Why Now Is the Worst Moment to Build Software (& Apps) In-House

As a B2B SaaS founder and CEO, I'm biased. But I also spend a good part (too much?) of my waking time thinking about this topic, so hear me out. 

With the inflection point that we hit with AI in software development at the end of 2025, I started to seriously question whether SaaS products could hold their ground against in-house builds. In the past month, my viewpoint has shifted by 180 degrees. 

I'm a computer scientist and mathematician by training, which means I tend to oversimplify the world with number games. That's exactly what I'll be doing below.

But before the math, a distinction. You should always build your core software in-house – quoting Jeff Bezos:

“Focus only on what makes your beer taste better”. 🍻

This essay is about context - the 90% of software required to run your business but which isn’t your core competence, such as a CRM or a Feedback Analytics Tool. 

🔥 Provocative take: there has never been a worse moment to build in-house versus buying context software. 

Here's why. 

Axiom 1: Surviving SaaS vendors are better at building their product than a buyer could be

My thinking starts from a simple premise: as a vendor for a specific software product, you have to be better at building and providing your service than a potential buyer would be in-house.  

The proof is in the mere existence of the software company. Assuming a free market that functions at least to some degree and isn't completely broken (e.g. by regulation or monopolies), you cannot survive as a vendor over a prolonged period – years – if your service is not materially better in some way than what a buyer could build themselves for the same money. That difference isn't just about features: it's the sum of the feature-set, reliability, customer service, security, and speed. 

Let's assume for the sake of argument that this difference in “output value” is very small: Say an in-house team produces 100 units of value per year and the vendor delivers 110 units. A modest 10% edge. Remember this number: it matters later. 

Note 1: This 10% estimate is almost certainly too low. In B2B, the path of least resistance is always to do nothing. To convince multiple independent stakeholders to buy - and keep buying – over several years, the vendor's perceived value-add must be vastly higher than a mere 10% edge to overcome the friction of purchasing. 

Note 2: Vendors don't produce "better" output because they magically hire smarter or more capable engineers. The advantage comes purely from focus: an entire company and its resources are singularly obsessed with solving this exact problem. 

Note 3: A common pushback is: "But we don't need all their features!" True, you rarely use 100% of any software (think Microsoft Word), and vendors already price this in. The 110 vs 100 unit comparison above applies strictly to the relevant feature set that actually matters to your business. 

Axiom 2: In software development, AI is not the great equalizer: it's a multiplier 

Many people now say: "The playing field is levelled. You can just build in-house what the vendor took years to build." The crucial flaw in this thinking is the assumption that the vendor hits pause while in-house teams are catching up. 

The reality: yes, it has never been easier to produce software. But software vendors are not twiddling thumbs. Believe me, we’re not. 

Talking to other engineers, product folks and SaaS CEOs – they are in code-red mode – pushing hard, releasing features at an unprecedented pace, rebuilding entire product experiences around AI like their survival depends on it. Because it does. 

So yes, a talented team could build the current state of a product like Caplena to 80% within, say, one-year with 10 talented engineers & product people - where it took us 8 years. But here's the flaw: in that year, our product will have been completely transformed in its capabilities. And the progress is exponential. 

This applies to people too. The best engineers who produced “300 units of software” per month before can now produce 900 units. Those who produced 100 units are now at 300. AI multiplies existing advantages. The gap widens. 

Axiom 3: A step-change in software development has happened 

This wasn’t a gradual trend: Something broke through in the last quarter of 2025.

I assume this one is mostly undisputed. It has been felt by almost everyone in the software development ecosystem.  The combination of frontier models, agentic coding tools, and AI-powered workflows hit a tipping point.  

AI-native companies that were already deep in this world. The ones who had the codebases, the domain knowledge, the feedback loops, who could absorb and deploy these capabilities most immediately. And we did.  

Axiom 4: B2B software has never been cheaper 

There has probably never been a better period to buy B2B software. 

Several forces are compressing prices simultaneously. Market uncertainty has brought down multiples. Competition among vendors is fierce – you better be shipping features in two weeks, not in two months, because in two months they might already be outdated. And the entire industry is still figuring out what pricing even looks like in an AI-native world. Credit-based pricing, outcome-based models, hybrid models – vendors and customers alike are experimenting. The new economics haven't settled yet and currently many vendors are eating their underlying AI / infrastructure costs. 

As a buyer, this is an excellent opportunity to lock in contracts at rates that won't last. The only challenge is to bet on the right horse – the vendor that produces a slightly better product than the others – because that difference is going to compound over time. 

What follows from this?

If you accept the axioms above, the outcomes are hard to avoid. 

Outcome 1: In-house teams lose ground, not gain it 

The simple math: Here's where my number games come together. If the vendor has even a small edge of value-output-per-money-unit-per year (remember 110 units vs 100 units), and AI multiplies the value creation for both at by 3x, the absolute gap between buy and build grows every year.  

And again, the 10% edge is probably an understatement (see axiom 1).  

The in-house team isn't gaining ground. It's losing it. 

Outcome 2: Expectations and capabilities of software are being rewritten 

This newfound efficiency in software development (Axiom 3) leaves the industry with two distinct paths: 

• Option A: We put in less effort to build software to similar levels as we have it today. 

• Option B: We maintain our current input efforts, but get vastly better software out. 

All my money is on B. Every major technology shift in history went this way.

👉 When fossil fuels dropped the cost of energy by an order of magnitude, we didn't just heat our existing homes for cheaper – we invented cars and airplanes.

👉 When broadband replaced dial-up, we didn't just download text emails faster – we built Netflix and Zoom. It is simply human nature to consume technological surplus to build more. 

Because vendors are pouring their AI-gained efficiency into Option B, the change in software from 2026 to 2027 will be staggering. Every vendor is scrambling to make their product smarter in every dimension. The rate of change will be very hard to match if this isn’t your full-time obsession. 

As a result, people simply won't accept the tedious, user-hostile, horribly manual software they tolerated for the past 20 years. Even in enterprises. Life is too short to spend it fighting frustrating interfaces.  

State-of-the-art and user-friendly software is no longer a nice-to-have for enterprises: 

• Talent demands it. If your stack feels like 2015, you won't attract the people building 2027. 

• The window to pull ahead is now: The gap between early adopters and laggards is never wider than during a paradigm shift like we are experiencing it currently. This is the maximum-leverage moment, the best chance to pull ahead, and the biggest risk of falling behind. Once the wave settles, as technology waves always do, the differences compress again.

But by then, the winners have already been decided. 

Outcome 3: Engineering talent should go where it moves the needle 

Let's assume a business has $100m in sales, and your vendor bill for a feedback analysis solution (or any other context software) is $80k a year. You put your engineers, product people and CX & management percentages on rebuilding that in-house to save money. Even if you miraculously reduce that software cost by half, you've moved the needle by exactly $40k. And that is assuming you’re fully accounting for internal costs, which, as Viktor Cessan lays out beautifully in this interactive calculator, is very often not the case. 

Meanwhile, every competitor is trying to figure out where AI has an impact on their core business. They're applying that same engineering talent to the thing that drives maybe 10% or 50% of their revenue. So how about multiplying a value bucket of $50m by a mere 1% instead? That’s $500k! 

Any exec putting engineering budget on the wrong initiatives is making a costly mistake. The opportunity cost is brutal. 

Outcome 4: The spend ratio between people and software shifts toward software 

We're dogfooding this ourselves. We have engineers (including myself) spending $400–600/month on coding tools and AI subscriptions because the ROI is absurd. A $100/month subscription on an engineering salary only needs to make that person 1% more effective to break even. The reality is closer to 2-3x. 

But notice: we spend massively on AI tools. We would never build our own development tools or our own AI models for software development (albeit being an AI company from the start..). People who spend their entire day on those problems will always be better at it than us. We use their tools to make our product better. 

Summary 

AI isn't leveling the playing field. It's amplifying existing differences. The people and companies with deep domain knowledge, engineering talent and established codebases are the ones getting the biggest multiplier. That's vendors. Which is great for buyers, because they get to profit from amazingly new or improved software. 

Buy now & negotiate. The window for locking in great B2B software deals is open. Vendors are competing harder than ever, prices haven't settled into their new normal yet, and product development is happening at unprecedented speed. Now the only challenge is betting on the right horse 🐴. 

So what should buyers do? 

Stop hiring thoroughbreds to pull a plow. The companies that dominate the next decade will be the ones with the highest Innovation-to-Maintenance ratio. 

Every hour your 10x talent spends rebuilding a feedback loop or a CRM is a gift to your competitors. Let the vendors fight their arms race; you should be the one reaping the spoils of their war. By offloading "context," you liberate your engineers to obsess over the 1% of code that actually moves your stock price. 

The ultimate competitive advantage is the organizational discipline to ensure your best minds are never wasted on a problem the world has already solved. 

If you're curious to see how we solved the problem of turning customer feedback into intelligence that all your teams can act on, do not hesitate to reach out to our friendly team.