‘Headless software’ signals further AI-led shake-up

There are plenty of unresolved questions about what part today’s leading software products will play if AI reshapes working life as profoundly as many in the tech world expect. But at least the worst of the indiscriminate selling of software stocks — a bloodbath known as the SaaS-pocalypse — appears finally to have run its course.

The latest sign of the relief rally was Thursday’s 20 per cent share price jump that greeted Salesforce’s quarterly earnings report, extending the rebound since the stock hit a low point in June to around 65 per cent.

Chief executive Marc Benioff has been declaring the SaaS-pocalypse over almost from the moment it started. He was at it again this week, this time apparently with more justification, though his company’s shares are still a third below their late-2024 peak.

Software investors could also take heart from comments made by Dario Amodei, CEO of Anthropic. The company’s release early this year of AI tools aimed at replacing some traditional software products triggered the worst of the sell-off. “We are not interested in destroying anyone,” Amodei declared in an interview on CNBC about a partnership with Salesforce — apparent evidence that a new period of “co-opetition” has set in.

The broad rebound in software stocks over the past two months shows that Wall Street has at least set aside some of its worst fears. One was that customers would start to drop their current vendors and use AI to vibe-code their own software. This always seemed a stretch.

Another realisation is that some software companies are less prone to immediate disruption from AI than others. That is particularly true if they hold data that represents the “ground truth” about what is happening inside a company — making them so-called systems of record — and if they are closely tied into a company’s workflows,

For general-purpose tools without a tight link to essential operations, on the other hand, the SaaS-pocalypse is still in full swing. That includes companies like Airtable, a once-hot maker of productivity software that has just been sold for around a fifth of its peak value, and Medallia, a generic customer survey tool that recently left private equity owner Thoma Bravo nursing a $5bn loss.

But even if the threat to more entrenched software products seems less immediate than it did, the longer-term challenge remains. If AI agents start to take over some of the work currently done by humans, they will not need the kind of software in wide use today.

The result is what is known as “headless” software — products stripped of the user interfaces that until now have governed how they are used, allowing AI agents to access the data they contain directly. Salesforce declared it would go headless in April, while companies like Workday and SAP have followed with partial moves of their own.

For companies that have always charged based on the number of people who use their software, this presents an obvious business model risk. But an even bigger problem may be that much of their value to users has come from the UI: the forms, dashboards, analysis tools and other features that determine how work is carried out and managed. Once AI agents skip the UI to tap into the data directly, much of that functionality will be delivered in new ways that have little to do with the old, human-centric software.

This risks turning today’s essential software into tomorrow’s back-end databases — necessary, perhaps, but no longer the sort of technology that commands a premium.

Recommended

Efforts to resist this are only just taking shape. For software companies, it means expanding the system of record function to turn themselves into essential authorities that control how data is accessed and used. Another is to expand from simply maintaining records to prompting actions, marshalling agents to control how work is done.

For Anthropic and OpenAI, meanwhile, moving further up into the application area as they look to embed themselves more deeply into their customers’ operations has become a priority, particularly as they try to outrun the threat of commoditisation in their core AI models. Co-opting the distribution of a company like Salesforce is one way to accelerate that, but it doesn’t resolve the inherent competition.

To judge from Wall Street’s reaction, investors are at least finally willing to give some software companies the benefit of the doubt. There is still much to be fought out before the final shape of AI-enabled work becomes clear.

[email protected]