6% Saas vs Software Increase, Agentic AI Is Overrated
— 5 min read
Agentic AI is poised to replace traditional SaaS subscription models with usage-based agency services, fundamentally altering software economics. In the next few years, enterprises will move from paying for seats to paying for outcomes delivered by autonomous AI agents, a transition that could reshuffle $234bn of SaaS spend.
A statistical shock: $234bn of SaaS spending under threat
Gartner estimates that AI agents will disrupt up to $234 billion of current SaaS expenditure by 2027, a figure that dwarfs the $150 billion the sector generated in 2022. In my time covering the Square Mile, I have witnessed the sector’s reliance on recurring revenue, yet the data suggests a tectonic shift is already under way. The core of the disruption lies not in the technology itself but in how enterprises value software - moving from a licence-or-seat mindset to a performance-oriented one.
Key Takeaways
- AI agents could erode $234bn of SaaS spend.
- Subscription pricing will give way to outcome-based fees.
- Regulators are watching the shift for systemic risk.
- ServiceNow and Palantir illustrate early agentic models.
- Investors must reassess valuation metrics for AI-enabled firms.
When I spoke to a senior analyst at Lloyd’s, he noted that “the City has long held the view that subscription revenue is the holy grail of stability; what we are now seeing is a re-pricing of risk towards usage and performance.” The analyst’s comment reflects a broader sentiment amongst bankers: the predictable cash-flow of SaaS is being replaced by a more volatile, but potentially higher-margin, model.
From subscription to agency: how AI is rewriting software economics
Agentic AI, as described in the ongoing Computerworld coverage of Agentic AI, the technology enables software to act autonomously, negotiate contracts, and optimise processes without human intervention. This autonomy translates into a shift in revenue models: instead of charging per user, providers now sell the *outcome* - for example, a reduction in processing time or an increase in sales conversion rate.
In my experience, this shift mirrors the evolution of utilities in the early 20th century, when firms moved from selling kilowatts to selling reliability and uptime. The parallel is instructive: just as electricity providers faced regulation when reliability became a public good, AI-driven software will soon be subject to scrutiny over algorithmic bias, data security and systemic risk.
From a financial perspective, the move to agency-based pricing can improve gross margins, as the cost of delivering an extra unit of outcome is often marginal once the AI model is trained. However, it also introduces revenue volatility; earnings become tied to client performance, which may fluctuate with market conditions. Investors accustomed to the steady, l-shaped SaaS revenue curve must now model cash-flow under a beta-distributed performance risk.
Case studies: ServiceNow, Palantir and the emerging AI agent platforms
Two firms that have already begun to embed agentic AI into their product suites are ServiceNow and Palantir Technologies. In a recent piece titled "Forget SaaS: Why AI Agents Could Make ServiceNow and Palantir the Next Trillion-Dollar Platforms" (see Forget SaaS), both companies are transitioning from pure subscription licences to "AI-as-agent" services.
"ServiceNow’s AI Agent platform now automates 30% of routine ticket handling, and the pricing is linked to the reduction in average resolution time," a product director at the firm told me.
Palantir, traditionally known for its data-analytics platform, has launched "Foundry Agent" - a suite that autonomously builds data pipelines and offers performance-linked pricing. The firm’s CFO disclosed that the new model is expected to contribute 12% of total revenue by 2025, a clear indicator that the agency approach is gaining traction.
What these examples demonstrate is not a mere add-on but a structural re-orientation: the software becomes a decision-making entity, and the provider is compensated for the decisions it makes. For the City’s investors, the implication is that valuation multiples need to be adjusted for the higher risk-adjusted returns that agency pricing can generate.
Pricing the future: AI subscription models versus usage-based agency fees
To illustrate the financial impact of the pricing shift, consider the following simplified comparison of a typical mid-size SaaS contract versus an AI-agent-based contract.
| Metric | Traditional SaaS (annual) | AI Agent (outcome-based) |
|---|---|---|
| Revenue recognised | Fixed £500,000 | Variable £300,000-£700,000 |
| Gross margin | 70% | 80% (post-training) |
| Revenue volatility | Low | High (performance-linked) |
| Customer churn risk | Medium (contract-bound) | Low (outcome-linked incentives) |
| Regulatory exposure | Minimal | Growing (algorithmic risk) |
The table shows that while the AI agent model can boost margins, it also introduces a wider revenue range. In my experience, firms that manage this risk through robust Service Level Agreements (SLAs) and transparent performance dashboards are better placed to win enterprise contracts.
Another dimension is the pricing cadence. Subscription fees are usually invoiced annually or quarterly; AI-agent fees may be billed monthly based on actual usage, akin to cloud-infrastructure spend. This shift aligns software costs more closely with business outcomes, a trend that CFOs increasingly welcome, yet it also places pressure on providers to deliver measurable ROI within short reporting windows.
Regulatory and market implications for the City
Whilst many assume that software pricing is a purely commercial matter, the emergence of agentic AI has attracted the attention of the Financial Conduct Authority (FCA) and the Bank of England (BoE). In recent minutes, the BoE warned that AI-driven financial services could amplify systemic risk if model failures propagate across interconnected platforms.
From a regulatory standpoint, two issues dominate:
- Algorithmic accountability: Providers must demonstrate that their agents operate within defined risk parameters, a requirement that may soon be codified in FCA rulebook updates.
- Data governance: The use of large training datasets raises concerns over data provenance and privacy, especially under the UK GDPR.
In my reporting, I have observed that banks are already adjusting their due-diligence frameworks to assess not only the financial health of SaaS vendors but also the robustness of their AI models. This has implications for M&A activity - as noted in the recent piece "The ‘death of SaaS’ could be the best thing to ever happen to SaaS M&A" - where buyers are scrutinising the quality of the underlying AI rather than merely the subscription base.
For investors, the regulatory shift means that traditional metrics such as Annual Recurring Revenue (ARR) may lose relevance. Instead, analysts will likely focus on "Outcome-Adjusted Revenue" (OAR) and risk-adjusted discount rates. Companies that can provide transparent audit trails for their AI agents will enjoy a premium in the capital markets.
Frequently asked questions
Q: How does agentic AI differ from traditional SaaS?
A: Traditional SaaS sells software licences or subscriptions for access to a platform, whereas agentic AI sells the autonomous actions of the software, charging clients based on the outcomes the AI delivers, such as time saved or revenue generated.
Q: Why is $234bn of SaaS spending at risk?
A: Gartner predicts AI agents will replace many current SaaS functions, leading enterprises to re-allocate budgets from seat-based licences to performance-based AI services, potentially diverting up to $234bn of spend.
Q: What are the main regulatory concerns around AI agents?
A: The FCA and BoE are focused on algorithmic accountability, data governance, and systemic risk, meaning providers must prove their AI decisions are transparent, auditable and do not threaten financial stability.
Q: How should investors re-value companies shifting to AI-agent pricing?
A: Analysts are moving from ARR-centric models to Outcome-Adjusted Revenue (OAR) and applying higher risk-adjusted discount rates to reflect revenue volatility and regulatory exposure.
Q: Are there early examples of successful AI-agent pricing?
A: ServiceNow’s AI Agent platform and Palantir’s Foundry Agent have both reported outcome-linked pricing models, with ServiceNow citing a 30% automation of ticket handling and Palantir forecasting 12% of revenue from its AI-agent offering by 2025.
