When ChatGPT graced the world in 2022, the word ‘hallucinate’ immediately joined our everyday vocabulary.
Like with AI hallucination, today in business I think there is a severe case of strategy hallucination going on, post the advent of AI.
Led by the vibe-coding revolution, many career executives are starting to feel empowered to build their own tech.
The only issue? Most of them have no clue what they are doing and should stay well away from it.
What is software anyway?
Writing software code was never hard in theory, so I don’t get why that framing is being widely used today.
Yes, it was hardER pre-AI than today, but it was never HARD. Hence obsessive kids in their garages built the internet.
As a business, you could always hire developers and start coding – and that was especially true of big business.
So why didn’t every business become a software house with deep technology expertise?
It’s simple – because the ‘specialisation of labour’ is a real thing. Coding is not all software building takes.
There are several elements to software building, but non-tech people tend to focus only on the bold items below:
- VISION
- Problem analysis
- Feature prioritisation
- Taste
- CODE
- Test
- DELIVER
- Support
So what has the new world of AI changed?
AI has made ‘CODE’ and ‘Test’ easier. But in other areas, AI cheapens the feel with generic thinking, and detracts from solving the intended outcome in the most creative way (it goes for the most obvious).
And sadly, like good grammar, good software delivery skills require having them before you can see those that don’t. As the old saying goes “a little bit of knowledge is a dangerous thing”.
Building good software ain’t cheap
Now AI is here, and coding has become easier, the executive-level reality distortion field is in full effect. Companies are no longer protected from their own worst impulses.
For some context on software project realities – McKinsey and Oxford University studied more than 5,400 IT projects and found that, on average, projects run 45% over budget and deliver 56% less value than expected.
The main failure points here are cultural. AI is not going to solve that! Executives should beware.
Given this lack of knowledge and experience, I think we are at risk of seeing the most criminal misallocation of capital in modern capitalist history: lots of companies building the same thing in silos, paying 5x more for the privilege, and having to deal with the long term costs of their tech debt to boot.
Will somebody wake these people up from their mind virus, before it’s too late and billions in value has been burnt?
Software is dead, long live software
Contrary to commentary about the ‘death of software’ earlier this year, it therefore won’t surprise you to learn that I think AI-native software businesses will become more valuable in future.
Sure, these AI-native companies may suffer some short-term sales friction while the AI ‘build ourselves’ mind virus is in full effect.
But for those that weather the storm, on the other side, in a sea of shit point solutions, quality AI-native applications will sail above the rest.
Good self-built AI tech will exist for sure – it will just be limited to non-critical, non-data sensitive, non-customer-facing efficiency helpers. Basic ‘point solutions’ are smoked (who buys to-do list software today?).
But that doesn’t mean most software is cheapening. Certainly not the major software I see being tackled by the new wave of AI-native software and service providers emerging today.
If there is a proliferation of half-built and half-supported software, it will provide stark contrast to solid, third-party built, AI-native products.
Now ‘AI-native’ is the important term here. One thing hitting the world like a freight train of efficiency is agentic workflows. Agents are a real trend. They are growing and good software must accommodate that.
But who is more likely to accommodate agentic workflows? The software company, whose life depends on adapting to the new world, or the big corporate’s CIO team, whose overpaid and underscrutinised internal developers sit each day working 9-5, tinkering under the guise of ‘innovation’, while finding out about new ideas from their managed services provider?
Sure, we will all make small applications ourselves in future, but substantial software projects should continue to be run by third party, AI-native software companies, whose costs are amortised across thousands of users, and whose capabilities are multiples higher than the software available a decade ago.
Boards should not confuse the ability to code with the ability to build software – to have high taste, user intuition and the willingness to support products for 10+ years. That’s what software building is.
So if you are a company executive and it’s a major software project – do yourself a favour – stay away from the self-build and go and find an awesome AI-native software vendor to lead the charge instead. Your P&L will thank you later.