Software and AI: the end of the uniformity illusion

The biggest surprise of the past two weeks may not have come from the hyperscalers but from software. For months, a narrative has been building: generative and especially agentic AI could destroy part of SaaS value by automating functions historically sold as seats, modules or specialised workflows. Often exaggerated, this narrative has nevertheless produced very concrete equity effects. AI is now widening the divide inside US tech, lifting some hardware names while pressuring several software stocks. For a family office, the message is clear, software can no longer be bought as a homogeneous block.

ServiceNow, the most telling example

On substance the quarter was strong: USD 3.671 billion in subscription revenue, up 22%, cRPO growth of 22.5%, 16 deals above USD 5 million in net new ACV, and customers spending over USD 1 million in ACV on Now Assist up by more than 130%. The company emphasised its agentic platform strategy, with an integrated context engine and the launch of Autonomous Workforce, including a service desk AI specialist resolving common IT requests autonomously. Yet the stock fell about 12% after-hours. Delayed government signatures in the Middle East weighed on the quarter, but the market mainly fears that part of historical enterprise software usage will be redefined by more general agents. ServiceNow is reassuring by emphasising the shift of new business toward non-seat-based pricing. That is exactly where the battle is being fought.

IBM, the paradox of a beneficiary that worries

The group reported USD 15.9 billion in first-quarter revenue, up 9%. Software grew 11%, infrastructure 15%. Management insisted that AI remains a tailwind, particularly when customers seek to orchestrate, deploy and govern AI across hybrid environments. The stock still dropped 6.5% after the release, as several investors worry that parts of software modernisation and consulting work will be progressively automated by more autonomous tools. IBM already benefits from AI, but the market simultaneously fears that AI could erode parts of its model.

SAP, a useful European counterpoint

The first quarter showed a current cloud backlog of EUR 21.9 billion, up 20%, cloud revenue up 19%, cloud ERP suite up 23% and rising operating profitability. Management links the performance to Business AI dynamics and the firm's ability to gain share as customers extend their use. ADRs rose 5.4% after the release. The market is not saying enterprise software is dead. It is saying that platforms with critical workflow, functional depth and a hard-to-reproduce data layer can still be rewarded, especially when they show AI embedded into a product already installed at the core of business processes.

Thomson Reuters and trusted AI

On 5 May, the company reported quarterly revenue of USD 2.09 billion, up 10%, above expectations, and reaffirmed its annual outlook. So-called fiduciary-grade AI is gaining traction in legal, tax and compliance use cases. For UHNW clients, the lesson is valuable: in fields where accuracy, auditability and legal accountability matter more than raw generation speed, some established players may emerge stronger by embodying the trust layer that large models do not provide on their own.

Four words to frame the lens: data, workflow, price, trust

The more proprietary, structured, historical and legally sensitive the data, the harder it is to disrupt. The closer software sits to a critical operational function, the more defensible it becomes. A vendor still locked in a per-seat model can be more vulnerable than one shifting to usage, ACV or outcome-based pricing. Finally, in regulated fields, useful AI is not the cheapest or the most spectacular, but the one that reduces error risk and documents its output. That is what paradoxically links SAP, ServiceNow and Thomson Reuters, despite very different short-term market reactions.