“You can't just take an LLM and throw it at a problem” - why Salesforce is pushing for a smarter way for everyone to do AI (www.techradar.com)

At Dreamforce 2025 Salesforce pushed the "agentic enterprise" as its roadmap for making AI practical across the business: not just shipping models, but embedding governed, managed agents into customer, employee and operational workflows via a unified platform. CTO Muralidhar Krishnaprasad and CEO Marc Benioff emphasized Salesforce’s “customer zero” approach — citing Agentforce pilots launched in 2024 that have handled more than 1.8 million conversations — to show real-world scale. The message was clear: you can't "just take an LLM and throw it at a problem" — deployment requires integration, grounding and governance to be useful and safe. For the AI/ML community this reinforces a shift from model-first to systems-first thinking. The technical implications: enterprise success depends on agent orchestration, data grounding, monitoring, human-in-the-loop controls, and lifecycle governance to avoid the high failure rates highlighted by studies (Krishnaprasad referenced an MIT finding that many AI projects fail). If done right, agents can offload repetitive work (sales/support/maintenance) and unlock developer time for innovation — Salesforce argues this will accelerate product and research cycles — but it also raises operational challenges around validation, observability, and compliance that teams must solve to realize those gains.