This topic covers how vendors and buyers structure access to generative AI for business use — from packaged subscriptions to consumption and outcome‑based contracts, plus the marketplaces and agent platforms that mediate deployment. As AI shifts from pilots to mission‑critical workflows, organizations face a mix of per‑seat and per‑API/token meters, per‑resolution and outcome fees, enterprise licensing with SLAs, and hybrid on‑prem / cloud options. Key categories include AI automation platforms (IBM watsonx Assistant, Observe.AI, Crescendo.ai) that combine orchestration, copilots and contact‑center automation; AI agent marketplaces and no‑code agent platforms (Lindy) that let teams build, deploy and govern autonomous agents; and AI tool and model marketplaces (Anthropic Claude, Google Gemini, Mistral AI, Together AI) that provide model choice, dev APIs, and infrastructure for fine‑tuning or private hosting. Productivity and collaboration products (Microsoft 365 Copilot, Notion) are integrating subscription tiers to expose generative features inside apps, while specialist tools (Perplexity AI, PulpMiner) supply retrieval, grounding and data extraction that feed agent workflows. Trends shaping procurement and architecture: growing demand for private or governed model access (self‑hosted or enterprise‑grade cloud), usage‑based billing to reflect bursty inference costs, outcome‑oriented pricing in managed services, and marketplaces that create new paths to source prebuilt agents or model variants. These choices interact with security, compliance, and vendor‑lock‑in tradeoffs — pushing buyers to balance cost predictability, integration effort, and data governance. For 2026 enterprise buyers, the practical question is not only which model performs best, but which subscription and access model aligns with usage patterns, risk controls and long‑term maintainability.