While tech visionaries and AI lab executives debate the existential risks and rapid pace of artificial intelligence development, everyday business leaders are facing a very different reality. At Salesforce’s annual Dreamforce conference in San Francisco, the disconnect between high-level safety debates and practical enterprise adoption was on full display. While industry pioneers like Nvidia’s Jensen Huang urged developers to push forward at maximum speed, and leaders from Anthropic and OpenAI discussed safety guardrails, attendees on the ground expressed a simpler sentiment: they are still struggling to implement the technology that already exists.

For many companies, the cutting-edge “frontier” models are far ahead of what is actually required for day-to-day operations. Representatives from various sectors noted that older, more cost-effective AI models are highly performant and perfectly capable of handling standard sales, customer service, and administrative tasks. Rather than chasing the absolute newest releases, businesses are focusing on budgeting, evaluating open-source alternatives, and integrating practical tools like Salesforce’s Agentforce, which often relies on previous-generation models to drive automated customer interactions.

The shift toward practical AI implementation is also forcing companies to rethink their financial models. The transition from traditional Software-as-a-Service (SaaS) subscription models to token-based pricing—where costs are determined by actual AI usage—presents a significant challenge to profit margins. To manage these expenses, companies like Docusign and Nice are utilizing “model routing” techniques, directing simpler tasks to cheaper, older models and reserving expensive frontier models only for highly complex, judgment-intensive workloads.

Ultimately, the consensus among many enterprise IT leaders is one of cautious integration rather than rapid adoption. System integrators and software developers note that they often wait several months before deploying the latest AI models to ensure stability and cost-efficiency. While tech giants continue their arms race to build superintelligent systems, the immediate future of enterprise AI lies not in the next breakthrough, but in successfully operationalizing the powerful tools already at hand.

Key Takeaways

• Enterprise buyers are prioritizing cost-efficiency and practical utility over cutting-edge 'frontier' AI models, finding that older models are highly capable for everyday business tasks.
• The transition from subscription-based SaaS models to token-based AI pricing is squeezing profit margins, forcing companies to adopt 'model routing' to manage costs.
• While AI safety and rapid development dominate executive keynotes, on-the-ground attendees are focused on catching up with and implementing existing technology.

Editor’s Analysis & Impact

The dynamics at Dreamforce highlight a growing maturity curve in the enterprise AI market. The initial hype cycle, characterized by a rush to adopt the most powerful models regardless of cost, is giving way to pragmatic financial and operational scrutiny. Businesses are realizing that ‘good enough’ AI is far more economically viable than paying a premium for frontier capabilities that exceed their actual needs. This shift poses a strategic challenge for top-tier AI labs like OpenAI and Anthropic, which must justify their massive R&D expenditures if the broader market prefers cheaper, older, or open-source alternatives. Furthermore, the transition to token-based pricing models threatens the historically high gross margins of traditional SaaS companies. Moving forward, the winners in the enterprise software space will not necessarily be those with the most advanced models, but those who can most efficiently orchestrate and route tasks across a hybrid ecosystem of diverse AI models.

Frequently Asked Questions

Q: Why are businesses opting for older AI models instead of the latest releases?
A: Older models are significantly cheaper, more stable, and highly capable of handling standard enterprise tasks like customer service and basic data analysis. Many companies find that frontier models offer capabilities far beyond what their current workflows require, making the extra cost unjustifiable.

Q: What is 'model routing' and how does it help companies save money?
A: Model routing is a technique where an automated system directs user requests to different AI models based on complexity. Simple tasks are routed to cheaper, older, or open-source models, while expensive, high-end frontier models are reserved only for complex reasoning and deep analysis.

Q: How is AI changing the business model for software companies?
A: Software companies are shifting from traditional subscription-based (SaaS) pricing to token-based pricing, where customers pay based on actual AI usage. This change introduces variable delivery costs, which can significantly lower gross profit margins compared to traditional software models.