The Software Apocalypse? Not So Fast.
Just ask corporate IT
Last week, stocks for software-as-a-service (SaaS) companies like Salesforce, Workday and Hubspot dropped by as much as 17% in what analysts called the software apocalypse. Skeptical investors worry that AI-driven productivity gains will mean fewer licenses sold. Worse yet, some may skip subscriptions altogether and build their own.
AI coding tools are like steroids for developers: One engineer built a DocuSign clone in 30 minutes. Twitter co-founder Jack Dorsey built an encrypted messaging app in two days. Both would have taken months without AI.
Going custom is enticing: You get exactly what you want, without years-long implementations or vendor lock-in.
For everyday tasks and quick internal tools, AI coding is a game-changer. It’s when you start replacing the systems your business depends on that things get risky.
Before you cancel your Salesforce contract, talk to your IT team. Here are five questions they’ll ask.
Are our systems safe?
Security incidents are IT’s nightmare. In 2021, hackers crippled Colonial Pipeline’s systems for five days. Gas prices soared across the eastern US. Service wasn’t restored until the company paid a $5 million ransom.
AI coding tools make it easier to miss security holes. When AI generates hundreds of lines of code in seconds, developers tend to skim and approve without catching vulnerabilities. And non-coders who build AI apps don’t have the expertise needed to spot gaps.
Remember Moltbook? Last week’s viral AI platform was built in two days without writing a single line of code. Security firm Wiz breached its entire database in 3 minutes, exposing 35,000 email addresses.
Can we get payroll right?
Payroll isn’t hard because it needs fancy code. It’s hard because you need to handle 50 state tax jurisdictions, complex bonus structures, union rules and equity vesting schedules. That expertise has been baked into mature SaaS over years of edge cases. You’re not buying software, you’re buying the scar tissue of a thousand implementations.
Then there’s the regulatory burden of audit trails, data retention, GDPR and SOX compliance. Just wait until your company can’t close their books because a vibe-coded app missed a German cost-accounting rule.
Who counts as a customer?
Most people flinch at the words ‘software implementation’. The process can last years and force departments to rethink everything from how they enter an order to how they submit a timesheet. But that process is valuable because it forces everyone to get on the same page.
Take the definition of a customer: For Sales it’s active contracts, for Marketing it’s trial signups, for Finance it’s paying accounts. Those definitions serve their departments well, except when the CEO asks “how many customers do we have?” and gets three different answers.
Enterprise software forces you to have that uncomfortable conversation. Build your own tools and you might never have it.
Who do we call at 3am?
Building software is one thing, supporting it is another. User needs change, infrastructure gets old, third-party integrations break. You need someone to cover when the person who built it goes on vacation, and a transition plan for people who move on or retire.
Startups carry the same risks. Only 29% of software businesses launched a decade ago are still around. And with over 30,000 AI software companies flooding the market, the startup behind your system could disappear and there’s no one to call at 3am.
Can we trust our AI chatbot?
AI models require a layer of oversight that traditional software never did. Unlike systems that follow a fixed set of rules, AI outputs are unpredictable. Worse, they can be exploited. In 2023 a prankster tricked a Chevy dealership into selling a $76,000 Tahoe for a dollar.
Even without manipulation, AI tools can stumble when they face a new situation. A customer service chatbot that handles a normal Tuesday just fine might confidently give wrong answers about shipping dates on Cyber Monday.
Keeping AI in check requires strict controls, ‘red-team’ testing and monitoring. SaaS vendors are building dedicated teams to do this work. Go in-house, and you’ll need to do the same.
Long Live Corporate IT
Corporate IT gets a lot of heat for slowing things down. But this is because they ask the hard questions: Are we secure? Is everyone aligned? Who’s going to support this?
SaaS vendors have spent decades answering these questions. Some have gotten comfortable, relying on lock-in instead of innovating. AI will force them to compete harder for your business, and Wall Street may be right to price this in.
But a software apocalypse? The people who actually keep companies running know better.
Dad Joke: What did the HubSpot engineer tell his precocious teenage daughter? Don’t SaaS me 😂
Did the $1 Chevy deal actually go through? That’s wild. 😂🤯