Code is cheap. Trust is not

OPENAI chief executive Sam Altman says software-as-a-service (SaaS) is entering its “fast-fashion” phase – cheap, disposable and quickly assembled. The stock market seems to agree.

Since the start of the year, investors have sold off SaaS companies en masse.

Shares of ServiceNow and Salesforce, two of the most prominent companies in the sector, have pulled back sharply, and the price of Atlassian shares have halved since the start of the year.

Panic has also spread to gaming platforms and cybersecurity companies over fears of a disruption in artificial intelligence (AI).

Roblox has dropped by nearly a quarter year to date, while stocks for Crowdstrike and Cloudflare were taken to the woodshed on Friday (Feb 20).

All of this happened before any of the accused had a chance to present their case. Here is what the fast-fashion analogy gets wrong: Enterprises are not buying T-shirts.

The visible and the valuable

The market sees the visible layer – and it is not wrong about what it sees.

Code is getting easier to write. AI agents can handle complex tasks. You can “vibe-code” an IT service-management module over a weekend.

After all, enterprise procurement does not work like consumer apps. Security reviews, vendor assessments, compliance requirements and multi-year contracts exist, as well as deeply embedded workflows that take years to build.

In this sense, the software layer has become commoditised.

But the institutional knowledge, implementation expertise and trust have not. This pattern holds in every sector that the market has condemned.

The technology layer is indeed getting cheaper.

But the trust layer beneath it is getting more valuable, precisely because the technology layer is becoming ubiquitous. When everyone has the access to the same AI tools, the differentiator is no longer the code. It is everything else.

The prosecution’s case

To be fair, the concerns are not frivolous. Anthropic’s release of AI agent Claude Co-work triggered a massive sell-off in SaaS stocks.

Here is a new breed of agentic AI – autonomous software agents capable of handling complex professional tasks – encroaching upon the territory of established platforms.

Meanwhile, Google’s Project Genie announcement spooked gaming stocks, sparked by fears that AI could recreate entire games in a few prompts.

The market has taken all of this to reach a swift conclusion: The incumbents are obsolete.

Swift conclusions, however, have a poor track record in the stock market. And the problem with passing judgment this quickly, is that the evidence has not even been presented yet.

The evidence the market dismissed

While the market was passing its verdict, these companies were busy reporting their results.

ServiceNow’s revenue grew by 21 per cent year on year for its latest quarter; Atlassian’s top line rose by 23 per cent over the same period.

On the gaming side, Roblox performed even better, registering a 36 per cent growth in revenue and 55 per cent higher bookings in 2025 compared with the year-ago period, while delivering more than twice the amount of free cash flow.

Meanwhile, its daily active users rose by 69 per cent to 144 million.

The gaming platform’s moat does not lie in game-creation tools – which AI could theoretically replicate.

Roblox believes that its advantage lies in its multiplayer cloud infrastructure, its social graph and the 144 million people who show up every day.

Simply put, these are not the numbers of companies under an existential threat.

The “trust evidence” is even more compelling. ServiceNow’s customer-renewal rates came in at 98 per cent. In a quarter where the narrative said customers would flee, virtually none of them left.

Evidence also indicates that AI is additive to SaaS businesses, not destructive. Atlassian’s cloud net revenue retention stayed above 120 per cent for the third straight quarter – a sign that customers are buying more, not less.

Across thousands of Atlassian Jira customers, those using AI code-generation tools create approximately five per cent more tasks, have five per cent higher monthly active users, and expand their Jira seats five per cent faster than those who do not.

In other words, the customers most adopting AI are using more of Atlassian’s platform, not less.

Its co-founder Mike Cannon-Brookes explained the logic: Efficiency does not shrink roadmaps, it expands them. Said another way, teams that finish their to-do list do not stop; they come up with more things to build.

And here is the most compelling evidence – the supposed disruptors are becoming customers of the incumbents or are partnering with them.

Instead of competing with ServiceNow, Anthrophic is partnering with ServiceNow to help customers build AI applications and accelerate deployment.

The seat compression dilemma

Elsewhere, another fear has taken hold – if AI agents handle the bulk of the work, companies will need fewer human users, and therefore, fewer seats to pay for.

If so, it would completely break the SaaS business model: seat-based pricing.

ServiceNow CEO Bill McDermott addressed this head-on. He noted that within the company’s target market alone, an estimated 1.3 billion available seats exist – of which ServiceNow has barely scratched the surface.

Far from shrinking, its active user base is growing by 25 per cent year on year.

Furthermore, customers for both Atlassian and ServiceNow are not rushing to abandon seat-based pricing for a fully consumption-based model.

ServiceNow chief operating officer Amit Zavery noted customers want flexibility, but also predictability.

Hence, going to fully consumption-based pricing may be premature in many cases.

Finally, a detail observers may overlook is that both OpenAI and Anthropic, the very companies whose products are supposed to trigger seat compression, are themselves using seat-based pricing with consumption limits.

In other words, the companies building the AI are pricing their own products the same way the “doomed” SaaS incumbents do. Seat compression remains a potential risk, not a present reality.

Get smart: Measuring what matters

Not everything that can be counted counts, and not everything that counts can be counted. The market measures what is easy to see – the cost of code, the speed of disruption and the impressive demos on social media.

But the durable edge in investing has always been in what is hard to see: trust, relationships, switching costs and institutional knowledge.

The real question is not whether AI can replicate what these companies do.

It is whether anyone can replicate what these companies are – deeply entrenched platforms with thousands of enterprise relationships, vast partner ecosystems and years of institutional trust baked into mission-critical workflows.

Code will keep getting cheaper; trust will not. For the patient investor, that distinction is where the opportunity lies.

The writer owns shares in Alphabet, Atlassian, Roblox and ServiceNow. He is co-founder of The Smart Investor, a website that aims to help people invest by providing investor education, stock commentary and market coverage