Palantir, ServiceNow and Toast: The AI Survivors Built on Execution

Why Palantir, ServiceNow and Toast Are Positioned to Outlast Cheaper AI

A Forbes España analysis co-written with Michael Lee Strategy founder Mike Lee argues that the most durable AI businesses will not simply generate answers. They will own the connection between AI insight and real-world execution. As large language models become cheaper and more interchangeable, the investment question shifts from who builds the model to who controls the operational layer where AI output becomes action.

Three companies are presented as examples of that thesis: Palantir, ServiceNow and Toast. Palantir organizes sensitive data and embeds AI-driven decisions into government, defense and commercial operations. ServiceNow provides the workflow and governance layer that routes, approves and documents work across a large enterprise. Toast operates an integrated software and payments platform for restaurants and local retail, using AI to improve marketing, staffing and menu performance.

Their latest reported numbers illustrate the momentum. Palantir's revenue reached $1.93 billion in its most recent quarter, a 93% year-over-year increase, with net dollar retention of 157%. ServiceNow reported $3.99 billion in total revenue, up 24%, and said its AI business surpassed $1 billion in annual contract value. Toast grew annualized recurring revenue 25% to $2.4 billion and added a record 9,500 net new locations.

The common thread is not that these companies own the most advanced models. It is that they already sit inside the operational loop where AI output becomes a decision, an approved workflow, a transaction or a restaurant promotion. That position, the analysis argues, makes them harder to replace as AI becomes cheaper and more capable.

The Execution Layer: Decisions, Workflows and Restaurant Transactions

Palantir: Owning the Decision Environment

The bull case for Palantir rests on more than growth. Its Gotham, Foundry, AIP and Apollo products span government mission support, enterprise data operations and AI deployment across classified and commercial clouds. The company is cited as part of the National Geospatial-Intelligence Agency's AI infrastructure, with NATO described as a buyer in 2025. Faster foundation models, the argument goes, help Palantir rather than replace it: a model may propose an answer, but Palantir provides the controlled setting where that answer is tested, governed and turned into a decision.

The financial figures are striking for an $8 billion revenue-run-rate software company. Revenue growth of 93% year over year and net profit growth above 200% are early-stage-style results at large scale. The key risk is also contained in those numbers: the stock's valuation already assumes that an exceptional 12-quarter streak of sequential revenue acceleration continues.

ServiceNow: The Governance Layer for Agentic AI

ServiceNow's role is different. It began in IT service management and has expanded into a workflow platform covering IT, employee services, customer support, security, risk and low-code development. In an agent-driven world, that context matters. A generic model can answer an employee or summarize an incident, but it cannot by itself verify permissions, trigger the right process, keep an audit trail and complete the work. ServiceNow's AI Control Tower is pitched as the place to manage both its own and third-party agents and models.

The reported AI ACV of more than $1 billion, with a management target above $1.5 billion by the end of the year, shows that demand for this layer is growing. The durability argument is that enterprises may swap models frequently, but they are unlikely to rip out the system through which work is approved and executed.

Toast: AI Embedded in the Transaction

Toast demonstrates that the thesis extends beyond governments and large enterprises. The company integrates point-of-sale hardware, payments, online ordering, delivery, loyalty, marketing, payroll, scheduling, inventory and lending for about 180,000 locations. Because it already sits in the transaction flow, it has context at the moment of action. Toast IQ Grow, for example, can detect a slow period, identify a target audience and launch a cross-channel campaign. CEO Aman Narang says it is the company's fastest-growing new product and is on track to become the fastest to reach $10 million in annualized recurring revenue.

The operational value is concrete: restaurants do not need another chatbot; they need fuller tables, better labor use, less food waste and simpler administration. Toast's physical point-of-sale presence and payments relationship make it stickier than a standalone software tool.

None of these companies is risk-free. Palantir must keep meeting extraordinarily high expectations. ServiceNow must defend its workflow position and margins as AI usage scales. Toast operates in a demanding restaurant and local-commerce sector and must prove it can grow beyond payments and point of sale. The analysis also draws a line between corporate strength and stock attractiveness: quality and valuation are separate questions, and a good business is not automatically a good investment at any price.

What Palantir, ServiceNow and Toast Have to Prove Next

• For Palantir investors: Test the valuation against the reported 93% year-over-year revenue growth, 157% net dollar retention and the 12-quarter streak of sequential revenue acceleration; the risk is that any deceleration hits the multiple hard.
• For ServiceNow investors: The gap between current RPO of $13.2 billion and total RPO of $29 billion is a concrete measure of future revenue conversion; watch whether the AI ACV target above $1.5 billion by year-end materializes.
• For Toast investors: The 25% ARR growth and 9,500 net new locations are already reported; the next test is whether Toast IQ Grow reaches $10 million in ARR as projected and expands beyond point-of-sale and payments.
• For operators building AI products: These three companies suggest that owning the governed last mile between model output and real-world action—decisions, workflows or transactions—creates more durable switching costs than owning the model itself.