Open-source models account for less than 5% of business spend on AI tokens, according to RAMP

Overview Debates Related claims Topics

Ken Griffin stated that companies are struggling to turn enterprise artificial intelligence expenditures into measurable productivity gains

Transaction data from Ramp showed that the top 1% of business customers accounted for 80% of artificial intelligence spending on OpenAI and Anthropic

Enterprises primarily adopt artificial intelligence workloads to avoid operational costs and automate productivity tasks such as customer service and fraud detection

Zip's legal superagent autonomously redlines contracts and communicates with supplier legal counsel via email using the Model Context Protocol

Zip's initial artificial intelligence deployment utilized discrete, single-shot agents designed to automate specific tasks within procurement workflows

Zip transitioned from single-shot agents to superagents designed around discrete personas capable of handling end-to-end procurement workflows

Zip recruits forward-deployed engineers to assist enterprise customers with artificial intelligence workflow transformations

Zip maintains a verified golden library of human-approved responses to prevent artificial intelligence hallucinations when responding to vendor requests for proposals

Zip's early SOC 2 review agent operated on predefined prompts to generate an output for human review or autopilot approval

Zip's sourcing superagent can autonomously identify suppliers from a vendor master or web search, generate a request for proposal, and issue the request to vendors

Nudify applications frequently operate as simplified user interfaces wrapped around existing open-source artificial intelligence image generation models

Open-weight artificial intelligence models such as GLM and Qwen can be deployed locally to develop ransomware without safety refusals

Stable Diffusion is an open-source image generation model that users can download, modify, and operate locally

Anthropic reported that the open-weight GLM-5.3 model achieved approximately 90 percent of the performance of proprietary frontier models

Security researchers run open-weight models locally to prevent proprietary exploit data from being uploaded to external hosted providers

Approximately 5 percent of large enterprises with over 1,000 employees have deployed AI agents into production for non-engineering workflows

Ninety percent of approximately 100 enterprise organizations surveyed by 10X deployed general-purpose large language models to employees

Sixty percent of enterprise organizations surveyed by 10X provide formal employee training and enablement for general-purpose large language models

Trioma integrates Chinese open-weight artificial intelligence models, including GLM, into its orchestration platform

Meta's Muse Spark Open model allows developers and researchers to download core model components directly