Software Got Cheap to Build. Now What?

Software got cheap to build. Now what?

AI-assisted development has cut the cost of shipping a product by orders of magnitude. Which sounds like great news for small software companies — until you realize it's equally great news for the other ten thousand small software companies...

Meanwhile, the buyer is still the same — they still buy software the way they always did: reluctantly, slowly, with incentives not always clearly defined. Drowning in offers now which all look the same, still cost money to implement and support, and might not survive the year. Their incentive is to wait and buy nothing ))

Market produces noise faster than quality and trust. So what's your business model, startup? ))

Will SaaS die?

Probably not, but "SaaS as a business model" and "SaaS as a moat" are quietly divorcing. The subscription mechanics will survive because recurring revenue is just good for accounting. What might not survive is the assumption that a working product plus a pricing page equals a business.

When your competitor can rebuild your feature set in a quarter, the product itself is table stakes. What's left is only:

• •your data,
• •your workflow lock-in,
• •your distribution,
• •and — increasingly — the buyer's trust that you'll still exist in 2028.

Will it all collapse into custom development and services?

Partially yes, and this is the interesting part of the story.

If AI makes bespoke software 100x cheaper to build, the old trade-off — "custom fits perfectly but costs a fortune, off-the-shelf is cheap but fits badly" — starts to dissolve. I'd expect a hybrid approach to dominate: a productized core (maintained and secure, regulations compliant) with AI-accelerated customization at the edges. Old good implementation engineering model which can be done 100 times faster.

The pure services model, though, still has the old disease: it doesn't scale, and margins get eaten by every hour billed. Services win deals, products win companies.

Attention and trust get scarcer

When production is abundant, the bottleneck moves to credibility. The small IT company's real asset isn't the codebase — it's deep knowledge of one boring, specific industry problem, and the relationships that come with it.

Vertical beats horizontal. "CRM for everyone" is dead on arrival. The future belongs to whoever makes the buying decision feel safe (which is not new of course).

Small companies that pick one narrow problem, get very close to their customers, ship a product but sell an outcome, and treat implementation as part of the product rather than a painful afterthought — they'll do fine. Giants can't be that close to anyone.

That's the opportunity.