AI has made building software easier than ever. Running a software company is another story, writes Kristian Waller, chief customer officer of media platform Rubii.
Over the last 18 months, we’ve watched a fascinating shift take place.
2025 was dominated by exploration. The entire media industry started experimenting with AI – testing tools, building proof-of-concepts and asking what was possible.
This year, the conversation has changed. The question is no longer “Can we build it?”, but rather “Should we?”
AI has dramatically lowered the barrier to building software. Today, almost anyone can connect APIs, generate code, build AI agents and automate workflows in a matter of days. That’s an incredible shift, but while AI has made building software easier than ever, it hasn’t actually made owning software any easier.
And that’s where many media organisations are about to discover the real challenge – the first version isn’t the hard part – version 147 is.
There is a growing assumption that every business should build its own tech. After all, it’s easy, right?
Not quite.
Don’t get me wrong – for some organisations, building their own solution is absolutely the right strategy.
If software is your product, or if building proprietary technology creates genuine competitive advantage, investing internally can make complete sense. But for many businesses – particularly media organisations – software isn’t the product, it’s an enabler.
The danger is that what begins as a helpful internal tool quietly evolves into something much bigger – a second business. Every custom platform creates an invisible business behind it – the business of maintaining the thing you’ve built.
Building software is only the beginning
The AI workflow that impressed everyone during a leadership workshop suddenly needs governance. The dashboard that saved hours every week now requires maintenance. The integrations that worked perfectly six months ago stop working because an API changed. The automation built by one enthusiastic employee becomes business critical – and then that person leaves, and so too does the knowledge.
What looked like a shortcut gradually becomes a free-flow of technical debt – and technical debt doesn’t care whether the software was written by developers, consultants or generated by AI.
The hidden costs rarely appear in the business case
When organisations estimate the cost of building software, they often budget for development, but what they rarely budget for is ownership.
Because once software exists, someone needs to own a very long list of tasks in order to keep it existing…Security updates, infrastructure, performance, monitoring, user permissions, authentication, documentation, support, bug fixes, product management, roadmap prioritisation, training, testing, compliance, AI model updates and API maintenance.
None of those disappear after launch. In fact, in many cases, they become the majority of the work.
A range of studies show between 60–80 per cent of a software system’s lifetime cost occurs after launch, through ongoing maintenance, support and adaptation.
Building version one is often the cheapest part of the journey.
There are even more factors to consider too.
For starters, AI isn’t free. Much of today’s excitement revolves around AI agents – and rightly so – they’re incredibly powerful. But deploying AI at scale introduces an entirely new layer of operational complexity.
Suddenly you’re managing costs associated with LLM APIs, prompt engineering, guardrails, hallucination monitoring, model updates, vector databases, embedding costs, performance optimisation, usage monitoring… The list goes on!
The very first demo might cost a few dollars, but running thousands of AI workflows every day is an entirely different equation.
Before you know it, you’re chewing through more LLM tokens than a Labrador through tennis balls.
The buck stops with you
The moment you ask customers or clients to log into software you’ve built your responsibilities fundamentally change. You’re no longer simply a media business – you’re also operating as a software provider. That means you’re now responsible for security, identity management, access controls, backups, disaster recovery, privacy, compliance, monitoring and incident response. The list is long, it’s constantly evolving, and none of it is optional. None of these creates competitive advantage, but every one of them carries a hell of a lot of risk.
And when internally built software fails, clients don’t blame the software – they blame you!
Like it or not, your technology becomes part of your brand, and it will be your reputation which absorbs the damage. Client trust is much harder to rebuild than software.
Just because you can build something doesn’t automatically make it the best use of your time, capital or expertise. Every hour spent maintaining internal software is an hour not spent serving customers, improving or innovating your product, winning new business or growing revenue.
Building software is one decision but maintaining it becomes an ongoing commitment. The real question is not whether your team can build it, but rather whether maintaining software is the highest-value use of their time. For many organisations, the smarter move isn’t building everything from scratch, it’s building where you create genuine competitive advantage and buying where specialists already excel.
There’s no doubt AI has democratised software development, which is incredibly exciting – but it hasn’t yet democratised software ownership.
The organisations that will thrive over the next decade won’t necessarily be the ones building the most internal tools, they’ll be the ones that understand where technology creates differentiation -and where it simply creates maintenance.
The winners won’t ask “Can we build it?”, they’ll ask something far more strategic… “Should we own it?”
