July 27, 2026
For decades, enterprise resource planning (ERP) software has been the backbone of corporate computing. Whether running on-prem or in the cloud, millions of businesses have relied on ERP components to automate their business processes and provide a competitive advantage. But the emergence of generative and agentic AI threatens to upend our relationship with ERP software. Where will ERP end up?
There is no doubt that AI is improving in capabilities and reliability, often by the week. The top frontier models from companies like OpenAI, Anthropic, and DeepSeek feature reasoning capabilities that few people could have plausibly predicted would exist just five years ago. Many of these reasoning agents can now perform tasks at a PhD level, and some of them have surpassed human PhDs in capability.
According to the latest issue of the Stanford HAI Center’s AI Index, frontier models gained 30 percentage points in a single year on Humanity’s Last Exam, a benchmark composed of questions from nearly 1,000 subject-matter experts, primarily professors, researchers, and graduate degree holders. “AI capability is not plateauing,” the AI Index authors wrote. “It is accelerating and reaching more people than ever.”
AI models are meeting, and some are exceeding, human baselines, according to the 2026 Stanford HAI Center AI Index.
If AI is demonstrating real utility in fields like math, chemistry, weather forecasting, and astrophysics, what is stopping it from handling business processes? AI is already working its way into the back office by way of coding co-pilots, such as Anthropic’s Claude Code and IBM’s Bob. Anthropic also launched a desktop app called Claude Cowork that can plan, execute, and complete multi-step workflows autonomously. So what is going to stop AI from taking over the core of business software, the ERP systems and related enterprise suites, including CRM, supply chain management, manufacturing, inventory and warehousing, logistics, and human resources?
In early 2026, the specter of companies “vibe coding” their own business software through tools like Claude Cowork prompted a major sell-off of shares in ERP software companies. The event, dubbed the “SaaSpocalypse,” resulted in the loss of hundreds of billions of dollars in market capitalization of SaaS-based ERP companies like Salesforce, Workday, and ServiceNow, as well as other business software makers like Adobe, Intuit, Atlassian, Elastic, and MongoDB.
One takeaway from the SaaSpocalypse was that point products were at-risk of being displaced by AI. Any individual software capability, like tax preparation (Intuit) or photo enhancement (Adobe) could be replicated by AI, but AI was not yet ready to replicate the precise coordination of data and business processes that occur within ERP systems.
In May, McKinsey & Company published an informative article that explored five possible ways that AI will impact the $100 billion global ERP market. Partners with the storied business consultancy put forward a range of possibilities, from a total disintegration of ERP at the hands of AI, the dreaded SaaSpocalypse, to an acceleration of ERP implementations thanks to AI agents and an end to ERP customizations.
The SaaSpocalypse view is extreme, as McKinsey partners admit, but it would seem to be within the realm of possibility. “In the radical view, ERP as it is currently known ceases to exist,” they write. “AI agents replicate ERP capabilities, creating and optimizing processes on the fly. Data is stored in microservices rather than large tables, and application logic becomes a commodity advanced by agents.”
Much of the preparatory work has been done to enable the full-scale ERP replacement theory. Some ERP vendors, such as SAP, will no longer allow customers to modify the software, effectively cementing the commoditization of application logic. The shift from monolithic code bases to collection of REST-based Web services also powers the disintegration theory. Training an AI agent to replicate how a human interacts with software screens, or making ERP systems “headless,” would seem to be a relatively minor implementation detail at that point.
McKinsey & Company envisions a five-tier architecture for AI-enabled ERP.
The McKinsey partners see a more plausible future where ERP continues to exist, but with AI characteristics. “Application logic will still enforce business rules, data structures will ensure consistency, and systems of record will provide auditability, but users won’t directly interact with them,” they write.
With an agentic operating model, workflow execution and decision-making are handled by AI agents and overseen by humans. “Instead of executing transactions, end users will work with agents to set intent, validate outcomes, and intervene in exceptions,” the McKinsey partners write.
McKinsey foresees something interesting happening on the data front, as new semantic layers emerge that encode business context beyond the schema that’s expressed in the ERP system’s database. “An AI operating system manages these ontologies, enabling agents to reason across domains and translate intent into consistent action,” the partners write.
The second possible future for ERP envisioned by McKinsey is that the software remains basically as-is, but the ERP implementations (i.e. the “business transformation”) become much less painful, thanks to AI agents that handle much of the digital grunt work.
This is a future that has the implicit backing of IBM, which foresees its Bob AI software helping IBM i shops with application modernization. (IBM also bristles if you call Bob a “coding co-pilot.” It’s much more than that, Big Blue says.)
AI works in a probabilistic manner, but there’s no reason it can’t be used with deterministic software, such as an accounts payable program or a general ledger. The potential efficiency gains to be had from using agentic AI software to not only make modifications to source code, but then to test them using established practices, are available to businesses now. Ultimately, McKinsey says this third way could result in a 90% reduction in effort.
The fourth way that McKinsey sees AI impacting ERP is the end of the need to customize ERP software (which is something that SAP, as previously mentioned, is already doing with its cloud-only business software, for better or for worse).
Today’s ERP implementations are lengthy, painful, and risky, thanks in large part to the need to customize the systems and ensure that the new software works with all the existing software surrounding the new ERP that is not being replaced. If the words “change order” and “system integrator” gives your CIO the cold sweats, then she likely has experienced this form of fun.
Wave hello to Bob, who is more than a coding agent.
McKinsey sees AI as a potential golden knight that saves an ERP implementation from these forces of digital darkness. “ERP vendors have an inherent interest in cheaper, faster, and better delivery,” the partners write. “They now have the opportunity to provide an agentic AI–supported, end-to-end deployment and migration solution to their customers and partners, regaining control over delivery quality.”
ERP vendors that want to survive the SaaSpocalpyse will need to embrace AI, McKinsey argues. The company recommends that ERP vendors take a cleansheet approach to reimagine business processes with agentic AI. “The underlying data structure and business logic may be largely reusable, but the process flow will completely change,” the partners write.
Arguably the most plausible scenario is that ERP continues to exist, but with new AI features and characteristics. There may be some “headless” functions, where AI agents handle simple workflows without human intervention, which used to go by the name robotic process automation, or RPA.
It’s doubtful that the need for software modifications is fully eliminated – indeed, the capability to build your own custom software has been a huge advantage for the IBM i server and its predecessors. Business models must evolve to meet changing business environments, and so software must change, too. Locking everyone in to commodity software functions that are frozen and never change is a recipe for the total homogenization of business, which would be no fun at all (and it would arguably, and ironically, give an advantage to those who refused plain vanilla software to boot). But reducing the pain and risk that comes with changing or modernizing software via agentic AI is something that could be a boon to businesses.
Just as the AI boom has raised the value of hardware (good luck finding GPUs or DRAM memory or HBM memory or NAND flash stock over the next five months), it also ultimately could raise the value of software. That’s the specter forecast by Goldman Sachs Research analyst Gabriela Borges.
“The market for customer service software – incorporating traditional software-as-a-service (SaaS) products and new AI agents – could expand by an additional 20 percent to 45 percent by 2030,” according to a Goldman report from 2025.
“We believe agents will drive productivity, and software companies will capture a portion of this value,” Borges said. Goldman Sachs Research said the application software market (a broader market than just ERP) could grow to $780 billion by 2030, a 13 percent compound annual growth rate from this year.
“Our technical deep dive illustrates the potential for agents to become the new user interface for knowledge workers,” Borges said. The agent portion of the software market could account for more than 60 percent of the total by 2030. So most of the profits will come from AI agents, but the entire software market will expand.
RELATED STORIES
Bob More Than Just A Code Assistant, IBM i Chief Architect Will Says
AI Is Coming for ERP. How Will IBM i Respond?