Every year ahead of Christmas, enterprise resource planning (ERP) vendors gather the faithful at their annual conferences to unveil the new features they have been working on for months. In late 2025, the message was unmistakable. AI has moved from the edges of enterprise software to the centre of every product roadmap.

In unison, vendors are promising that AI will reshape the way finance teams work and, while each is taking a slightly different path, three clear themes are emerging.

The first is the rise of chat-based interfaces as the front door to the ERP. Instead of navigating menus and clicking through screens, users will increasingly ask the system to do the work for them.

Generative AI sits behind this shift to a conversational interface. These same models are also adding the ability to write month-end commentaries, summarise anomalies, draft contracts, interpret invoices and produce product descriptions.

The second is the rise of AI agent workflows inside the software. Vendors are experimenting with agents that can prepare payment proposals, chase customers, classify transactions or reconcile accounts with far less human intervention. They point to a future where accountants spend more time reviewing outputs than actively doing rote tasks.

The third is the development of agent-to-agent capabilities (known as A2A), where ERP systems expose their data so independent AI platforms can interact with them. This is a more radical shift. Rather than keeping automation locked inside the ERP, vendors are tentatively opening the door to external agents that can read, analyse and act on ERP data. It is early and uneven across vendors, but it signals a broader move towards AI layers that sit above the ERP, rather than inside it.

Against this backdrop, Acuity spoke to three vendors – Oracle NetSuite, Workday and BlackLine – who are each trying to define what AI-enabled finance and operations look like. And we spoke to one ERP services company, Rimini Street, which believes AI could spell the end of ERP as we know it.

Oracle NetSuite

At its 2025 SuiteWorld conference in Las Vegas, Oracle NetSuite outlined a sweeping AI roadmap, codenamed NetSuite Next, that included a redesigned interface, agent-driven workflows and a framework for letting external AI systems interact with ERP data.

The centrepiece is a new AI-first interface under Oracle’s Redwood design system. NetSuite is replacing its 20-year-old user interface (UI) with a layout built around conversational navigation and natural-language prompts.

The flagship feature is Ask Oracle, a chat-style assistant that lets users search, navigate, analyse and act across the suite in plain English. It returns context-aware answers, visualisations and explanations of how it reaches its conclusions, although is still in early stages and not broadly available.

“Conversations are the future of interacting with business systems,” says Evan Goldberg, executive vice president, Oracle NetSuite.

The new AI Canvas is a collaborative whiteboard inside NetSuite. Teams can pull in live data, discuss issues and then ask AI agents to take the next step. The workflows run by those agents are built elsewhere in NetSuite; the ‘canvas’ is where teams explore a problem and trigger the action.

Business users will also be able to create their own prompts to automate tasks through NetSuite Prompt Studio. The Prompt Studio is a no-code environment for building, testing and deploying custom AI prompts directly into workflows, data interactions and the native UI. One of several architectural changes is NetSuite’s adoption of the Model Context Protocol (MCP). This is a standard for giving external AI agents controlled access to ERP data, and for enabling agents to pass information between systems.

NetSuite claims to be the first ERP with a fully operational MCP server. If delivered as described it would allow tools such as Anthropic’s Claude to act on NetSuite data with appropriate permissions.

It’s a major shift that could redefine the ways we use ERPs, but it’s still early. NetSuite has not released customer case studies and MCP support will likely roll out in stages.

One of the more practical releases is NetSuite Subscription Metrics, for software as a service (SaaS) and subscription businesses. It provides out-of-the-box monthly recurring revenue (MRR)/annual recurring revenue (ARR) reporting, cohort analysis, roll-forward reporting and AI-generated explanations.

For CFOs of recurring revenue businesses, this is a tangible improvement.

Workday

Taking a broad approach to AI, Workday is combining traditional machine learning, large language models and a shift toward agent-based automation.

Workday uses a range of techniques, including established machine learning capabilities, which classify and cluster data to spot inconsistencies and miscoded transactions.

An example of using AI in finance involves Workday’s Worktags feature. These let finance teams add extra dimensions derived from spend and revenue data – such as region, project or department – directly inside the finance system of record.

Instead of tagging in a business intelligence (BI) tool or a spreadsheet, Worktags stay with the finance data itself. And rather than manually tagging transactions, the model suggests tags, users confirm or adjust them and the system learns from each correction. This means that you can run very detailed reports within the ERP itself, without having to switch to a BI tool.

The biggest shift is toward agentic AI. Workday says that business users will gradually move towards using agents to interact with the ERP. It bills this as “AI as the new UI”, where users ask for actions and the system carries them out.

“We’re at the start of a move in that direction: not just answering questions in natural language, but generating pages and charts and visualisations dynamically. It’s going to be increasingly the way that end users interact with ERPs,” says Adam Krebet, business architect at Workday.

New contract-intelligence agents can search and summarise agreements, prepare drafts using corporate templates and follow instructions expressed in plain English. Other agents can trigger tasks, schedule workflows or respond to events across the system.

To help organisations stay in control, Workday has introduced the Agent System of Record (ASOR), a central place to manage Workday-built, partner-built and customer-built agents. It helps to manage the lifecycle of agents, allowing you to govern what agents are used, what they can do and track their usage, so the work stays visible and auditable.

A series of acquisitions has built out Workday’s agentic vision. Sana adds a no-code platform for creating simple conversational agents. This makes it easier for finance and operations teams to design helpers, without relying on developers.

Flowise gives software developers the tools to create more complex, multi-step agent workflows.

Pipedream provides a library of 3000 connectors so agents can work within ERPs, procurement systems, banks and other tools.

Tackling the data problem, which sits underneath every AI initiative, Workday has a new zero-copy integration model that lets organisations analyse Workday data in BI tools, without moving or duplicating it. It means that companies no longer need to invest in expensive data warehouses and data migration projects.

BlackLine

BlackLine has moved early on agentic AI inside finance operations, introducing AI-driven assistants into the close, cash application and the long tail of collections. Its recently released Verity AI program spans the record-to-report and invoice-to-cash cycles, with the company describing it as a way to build an agentic workforce that operates alongside finance teams.

“It’s all about that currency around time and understanding the data. Verity can look at masses of data easily and quickly, and provide insight back to the user,” says Brian Morgan, vice-president, strategy and go to market – invoice-to-cash, at BlackLine.

At the centre is Vera, an AI team lead that can coordinate tasks across these agents, oversee their work and provide a clear audit trail.

BlackLine has looked for ways to use AI to improve accuracy in reconciliation. During month end, Verity scans journals and reconciliations, and flags items that fall outside expected patterns. The aim is faster exception handling and fewer hours spent digging through detail. In invoice-to-cash, BlackLine has replaced traditional optic character recognition (OCR) with large language models to read remittance data in a wider range of formats. The company says this has lifted auto-match rates to about 85%, with fewer manual keystrokes for finance teams.

The most ambitious feature is a new AI collections agent aimed at smaller debtors. The agent can call customers, send follow-ups, record and transcribe conversations, summarise outcomes, and push actions back into the ledger and contact record.

The agent is designed to handle the low-stakes, repetitive workflows, which frees up the finance team to apply their skills to more impactful, strategic work that AI can’t do.

BlackLine is also trying to address concerns about trust by certifying its AI processes under ISO/IEC 42001, the new management system standard for AI. BlackLine claims it was one of the first providers to achieve this standard.

The certification does not guarantee that Verity’s outputs will always be correct – no large language model is fully predictable – but it does confirm that the company follows documented, auditable processes for how its AI is designed, deployed and monitored.

For finance teams, this governance layer matters: it increases transparency, reduces operational risk and should, over time, improve the reliability of AI-generated insights. (As of writing, Salesforce, Oracle, Workday, SAP, Infosys and Microsoft had also certified to this standard.)

Does AI sit on top, not in ERP?

While vendors hail AI as the catalyst for ERP’s transformation, other voices believe it could spell the system’s decline. The next five years won’t be about smarter ERPs, argues Joe Locandro, executive vice president and global CIO at Rimini Street, a NASDAQ-listed support provider for SAP, Oracle and other enterprise platforms.

Locandro says most users will eventually work through AI agents to get their tasks done and those agents will move across applications, including ERPs.

“It doesn’t matter what’s at the back end. The modern UI [user interface]/UX [user experience] and the agentic AI is doing all the work and is running workflows that are custom to each organisation, regardless of what’s beneath,” he says.

The vision is similar to what ERP vendors describe, except Locandro believes it will be cheaper and more effective to use an independent AI platform, rather than one tied to a specific ERP vendor.

This is the basis for what is known as ‘headless ERP’, where the ERP remains as a database of financial and operational records but users rely on new UI/UX and agent workflow to interact with it. ERP becomes a compliant data store rather than the system where work happens.

“Your ERP ends up becoming a system of record. The functionality that’s left is pretty well transactional records and the data is being pulled from them with agents,” Locandro says.

As more processes shift to AI platforms, “the core below ends up being shrunk to transactional databases, which is why we’re saying ERP of today will probably be dead in the future”.

Consequently, companies therefore face a choice: upgrade their ERP to access vendor-native AI features, or keep their existing version and invest the upgrade budget in an independent agentic platform that can work across multiple systems.

Because ERP upgrades are costly and take time, Locandro argues businesses will get faster results by spending part of that budget on agentic tools and the rest on business growth like factory, machinery or other operational improvements.

In the meantime, finance leaders should brace for a furious marketing push from both sides. ERP vendors will argue their embedded AI keeps data, controls and workflows in one place, while the AI platforms trying to break into enterprise automation – from OpenAI and Anthropic to Palantir and others – will claim they offer greater flexibility and reach across systems.

Each camp will pitch itself as the natural home for an organisation’s future AI agents. The challenge for CFOs will be understanding where AI can deliver its promised lift in productivity and where it simply adds another layer of complexity.