Skip to the stories
Microsoft 365 & SharePoint

When Your Platform Starts Writing the Requirements, Revisit the Build-vs-Buy Call

  • Nickola Naous
  • 572 views

A website rebuild that bent requirements around a commercial CMS raises a wider question: has AI changed the build-vs-buy math behind your older platform choices?

When Your Platform Starts Writing the Requirements, Revisit the Build-vs-Buy Call

The most useful warning sign in this story isn't the missed deadline. It's the moment the tool starts deciding what the work should look like. Nickola Naous, co-founder and CTO of Flashcloud, a web hosting company, describes exactly that pattern in a rebuild of his company's website, and he argues it applies well beyond websites.

A rebuild that moved the wrong way

The goal of a site rebuild is simple: a better experience for visitors and an easier job for the people who run the site. Flashcloud commissioned an outside agency to build the new site on a well-known commercial platform. Instead of improving things, the project ran months behind schedule, and once it got back on track, the new site was slower than the one it replaced.

Overruns aren't unusual. What Naous missed at the time was how the project was drifting. Somewhere in the middle of the build, the team started bending the plan around the platform. Features were simplified, interactions were dropped, and small compromises piled up because each one looked like the practical route to launch. By the end, the site resembled what the platform could comfortably support, not what the team had set out to build. The workarounds also made routine updates harder than they should have been.

Requirements written backward

Here's the thing about that drift: it rarely shows up as a single decision. It shows up as a sentence in a meeting. Someone says, "We can't do that in the system," and the room adjusts the requirement instead of questioning the system. One instance is reasonable. A pattern is a problem, because at that point the platform is setting the boundaries and the business is following them.

Teams in this position often don't notice because the trade-offs look small one at a time. The cost of switching platforms is obvious: migration effort, retraining, disruption to live work. The cost of staying is spread thin across slower delivery, awkward workarounds and features that never ship. That imbalance is why many organisations keep software that made sense when it was bought but no longer fits how the work is done.

The fix created a new problem

Naous's team moved to a static framework and changed hosting, and that solved the performance problem. It also created a new one. The commercial platform had done one thing well: it let non-technical staff edit and publish content without a developer in the loop. A faster, more controllable site that only engineers can update is a trade the business might not accept.

A few years ago, that tension would probably have ended the discussion. Building and maintaining a content management system is real work, and most teams would have accepted the platform's limits rather than take that on. Naous says AI assistance changed the calculation. He built the replacement with AI help, not because he is a career developer, but because the cost of building certain internal tools has dropped far enough that old buy-versus-build decisions deserve another look.

Assumptions with an expiry date

This is the part I'd push hardest on for anyone making technology decisions beyond a single site. Many build-versus-buy choices were made on assumptions that were accurate at the time: how long a custom build would take, what it would cost, how hard it would be to maintain. Some of those numbers have moved, especially where AI is part of the build. Yet the decisions usually stay put because nothing forces a fresh look.

Vendor contracts get reviewed at renewal. The reasoning behind the original choice almost never does. So a platform bought five years ago can keep running on a set of assumptions nobody has tested since.

That said, this isn't an argument for building everything. Naous is clear that buying is still the right answer most of the time, and a custom tool nobody maintains can become a bigger headache than the subscription it replaced. The point is narrower. The reasoning behind a platform choice has a shelf life, and that's worth checking when the platform starts forcing compromises on important work.

Two questions for your longstanding platforms

Naous suggests two questions for any platform that's been in place for years:

First, are we changing our requirements to fit the tool, and when did that start?

Second, if we were choosing today, with what building now costs, would we make the same decision again?

If the honest answer to the second question is no, that doesn't automatically mean migrating. Migration has its own cost, and staying can still be the right call. But the decision then deserves a formal review rather than sitting on autopilot. Otherwise the platform survives by default, not because it still fits.

In practice, that review can start small. Pull the list of workarounds and deferred features from the last year of change requests. Count how many of them trace back to a platform limitation. Then price the honest build option against the current renewal, using today's AI-assisted effort estimates rather than the ones from the original business case. Nothing about that requires a decision on day one. It just means the next renewal conversation starts from evidence instead of inertia.

The lesson Naous takes from his own project is that the mistake wasn't picking the platform. It was failing to notice when the platform had stopped being the right one.

Source: Nickola Naous · www.informationweek.com

Neighbouring cells