From Generic Models to the AI Execution Layer
The AI execution layer is the set of data, workflow and action capabilities that let AI agents operate across business systems and complete end‑to‑end tasks, instead of only summarizing or recommending information. This layer connects models to real processes, with tools for orchestration, controls and feedback so software can take grounded actions in ERP, CRM, spend and procurement platforms. Recent enterprise software acquisitions show that vendors are no longer focused on building their own large language models; they want the connective tissue that lets agents trigger workflows, negotiate with suppliers, assemble compliant content and process complex documents. Asana, Coupa, Salesforce and Vertice are all pursuing this direction, buying cross‑system automation, document intelligence, content orchestration and procurement AI platforms to move from passive insights to practical, agentic ERP systems embedded in daily work.
Asana, Coupa and Salesforce: Building Agentic Workflows, Not New Models
Asana’s acquisition of StackAI brings no‑code AI workflows that can execute across ERP, CRM, ITSM and document systems, turning its Work Graph into a command center for “human‑agent teams.” StackAI agents can pull project context from Asana, act in external tools like Salesforce or Oracle, then send results back into Asana so more complex business processes can be “agentified” end‑to‑end. Coupa’s purchase of Rossum focuses on intelligent document processing, powered by a specialized transactional LLM trained on tens of millions of documents, to automate complex invoicing and extend document intelligence across its source‑to‑pay portfolio. Salesforce’s deal for Contentful adds a composable content layer to Customer 360, giving Agentforce structured, approved content it can query and assemble to deliver personalized experiences without manual publishing. Across these enterprise software acquisitions, the pattern is clear: vendors are buying execution and workflow automation, not generic AI models.
Vertice, Vendr and the Rise of Procurement AI Platforms
In procurement, Vertice’s acquisition of Vendr shows how data‑rich execution layers are becoming strategic assets. Vertice already runs an AI procurement platform with agentic intake, workflows, expert buying talent and procurement AI agents. By integrating Vendr’s software pricing insights, Vertice says it is creating a procurement intelligence dataset representing more than $75 billion in global indirect spend, more than 2 million pricing data points and 250,000 negotiated contracts across over 32,000 vendors. According to Vertice, this data will feed its autonomous negotiation agent, Ana, and more than 60 other AI agents across benchmarking, vendor consolidation, third‑party risk, renewal management and procurement orchestration. Customers such as ARM, Brex, Duolingo, Twilio and Santander are expected to see insights surfaced directly at the point of decision, turning procurement workflows into an AI‑enhanced system of action rather than a separate analytics tool.
What Agentic ERP Systems Mean for Vendor AI Strategy
Taken together, these moves signal a shift in vendor AI strategy toward embedded agents that can trigger, track and complete work inside ERP, CRM and spend systems. Asana is buying cross‑system action, Coupa document understanding, Salesforce structured content orchestration and Vertice procurement intelligence, but each investment aims at the same outcome: AI that acts within existing controls, data models and approval paths. This is reshaping expectations for agentic ERP systems, where AI agents sit natively inside core platforms instead of operating as standalone copilots. For buyers, the question is no longer which model a vendor runs, but how well its AI execution layer connects to contracts, invoices, cases, projects and content. Enterprise software acquisitions are therefore becoming less about inheriting algorithms and more about owning the workflows, datasets and orchestration engines that turn general‑purpose models into reliable, business‑specific agents.
