What parts of ERP will be left after AI takes over?

AI agents and copilots have begun to transform the way employees interact with ERP systems by reducing much of the repetitive tasks needed to get the most out of the software, observers say.

In recent years, IT analysis firm Forrester and other experts have predicted that ERP systems are ripe for an AI-driven overhaul, and the revolution has begun. AI is helping some companies cut ERP-related manual labor by up to 20%, says Christopher Combs, senior AI business consultant at consulting firm Columbus.

Combs sees AI reshaping ERP systems in five areas: process automation, predictive analytics, decision support, user experience, and adaptive learning.

“We’re seeing automation take over repetitive tasks like data entry, invoice processing, and financial reconciliations,” he says. “AI-powered tools are also streamlining month-end close with transaction matching and anomaly detection, freeing finance teams to focus on more strategic work.”

Testing the waters

In recent months, ERP vendors, including SAP, Oracle, and Microsoft, have integrated gen AI into their products to automate many functions, including financial reporting, procurement summaries, and meeting notes, Combs says. Embedded AI assistants like SAP Joule and Microsoft Copilot are allowing for conversational ERP interactions.

As a result of these integrations, many organizations use gen AI to generate narrative reports, board decks, and customer communications, including collection emails and policy documentation, he says.

“Forecasting and planning are also significantly improved, with AI analyzing historical and real-time data for more accurate predictions,” he adds. “Intelligent procurement recommendations and automated responses to supply chain disruptions are becoming standard.”

Fast changes to ERP

The transformation of ERP through AI is happening faster than many experts expected, adds Garrick Keatts, global SAP practice leader at IBM Consulting.

“AI is transforming ERP from a static system of record into a dynamic system of intelligence,” he says. “Today’s platforms offer natural language interfaces and embedded AI copilots that not only automate transaction execution but also assist users in navigating complex decisions. ERP is no longer just about entering data, it’s about engaging with intelligent workflows that learn and adapt.”

IBM Consulting has deployed AI agents capable of automating routine tasks, such as reviewing supplier invoices against purchase orders, good receipt slips and contractual documents, Keatts says. The AI agents can find billing discrepancies in volumes and at detailed levels not possible through human driven review. 

In addition, IBM is working with clients who use gen AI to generate board-level financial reports straight from the general ledger, automate emails based on contract details and payment history, and summarize variances across thousands of accounts. “These aren’t pilots anymore,” Keatts says. “These are in production and the productivity gains are very exciting.”

In a recent report, IBM’s Institute for Business Value suggests that AI isn’t only transforming ERP functions, but also that organizations bringing AI into their systems through their ERP systems are improving their ROI and operating margins.

“ERP platforms enable regular, widespread use of AI in a way few other solutions can,” the report says. “Enterprises boldly using AI for transformation have a larger share of their enterprise platform users engaged in the use of AI and gen AI in their day-to-day tasks.”

Evolution, not revolution

While some experts see integration of AI with ERP systems as a major change in recent months, others see it as more evolutionary than revolutionary. The use of AI alongside ERP systems appears to be a quiet and structural change, with AI changing the workflows around ERP, rather than ERP itself, says Roman Rylko, CTO at software development and consulting firm Pynest.

“We’re seeing gen AI used to clean up the mess ERP never really solved: long audit logs, repetitive finance summaries, cross-system reconciliations that used to take half a day just to prep,” he says. “AI isn’t replacing ERP, but it’s taking pressure off the parts that used to rely on manual patchwork, and that’s where change is actually happening.”

One of Pynest’s clients built an AI-powered tool that connects to both NetSuite and Salesforce, and pulls revenue and pipeline data, runs basic checks, and drafts the monthly finance summary. “The controller still signs off but they don’t start from a blank page anymore,” Rylko says.

Another company using AI with its ERP system is OEM Source, an IT asset disposition (ITAD) and electronics recycling provider. AI is beginning to rewire team interactions with the ERP system, says Gene Genin, CEO and founding partner of the company.

AI is assisting with documentation of compliance and audit trails, and it’s starting to eliminate some of the grind work associated with ERP systems, he says.

“Such processes as internal summaries, duplicating messages, as well as some classifications are being processed without involvement of employees,” Genin adds. “It’s not swapping employees, it’s the backlog clearance.”

OEM Source is in the early phases of turning over dynamic dashboards to an AI system that can provide summarization and suggestions without employees needing to generate a manual report, he says.  “It lessens delays and makes us trust in the quality and on-time sharing of data with clients more,” he adds.

Taking up the slack

Genin, like Pynest’s Rylko, doesn’t see AI as replacing traditional ERP systems, as some experts have suggested, but the software will need to evolve.

 “Siloed systems of the past will start weighing heavily in comparison to more fluent and AI-integrated platforms,” he says. “We’re probably on the road to ERP functionality being more modular and task-specific as agent-based AI elements understand the business better and speak the language of each department.”

Rylko predicts that emerging ERP systems will be collections of domain-specific AI agents that pull from ERP as a backend, not a control tower. One agent may explain variances in the past month’s sales, another may forecast procurement risk based on vendor behavior, and a third may write a board summary.

In this scenario, ERP becomes a source of truth, but the real action happens in the AI layers on top of it, he adds.

“We don’t think ERP dies but the interface definitely does,” Rylko says. “Standalone ERP only survives if it gets open and modular fast. The ones that wall off access will get wrapped and routed around.”