forward by leaps and bounds in the last 12 months.
I believe that by 2027, AI will be capable of writing better software than most professionals could ever hope to output. The trend is heading rapidly towards instant, near-free software whenever we want it.
A great example of this is Google Disco. Released in May 2026, it stands as an affront to the entire idea of software development as a career choice. In short, it lets you turn any of your Chrome browser tabs into an ‘app’. You can prompt it to edit and expand the app, however you want, in real time. When you’re done, just close the app. It’s effectively disposable software. That was once as unthinkable as building a disposable airport. Is it time to grab our pitchforks, yet?
Where value still remains
There’s a phenomenon I often notice, when someone’s internal model of something jars with hard evidence that they were actually way off the mark.
When we think clearly, we’re able to update our priors pretty much instantly. New evidence leads us to rapidly adjust our thinking.
But when we’re emotionally tied to that certain something, this process can slow down massively. This could be finally seeing the ‘true face’ of a friend or a lover.
I believe we’re in one of those moments right now with our relationship to technology and the identity of career software developers. It’s a bit like that running gag in Looney Tunes, when Wile E. Coyote runs off the edge of a cliff. There’s always a brief moment when he looks around, suspended in mid-air, before reality sets in.
But it’s not all doom and gloom. There is still enormous value in tech. It’s just not necessarily in the same places as before:
- Firstly, if your moat was purely built of software, you’re in trouble. Lots of people have been saying this. Sheer software development complexity is no longer meaningfully defensible.
-
There is still value in rare IP. Take Netflix or Spotify for example. Signing such a wide range of agreements with production studios for upcoming and legacy rights is a deeply political human art. It’s true that the powerhouses of IP have rapidly consolidated in the last few years, but those IP rights are extremely hard to gain access to and they remain safeguarded by humans. While many companies will want to keep their IP under lock and key, others are in the business of selling access and so require distribution platforms to leverage the wealth that they own.
- There is value in incumbency. If you’re already a highly-established player with thousands (or millions) of customers, you’re in luck. In a world that’s drowning in AI slop, we need large organisations to start taking on the role of gatekeepers against a tsunami of generative crap. Apple’s App Store sets the gold standard for this. They are able to maintain an incredibly high level of quality across all apps, with 24-hour human-review turnarounds. For those willing to take on this type of challenge, there is value in the reputation that is to be gained here.
Professional services firms
It’s both good and bad news for the world of professional services, the LLPs. The bad news is that they are likely to be squeezed; AI will see their competition get more intense and their clients become more demanding, expecting more outputs, faster and for less.
The good news is that there’s no way to fully automate these firms away. Humans still have one irreplaceable quality that AI simply cannot – they can be sued.
Humans in consultancies already act as a sponge for risk, taking ultimate responsibility for the information and sign-offs they produce. People can be insured, taken to court, fined or even presented with jail time.
The legal sector is slow – lawsuits against tech/AI firms will remain niche and it will take many years for meaningful case law to really establish itself in this fast moving field. Nobody will insure an AI against its own misdeeds. For now, at least.
There is also opportunity in this. Since humans are required for sign-off, it might finally be the time to start taking risk incredibly seriously. The next wave of popular SaaS products could very well be risk assessment studios. Lots of pretty dashboards and red flashing lights. You get the vibe.
Physical labour
The obvious conclusion one reaches is that all physical work is safe from AI, at least for the near future. As long as it remains screen-bound, the builders, hair stylists and chefs shall remain employed. We all know that robots are eventually coming, and they’ll probably get pretty good at even the fiddliest of manual tasks (robot surgeons are already a thing).
My best guess is that any tasks that are manual, involve adaptation to highly custom requirements and where the outcome is highly sensitive will be safe from AI in the longer term.
I would have once said that opening a child/daycare centre should be pretty AI-proof. But actually I could be wrong there. A great deal of light-touch office jobs (HR, marketing, socials, finances) are likely to be automated away within the next five years. The vast majority of those working in these areas may soon decide that raising and educating their children themselves is a pretty ok Plan B after all. It occupies a lot of time and effort and can be a deeply meaningful kind of work.
Now, if I were to bet, I’d focus in on something more like historic building renovation. The clients often have deep pockets, are traditionally-minded and the work is manual, custom and highly sensitive. Not to mention the worksites are often pretty beautiful.
The hollowed-out company
In April 2026, Salesforce announced a new offering, Headless 360. They’ve been the largest company to offer their entire platform as essentially a service for agents.
The impact this new direction of travel could have on tech is profound. Very quickly, we realise that most of software (UI/UX) exists only to aid humans in using it. But agentic AI is perfectly happy with an invisible MCP server. Soon, entire platforms might be reduced down to a suite of webservices, quietly humming away and making money.
The same phenomenon is happening internally within companies. It’s a bit shocking when we realise how very many of our ceremonies, functions and positions exist purely to facilitate the needs of other humans working together. When we start using AI to boil away our organisations, we’ll be left with a very thin reduction at the end of it all.
The headless Internet
If we follow this same thinking through, we arrive at the idea of the Headless Internet. The Internet becomes an invisible layer, solely traversed by agentic helpers, with no humans eyeballs left. This is sort of what the Rabbit R1 was getting at back in 2024.
It’s a strange idea. As an architect who grew up on the developing Internet in the early 2000s, it’s always been a kind of place to me. We speak about it as such. Websites were something we visited. We surfed the web. Each website was its own little world. The idea of that alternate universe collapsing in on itself feels very strange and a little sad, like we’re losing something important.
I hope this is wrong; it’s entirely possible that the creatives among us will use AI to bring fun back into our online lives The web has long-mourned its early, creative, fun days, when most websites used a Flash plugin. Maybe we’re heading back that way; it might just be the good ending.
Finally, in a web that is largely traversed by agents, what happens to ad revenue? I read an article recently about a company that generates fake podcast-slop about knitting that somehow still runs real ads. There are over 3000 episodes. All of it hallucinated. Is anyone listening to this? Can the advertising economy that’s paid for the entire Internet continue to exist in this new paradigm? Who knows.