Introduction

The debut of GPT-6 Astra sent shockwaves through the enterprise software market, triggering broad selloffs in two of the sector's most prominent names. As general-purpose AI agents demonstrate the ability to execute complex tasks across platforms, investors are questioning whether traditional SaaS models can withstand the shift toward autonomous automation.

What Happened

On September 8, 2026, shares of Salesforce, Inc. (NYSE: CRM) declined roughly 4% and ServiceNow, Inc. (NYSE: NOW) dropped about 5% following the public launch of OpenAI's Astra on September 3. Astra was explicitly designed to operate software, introducing improvements in coding, research, and computer use that position it as a direct competitor to specialist applications. The selloff reflected concerns that general AI agents could reduce the need for expensive software seats and custom workflow development.

Procurement teams are now comparing the cost of agent-driven outcomes against traditional license pricing, creating pressure on vendors that rely on per-user seating models. Both Salesforce and ServiceNow have built layers of defense around their ecosystems, but the threat of unbundling looms large.

Why This Matters

The market reaction highlights a fundamental shift in how enterprises evaluate software value. If an AI agent can complete end-to-end processes across multiple applications, the per-seat pricing model that has long defined the SaaS industry becomes vulnerable. Companies may shift toward consumption-based spending, paying for outcomes rather than access to interfaces.

For Salesforce, the strength lies in its deep customer data and established workflows. The company reported 11% revenue growth to $11.3 billion in its latest quarter, while Agentforce annualized recurring revenue surpassed $1.5 billion, up 240%. Agentforce combined with Data 360 approached $3.9 billion in ARR, and agentic work units grew 97% sequentially to 3.2 billion. Salesforce already blends per-user licensing with consumption through Flex Credits and per-conversation pricing, positioning it to monetize work even if seat counts plateau.

ServiceNow, meanwhile, benefits from governed enterprise processes and strong institutional safeguards. Second-quarter subscription revenue rose 24.5% to $3.88 billion, current remaining obligations reached $13.20 billion, and its AI offerings crossed $1 billion in annual contract value. Production context and permission controls make it easier to deploy general models safely within large organizations. However, Astra's ability to bypass traditional navigation layers could pressure ServiceNow's differentiation strategy, as customers may demand lower prices when agents complete tasks without clicking through each product screen.

Key Takeaways

  • GPT-6 Astra's launch intensified scrutiny on SaaS valuation models that depend on seat-based pricing.
  • Salesforce's defensive moat includes proprietary data, a growing agentic ecosystem, and a hybrid pricing strategy that mixes licenses with consumption credits.
  • ServiceNow's advantage rests on governed processes and strong compliance frameworks, but faces pressure if agents reduce the need for interface navigation.
  • Fund activity showed modest shifts: CRM holder count fell slightly, while NOW holder count increased, with notable positions from Harris Associates and Fisher Asset Management.
  • Short interest in Salesforce shares remained elevated, reflecting bearish sentiment that preceded Astra's debut but also highlighting that extinction-level disruption requires more than just technical capability.

Conclusion

The emergence of general AI agents like Astra does not automatically signal the end of the SaaS model, but it does force vendors to rethink how they price, package, and deliver value. Companies that integrate AI natively, expand consumption-based offerings, and protect proprietary data and workflows will likely maintain an edge. For investors, the near-term volatility underscores the need to distinguish between companies adapting to the AI shift and those exposed to rapid unbundling.