From Chat to ERP AI Execution
ERP AI execution is the shift from AI that answers questions to AI-native enterprise software that autonomously runs business processes, connecting data, workflows, and domain rules so enterprise AI agents can complete real tasks across finance, procurement, and operations without human clicks on a user interface. After years of hype around conversational interfaces, enterprise vendors now compete on which system can act, not talk. Agents retrieve data, reconcile records, launch workflows, and close loops between operational and financial systems. This move is pushing vendors to rethink their architectures around APIs, orchestration, and decision engines instead of screens and forms. The winners will be platforms that hide complexity behind ‘dark software’ layers while still giving leaders clear oversight and control over how automated decisions are made and audited.
Coupa’s Autonomous Spend Management Bet
Coupa’s Inspire 2026 event highlighted how procurement and finance are becoming test beds for autonomous spend management. The company set out a vision of a spend “network effect,” where buyers and sellers transact through embedded AI rather than manual workflows, and backed that up with an aggressive acquisition streak. Recent purchases of Cirtuo, Scoutbee, Rossum, and Tonkean show Coupa buying the ingredients of ERP AI execution: category strategy, supplier discovery, intelligent document processing, and workflow automation. By bringing Rossum’s transactional large language model into its source-to-pay portfolio, Coupa is wiring invoice capture and document understanding directly into execution flows. The message to customers is that spend management will move from monitored automation to self-steering networks that identify opportunities, onboard suppliers, process invoices, and flag exceptions with minimal human input.
Owning the Execution Layer in ERP
Across the ERP market, the most active deals are clustering around the execution layer rather than chatbots. Asana’s acquisition of StackAI, Coupa’s move for Rossum, Salesforce’s agreement to buy Contentful, and Vertice’s purchase of Vendr all target the same goal: give enterprise AI agents the domain-specific data, workflow context, and content they need to act across systems. According to ERP Today, the point of AI in ERP is “not to explain what happened; it is to do the work.” Hyperscalers are reinforcing this direction, with Google Cloud integrating Gemini Enterprise into core enterprise platforms and Microsoft wiring Dynamics 365 Field Service to Project Operations and Financials to close the gap between service delivery and financial outcomes. The race is now about which vendor becomes the default execution fabric inside day-to-day operations.
Intuit’s AI-Native Enterprise Suite and the Mid-Market Gap
Intuit is using execution-first AI to attack a large gap between small-business tools and heavyweight ERPs. Intuit Enterprise Suite is an AI-native ERP aimed at companies with USD 10 million–100 million (approx. RM46 million–RM460 million) in revenue that need depth without the cost and complexity of traditional systems. The suite runs the full CFO stack as a single source of truth and uses AI agents built on Intuit’s GenOS platform to act on financial issues, not only surface them. One quotable claim from Intuit is that the suite “runs at roughly USD 12,000 (approx. RM55,200) per year versus USD 80,000+ (approx. RM368,000+) for legacy ERPs,” with more than 90% of customers live within 30 days. That pricing and deployment model makes ERP AI execution accessible to mid-market firms that previously avoided enterprise platforms.
Dark Software and the Era of Enterprise AI Agents
As ERP AI execution matures, user interfaces are starting to disappear behind agents and APIs. Younglimwon Soft Lab executive Ho Woong-ki describes the next phase as “dark software,” drawing an analogy to fully automated “dark factories” that run without human workers. In this model, enterprise systems are built as services chosen by AI agents rather than humans clicking menus. Agents pull data, trigger workflows, and complete tasks across ERP, procurement, and finance, with the software stack exposed through APIs instead of screens. Ho argues that “we are entering an era where software providers can survive only if they are selected by AI agents.” For vendors, that means redesigning products around reliability, clear machine-readable contracts, and strong audit trails, so autonomous systems can trust and continually use them in the background.
