EPM Market Trends 2026: What the Finext Line Up Revealed About AI, Data and Tool Choice

What changed in the planning and consolidation software market in 2026?

The show opened with a look back at a busy year. Performance management software, the planning, consolidation and reporting layer that sits on top of an ERP system, went through a year of mergers, rebrands and acquisitions, and that makes it harder for you to compare tools on their own marketing.

Private equity firm Hg took OneStream private in a deal worth about 6.4 billion dollars, completed on April 1, 2026, only 17 months after the company went public. On September 16, insightsoftware launched Lineos, a single brand for the office of the CFO that bundles the finance products it had acquired over the years. And Vena, long known as a planning solution, acquired Power BI planning specialist Acterys in March and agentic AI company Morpheo AI in August, while also building its own consolidation application. Pigment is building a consolidation engine as well. The table expected little short-term impact from the deals for customers; the effect on roadmaps will take longer to show.

New vendors are entering on the planning side. AI-native tools such as Drivetrain, Aleph and Sapien let finance teams build models by prompting instead of configuring them by hand. For the guests, most of these names were new, and that is the point: they come up faster than any finance team can follow. Their customers today are scale-ups, technology companies and the subsidiaries of large groups that show up in their logo walls. The hosts also hinted that a well-known name in consolidation is building something new, still under embargo, for the 2027 edition.

Regulation moved as well. The EU Omnibus agreement of December 2025 postponed CSRD sustainability reporting for wave-two companies by two years, to 2028 over the 2027 financial year. The table described the mixed feeling of a program that ran for two to three years and then got two more. Two of the three organizations at the table chose to keep reporting as if the rules still applied, partly to keep the teams who did the work motivated.

How does Finext classify planning and consolidation tools?

Comparing a point solution to a modular platform on the same criteria tells you little, so Finext first groups tools by how they are built. The classification below is Finext's own, and Finext reviews it every year.

• ‍Organic platforms grew over time into one platform that serves several processes, such as consolidation, planning and reporting, without integrating separate products. For organizations with several business groups, the table saw the appeal: each group can report on its own while everything rolls up to one central model. In Finext's 2026 classification: OneStream and CCH Tagetik.‍
• Modular platforms bring specialized products together under one brand, often through acquisition. You get a specialist for planning and a specialist for consolidation, but under the hood they remain separate technologies with different data models. Someone has to maintain the integration between them, and financial control and business control have to coordinate every model change or the integration breaks. In Finext's 2026 classification: Anaplan (with the former Fluence for consolidation), Oracle EPM, SAP Analytics Cloud with SAP Group Reporting, Lucanet and Prophix.‍
• Point solutions specialize in one process, often with a look and feel close to Excel that lowers the barrier for users. One guest's organization runs a point solution for sales planning in a single business group, and the table agreed it would not fit at group level, where planning runs on a different level of detail. In Finext's 2026 classification: Abacum, Board, Jedox, Pigment, Una and Vena. Because Pigment and Vena are both building consolidation engines, they may move to another category in 2027.

How are EPM vendors approaching AI?

Almost every vendor now claims AI capabilities. To make those claims comparable, Finext presented a second classification at the 2026 Line Up, with four categories.

• ‍No AI (CPM only) means the tool produces financial insight the way it always has.
• ‍AI-centric tools include an AI component, for example querying your data in natural language, but keep it inside the product, so no external AI model touches the data.
• ‍AI-integrated tools open the platform to external large language models, usually through the Model Context Protocol (MCP), an open standard that lets AI assistants read from and act in other software.
• ‍AI-native tools were built in the era of generative AI, roughly the last four years, and treat prompting as the main way to work.

In Finext's 2026 classification, CCH Tagetik, Oracle, SAP and Board are AI-centric. OneStream is AI-integrated: it has its own built-in AI, SensibleAI, but also opens up to external models through MCP, with a focus on Microsoft Copilot. Drivetrain, Aleph, Sapien and Concourse are AI-native. In Drivetrain, for example, you cannot build a cube by hand: you ask for one and the tool builds it.

Finext does not rank these categories, because AI-centric and AI-integrated vendors each trade control against openness in their own way. That trade-off is where the table started to disagree.

Should finance prefer an open or a closed AI approach?

The case for a closed, AI-centric approach starts with consolidation, where a balance sheet that is 90 percent right is 100 percent wrong. Vendors in this category argue they can only guarantee correct output, and take responsibility for mistakes, when they control the technology end to end. If an external model queries the database and returns something odd, nobody can say whether the fault lies with the model or with the vendor. Risk-averse consolidation teams tend to value that clarity.

The case for an open, AI-integrated approach is that people will use AI on finance data either way. One guest argued that a platform needs some openness for exactly that reason: if it stays closed, people export the numbers and build their own version of the truth with other tools. An open platform that respects user rights, so a user sees no more through an AI assistant than in the tool itself, keeps AI working on governed data. Another guest looked forward to the self-service that follows, with business users answering more of their own questions before they come to finance.

Not everyone at the table was ready to open up. Another guest pointed out that an organization holding sensitive research data has to decide what an AI model may access before connecting anything, or the data ends up in places nobody intended. The hosts added the geopolitical angle: you may not mind operational figures passing through an AI assistant, but you probably do not want an HR planning model with personal details running over an MCP connection to a server in another jurisdiction. One guest pushed back on the fear itself, comparing it to the early days of cloud: unknown does not automatically mean dangerous.

Two practical checks came out of the discussion. First, check how mature a vendor's MCP connection is. Some only let an AI model read data, while others let it write data back or run a command such as "run a scenario with two percent more volume." Second, set the guardrails before you connect anything. Map the AI use cases that already exist in your organization, then decide what to allow. The table agreed that organizations that opened up early now have to go back and add these rules afterwards, which is harder than starting with them.

Why does master data decide whether AI works in finance?

If the table agreed on anything, it was this. Every AI use case in finance depends on data that is structured and trustworthy. When master data such as accounts, cost centers, entities, customers and vendors contains duplicates or entries that do not match across systems, your analysis falls apart, and AI makes it worse by giving confident answers at speed.

Reports and spreadsheets always depended on clean master data. AI makes the errors public: when a model returns a wrong answer, everyone in the meeting sees it. Meanwhile, more and more employees run AI tools on their own laptops, with or without a group policy that allows it, and weak master data makes that shadow use riskier. One guest described the payoff in meeting terms: the slightest doubt about a data point sidetracks the entire decision, so every hour invested in master data comes back several times over.

The practical advice from the table: set definitions, quality rules and ownership from the start, and consider a central chart of accounts that rolls up across ERP and EPM systems. It is the less exciting work, as one guest admitted, and the one with the best return.

The same foundation explains the gap between AI spending and AI results. According to Bain & Company's 2026 CFO research, 83 percent of CFOs plan to raise their AI budgets by more than 15 percent over the next two years, yet only 31 percent rate AI outcomes in finance as strongly positive. The hosts put that number to the table and joked that, statistically, at least one of the three should be satisfied. The answers spread. One guest was satisfied with a specific use case but not yet at the level of the whole organization. One called the potential itself the reason to be happy. The third, optimistic by nature, could not say the organization was there yet with costs rising as they are. Bain's own data points the same way: satisfaction rises to 41 percent at companies that have scaled AI, against 25 percent of those still in pilot mode.

What does AI change about the finance role?

At the start of the show, one of the hosts gave an AI coding assistant a single prompt and a written build specification, and left it running. By the second half, it had produced a working planning prototype with KPI cards, a revenue plan, a workforce plan and a profit bridge. The recording shows the build from prompt to result, and you can download the same specification to run it yourself.

What that means for vendors split the table. The hosts called it a real threat to established software, because building finance applications has become cheap. One guest saw more competition and lower software prices instead of collapse, since established vendors can use the same technology on top of the foundation they already have, as long as they stop promising features "in about two years." Another expected software built this fast on so little foundation to run into problems of its own, so how quickly a tool was built says nothing about how it holds up during a close.

On the finance role itself, the table was closer together. AI can already draft a forecast, write the story behind it and prepare variance and driver analyses, and the time that frees up goes into working with the business on decisions. Viewers agreed in a live poll: asked whether finance adds most value by producing reports or by assessing results and supporting decisions, most said both remain important, and when forced to choose, they picked assessing results.

The shift also asks more of you. You still have to know how every number came about, with or without AI, because the business will keep looking to finance as the judge of what the report says. One guest stressed that plans are an agreement between people: AI can speed up preparing the numbers, but the people who own a budget need to keep owning it. A viewer asked how a non-deterministic technology fits a deterministic finance world, and the table answered with a comparison to people: ask a colleague the same question twice and you may get two answers too, so you set boundaries and let them work inside those.

How do you choose a planning or consolidation tool that fits your organization?

The right tool depends on your processes, your data foundation, your appetite for open AI and the rest of your architecture, so no single tool wins for every organization.

The last split at the table was about architecture. The hosts argued for a composable architecture, where several specialized solutions each support a use case and you can replace one without rebuilding the rest. They warned about vendors that resist integration: if a tool becomes too rigid, customers replace it with one that fits their architecture. One guest preferred the other route: one platform that covers most needs, even if it is not best in class at everything, because a manageable software landscape with fewer global platforms is worth that trade. Composable gives flexibility at the cost of more interfaces and maintenance, and business units rarely want to wait for group to pick a tool. Both sides agreed the answer has to fit the organization.

To test that fit against your own situation, try the free Financial Planning Tool Advisor. Finext launched it at the end of the show. Answer a few questions about your requirements and it scores the long list of planning tools against the criteria Finext uses in its own selections, then returns a shortlist.

Frequently asked questions

What is the Finext Line Up?

The Finext Line Up is an annual, independent talk show in which Finext compares the planning and consolidation software market with no vendor interest. The 2026 edition took place on September 24, 2026, as a hybrid event with an onsite audience and online viewers, and focused on market consolidation and how EPM vendors approach AI. Asked at the end whether the Line Up should return in 2027, the table answered yes.

What is the difference between AI-centric and AI-integrated EPM software?

AI-centric EPM software has AI built into the product, such as natural-language queries, but does not let external AI models access the data. AI-integrated software opens the platform to external large language models, usually through the Model Context Protocol (MCP). Many vendors offer both; Finext classifies them by whether they open up to external models.

What is an AI-native EPM tool?

An AI-native EPM tool was built in the era of generative AI, roughly the last four years, and uses prompting as the main way to build models and plans. Examples in Finext's 2026 classification are Drivetrain, Aleph, Sapien and Concourse. Most serve scale-ups and technology companies rather than large enterprises.

Can AI in EPM tools write data back into the system?

It depends on the tool. AI-native tools and some AI-integrated tools can write data back or run commands such as a scenario calculation. Others only let an AI model read data. Check the maturity of the vendor's MCP connection, and whether it respects user access rights, before relying on it.