What ERP vendor roadmaps reveal about AI in 2026

For years now, AI in enterprise systems has been showing up as isolated capabilities. Enhancements in forecasting, automated classifications., a generative summary here and there. All has been incredibly useful in pockets, but easier to ignore operationally. 

In 2025, that pattern broke. Across major ERP platforms, AI moved from blingy add-on features to something closer to an operating layer. Systems began explaining results, proposing actions and, in controlled cases, executing steps on behalf of users. 

And its t set up 2026 nicely as the year ERP AI starts to affect how finance and operations teams actually work, not just what tools they have access to. 

What changed across ERP platforms in 2025 

Three changes stood out across the market, regardless of vendor. 

First, ERP systems moved from reporting to explanation, with dashboards no longer the end point but the starting point for understanding performance. We saw AI increasingly used to surface what is genuinely different and why it matters, through narrative explanations of variances, drivers and forecast changes embedded directly into analytics and reporting layers. Instead of asking users to interpret endless variances, systems started to highlight the small number of movements that explain most of the outcome, reducing reliance on manual commentary and offline analysis. 

Second, automation moved from workflows to agents. Traditional ERP automation has always been rule-driven, following predefined logic where if X happens, Y is triggered. In mid 2025, platforms began introducing agent-style capabilities, including conversational interfaces and role-aware assistants, that can reason across steps, evaluate context and propose end-to-end actions rather than automating tasks in isolation. Announcements from vendors such as Oracle, SAP and Workday all pointed towards systems that support decision-making and orchestration, with humans reviewing and approving outcomes rather than manually stitching processes together. 

Third, we began to see the early foundations of a shift away from strictly period-based finance. Rather than waiting for month end to reconcile, explain and adjust, AI is starting to be used to monitor transactions closer to real time, allowing some exceptions to surface earlier and parts of reconciliation to happen progressively. Recent releases focused on automated reconciliation, exception monitoring and close optimisation signal a direction of travel towards continuous accounting, even if most organisations are still some distance from fully realising it. 

What AI we can expect to see in ERPs throughout 2026 

As we move into 2026, the most durable value from AI in ERP will come from agent-assisted work that stays inside governance, approval and audit controls. Vendors are signalling this direction through releases of purpose-built agents for finance and operations, plus the tooling to connect external models safely to ERP data and actions.  

The first area is exception detection and prioritisation that runs continuously. Instead of teams trawling reports, we will have agents that can monitor transactions, master data and workflow states to surface the small number of items that actually need attention. This is already emerging in close optimisation, reconciliation support and finance agents that sit closer to daily work rather than month-end projects.  

Another is explanation at the point of review. In 2026, more platforms will generate structured narratives that show what moved, what drove the movement and which records support the conclusion. The practical impact is fewer hours spent writing commentary and fewer meetings spent debating whose spreadsheet is correct, since the explanation is tied directly to system data and workflow context.  

Then there's agent-led orchestration with human sign-off. This is where the shift becomes operational. Agents will increasingly assemble a proposed end-to-end outcome across multiple steps, then route it for approval based on policy. Think "prepare the payment run and flag anomalies for review" or "draft the journal set and highlight entries above threshold" rather than single-task automation. For example, SAP has explicitly positioned Joule Agents for specific finance functions, and Workday has announced Illuminate agents for finance with availability planned for 2026.  

Finally, we will see safer extensibility through controlled connectors. Vendors are moving towards standardised ways to let external AI interact with ERP data and actions inside existing roles and permissions, which makes agent adoption more practical in real organisations. NetSuite's AI Connector Service, built around Model Context Protocol, is an exciting and concrete example of this approach.  

The common thread across all vendors is not autonomy for its own sake. It's shifting human effort away from navigating systems and stitching steps together, towards reviewing exceptions, approving outcomes and owning the judgement calls that still matter. 

What AI is still not good at 

Even as agent-based capabilities mature in 2026, ERP AI remains weak at areas that require contextual judgement, accountability and strategic trade-offs. It does not understand business intent beyond the parameters it is given, cannot assess materiality or risk in the way finance leaders are required to, and cannot own the consequences of decisions it proposes. AI can assemble options, surface implications and recommend actions, but responsibility for approving outcomes, managing exceptions and setting boundaries still sits firmly with humans. Organisations that blur this line risk faster processing paired with weaker governance rather than better decision-making. 

So what makes the 2026 phase of AI in ERP different? 

As platforms continue to embed agent-led capabilities, organisations will need to be more explicit about where judgement sits and how outcomes are governed, because those choices, rather than the models themselves, will determine whether AI improves decision-making or merely accelerates activity.  

EBOOK: Building an AI in finance strategy 

A business leader's guide to unlocking AI's potential in finance. 

Curious about what AI can do for finance? Download our free eBook to discover practical strategies and real-world insights on finance-led AI initiatives that will transform the finance function into a true powerhouse of data-driven insight and efficiency. Download your copy.