2026 Predictions vs. Reality: What SaaS Finance Teams Should Fix Before 2027

The short answer: In December 2025 we predicted three things for SaaS finance in 2026: AI would become the finance backbone, usage-based and hybrid pricing would become the norm, and the metrics stack would grow beyond ARR and NDR. All three happened. What we didn't predict was how exposed the foundation underneath them would be: messy data, billing that can't keep up with pricing, and metrics nobody fully trusts.

Key takeaways

• AI: Adoption is not the problem, oversight is. Most finance teams use AI, but far fewer trust it to run without a human checking the work.
• Usage-based pricing: The shift is real, and hybrid (seats plus usage) is the winning model. Billing infrastructure is what's lagging.
• Metrics: Everyone agrees on what to track. Far fewer trust the number once it's tracked.
• The common thread: In every case the ambition was right and the foundation underneath wasn't ready. Fix the data layer and AI, pricing and metrics all improve together.
• Looking to 2027: The next step is giving AI governed, live access to clean revenue data, which is where MCP comes in.

What we predicted in December 2025, and what happened

As our CEO Niclas Lilja puts it: "I believed the winners would be whoever had the smartest model. Now I know it's whoever has the cleanest revenue data for the model to stand on."

Is AI in finance working? Adoption is real, oversight is the job

Ask a CFO if the team uses AI and most will say yes. Ask if they'd let it run with nobody checking its work, and the room goes quiet.

The numbers back that up. KPMG (2026) finds 93% of US finance teams plan to scale AI within 18 months, and nearly three-quarters already report meeting or exceeding ROI expectations. But Rillion's 2026 survey of 250 CFOs and finance leaders shows what the headlines miss: 68% of finance teams actively use AI, yet only 39% of CFOs trust it to run without human oversight, and 45% still require manual review after automated invoice processing.

Our own finance team is living this. Younium CFO Johan Castevall has set a target of cutting month-end close effort by 75%, and the team ran its first AI-powered close in August. They are moving in stages, and even in the later ones a person still makes the final call. His take: "Human oversight isn't being removed, it's being redesigned around where the judgment actually matters."

Martijn Plessers, Co-Founder of BONANA, told us in December that AI won't fix broken data until the foundation is solid. Months later he says the problem hasn't gone away, it has moved. Teams bolted AI onto the same messy CRM and ERP data, so AI now amplifies the mess faster. His summary: "That's not an AI problem, it's a data governance problem wearing an AI costume."

What to do: Treat AI as a data-readiness project, not a tool purchase. Before scaling AI-driven forecasting or reporting, check whether your billing and revenue data is clean and structured enough to feed it.

Is usage-based pricing leaking revenue? The model won, the billing lagged

Usage-based pricing won the argument in 2026, just not cleanly. Kyle Poyar's State of B2B Monetization survey (via ValueAdd VC, 2026, 230+ companies) puts usage-based pricing at 38% adoption, up from 27% in 2023, with 61% now running hybrid seat-plus-usage models. Hybrid, not pure consumption, is the actual winner.

The strain shows up downstream. Zylo (2026) found 78% of IT leaders were hit with unexpected usage-based charges last year, and 61% cut other projects just to cover them. Buyers are getting burned by their own bills, and the finance teams sending those bills are rarely better prepared.

Sanne van Deudekom, Revenue Director at Blinqx, had to switch to a new revenue management tool just to invoice usage-based products correctly. Now that more of the market has moved to hybrid pricing, she sees other companies hitting the same wall. Her most useful point is that the hardest part isn't the technology: "Transitioning to usage-based pricing requires a different way of working for an entire organization and the people within it."

What to do: Don't let pricing outrun billing. Before adding another usage tier or hybrid model, pressure-test your metering accuracy and invoice reconciliation.

Can you trust your SaaS metrics? Everyone tracks them, few trust them

Every board deck we've seen this year carries two scorecards: growth and efficiency. Fewer of them agree on what the numbers actually mean. CFO Advisors (2026) finds the same problems on repeat in Series A finance reviews: revenue recognised in the CRM doesn't match the ERP, CAC calculations inconsistently include S&M headcount, and NRR varies depending on who runs the query. G-Squared CFO (2026) adds that many early-stage SaaS companies lack the internal finance capacity to produce modern metrics reliably.

Linda Nordstrand, CFO at Inexchange, applies the same logic to AI-flagged numbers: "I trust it, but not blindly: the number is only as good as the human judgment layered around it..."

What to do: Pick two or three metrics beyond ARR and NDR that you can trust today, rather than adding five more you can't calculate reliably. Fix the data consistency problem first.

MCP and finance in 2027: giving AI something real to work with

If clean data is what turns AI from a headline into a habit, the next question is how AI actually reaches that data. This is where the Model Context Protocol (MCP) comes in.

What is MCP? Model Context Protocol (MCP) is a standard that allows AI tools to securely access and interact with data from business systems. Instead of building separate integrations for every AI tool and application, relying on exports, copy-paste or one-off custom integrations, MCP provides a common framework that enables AI models to retrieve and understand structured data more efficiently.

Why it matters heading into 2027. Agentic AI is moving from experiment to expectation, and not every project will survive. Gartner has predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value or inadequate risk controls (Gartner, June 2025 forecast).

Our view is that the projects most likely to last are the ones built on trusted data, clear permissions and human approval where it counts. That is also what our report found: the foundation decides the outcome, not the model.

What a finance-ready MCP looks like. We think of it as three requirements:

1. Live data from the system of record, not exports.
2. The user's own permissions, so AI can only do what that person could already do.
3. Approval before anything with financial impact, with an audit trail.

That is how Younium's MCP server works. It connects AI assistants such as Claude, ChatGPT and Cursor to live customer, subscription, quote, invoice, payment and usage data, limited to what the user's Younium role allows.

Ask “What’s our current average CMRR?” or “[Customer Name] has signed. Activate their subscription.”

“Read” prompts run instantly, meaning that questions about your customers, revenue and billing are answered straight from your system of record. However, “write” prompts (i.e. making, drafting or creating changes) come with guardrails:

• Draft changes run directly if you have full access.
• Actions that affect customers or your books always show a preview first and only go through once you approve them. These are posting invoices, activating subscriptions, sending quotes and registering payments.
• New quotes, products and usage records can be hard to undo, so review those requests carefully before they run.

Our Co-founder and VP of Product, Björn Schlingmann, puts it this way: "As AI agents take on real work across every process, a well-built MCP, with the right tools and clear documentation, isn't a nice-to-have, it's the foundation for getting implementations right."

A simple MCP readiness checklist for 2027 (Younium's recommendations):

• Decide which finance questions you'd want AI to answer from live data first (unpaid invoices, renewal pipeline, ACV, churn exposure).
• Confirm your billing and revenue data is consistent across CRM, billing and ERP before connecting anything.
• Start read-only, and only enable write access once your approval and audit process is clear.
• Define where a human must stay in the loop, especially for pricing, forecasting and anything that posts to the ledger.
• Measure whether AI answers match your own numbers before you rely on them.

The takeaway: fix the foundation once, and everything moves together

AI, usage-based pricing and the metrics stack all landed in 2026. What is still missing is the foundation underneath them. The ambition was right in every case, and the infrastructure wasn't ready to carry it. Teams aren't wrong in their predictions, they are racing to build the foundation under decisions they've already made.

Want the full picture? The full report, 2026 Predictions vs. Reality, goes deeper: longer commentary from Martijn Plessers, Sanne van Deudekom, Linda Nordstrand and Johan Castevall, and how our own finance team is rolling out AI in the month-end close, step by step. Download the report →

FAQ

Did SaaS finance predictions for 2026 come true?

Yes. AI became a core part of finance work, usage-based and hybrid pricing spread, and teams widened their metrics beyond ARR and NDR. But oversight, billing infrastructure and data trust all lagged behind the ambition.

Do finance teams trust AI to work without human oversight?

Mostly not. A 2026 Rillion survey of 250 CFOs and finance leaders found 68% of finance teams actively use AI, but only 39% of CFOs trust it to run without human oversight.

How common is usage-based pricing in SaaS in 2026?

Kyle Poyar's State of B2B Monetization survey puts usage-based pricing at 38% adoption (up from 27% in 2023), with 61% of companies running hybrid seat-plus-usage models.

What is the biggest problem with SaaS metrics in 2026?

Trust. Teams agree on what to track, but the same metric often differs between CRM and ERP, and NRR can change depending on who runs the query.

What is MCP, and why does it matter for finance teams?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to business systems in a structured, permissioned way. For finance, it means AI can query live billing and subscription data directly, under the user's own permissions, instead of relying on exports.

Is Younium's MCP server read-only?

Organisations choose between read-only and full access. With full access, AI can draft and update quotes, subscriptions and invoices, and anything with financial impact needs approval first. See the Younium MCP page for the current details.