How are finance teams using AI right now? [2026 update]

The future is here, it just isn’t evenly distributed. Some finance teams are at the bleeding edge, using multiple agents and Claude MCP servers to generate reports autonomously. But a great majority are still struggling to overcome legacy workflows.

There are three reasons that adoption has looked like a misshapen barbell:

A. Skill mismatches and shortages

The world still has too few accountants. But that shortage is nothing compared to trying to find a genuine AI expert with equivalent expertise in accounting or FP&A. Yes, financial data is perhaps the most obvious place to apply algorithms. But it also requires domain knowledge. It’s difficult for a developer to build AI agents to help with the period-end close if they don’t understand why the revenue reconciliation was happening in a spreadsheet. Or know what Sarbanes-Oxley requires. 

Further, “AI” is not a monolith, and many finance leaders confuse large language models (LLMs) with machine learning, deep learning, or reinforcement learning. There is sometimes a bit of magical thinking that occurs in saying, ‘Surely AI must be able to do this or that,’ and too little rigor or understanding of the tools. (If a use case involves detecting cancer, as you so often hear, it probably isn’t an LLM.)

B. Their data and systems are incompatible

Traditional ERPs were built to resist change. Most suffer from inherent data silos, where an inventory tool doesn’t speak to the core system. Any AI you deploy in your ERP can’t see both, and no amount of integration work and APIs can overcome that outdated architecture.

C. They don't have controls to govern the AI's work

Then there is the question of risk tolerance. Older ERPs like SAP, NetSuite, and Oracle’s products were not built to be auditable, and you cannot ‘undo’ mistakes. Most teams just aren’t willing to allow an AI anywhere near their general ledger.

That’s the reality, but don’t be daunted. There are ways to launch useful AI pilots even if you aren’t planning on migrating your ERP.

The clever approach: Use AI for what it excels at, and only that

The real ‘AI in finance’ innovation is happening where AI engineers work directly with controllers and CPAs. Finance people explain the domain, and engineers find clever ways to bend AI to their purpose.

Take LLMs, for example. These algorithms operate by guesswork, reducing text strings into tokens and pattern-matching to guess the next-best character. This means they can synthesize vast troves of information. But it sometimes makes them bad at math. They deal in probability, not certainty.

This is fixable with the right tech team. You can use LLMs as the chat interface, but give that LLM access to other tools, like a calculator. Or specific, deterministic workflows in your ERP.

Everest applies this logic in AI requests you make within the software. Go into NetSuite and ask the AI for a report and that report will change slightly each time you request it, which is obviously no good. Whereas Everest’s AI knows to write invisible scripts to fetch that report, so it’s accurate, and save that script, so every time you ask, the report is exactly the same.

An even better approach: Rethink workflows from scratch

Applying AI to your existing operations may incrementally improve them. But for real step change, finance teams are going to have to rethink those workflows entirely. Instead of coding transactions, why are accountants involved in coding at all? If a customer's AI will be processing your PDF invoice, must you generate that PDF at all? If reporting is just an intermediate step to insights, why can't everyone simply chat with the ERP's AI?

Here are a few ways teams are applying all the above logic to get actual work done.  

12 use cases for AI in fintech

What follows are credible, attainable, realistic AI use cases today, mostly with LLMs for they are the easiest to get started with. All depend on the platform or model of your ERP.

1. Reduce the need for ERP training

When your company can use an LLM to chat with your ERP, and it can actually answer questions, the interface becomes far less important. Training becomes less necessary and more people can use it. The new wave of ERPs feature this natively. For older ERPs, you must rigorously test what it actually has answers to and instruct it to tell you when it can’t generate an answer.

2. Auto-generate reports for functional leaders

Brady, a senior accountant at Spectora, uses simple LLM prompting to build reporting agents within Everest. The agents automatically generate reports for his functional leaders and check each other’s work. That’s why Brady says he’s gone from AI skeptic to evangelist: He can set guardrails, audit, and revert if something goes wrong.

3. Chat to build new intra-ERP functionality or apps

Within Everest, you can use the AiSpecify feature to request specs or build entire modules. It’s part of our theory that non-technical finance users should be able to extend the functionality of their ERP, adding fields or adjusting the interface, without sending tickets to IT.

Examples of applications Everest users have created on their own:

• A treasury management system

• Integrated sales forecasting

• An automated expense manager

• An HR reporting suite

• AI-powered cash flow predictor

4. Auto-generated summaries

Most finance systems involve a great deal of scrolling and reading. But you can potentially save your team time by adding “AI summary” fields to accounts, bills, or reports. This can, for example, explain to everyone who views a customer page that they are frequently delinquent by 43 days. Or, connect fulfillment data to tell a user that an order is ready.

5. Contract comparison and rewriting

Most LLMs now have optical character recognition (OCR) and can read handwriting and PDFs. Use them to compare contracts side by side to highlight differences. Unlike Microsoft Word’s document compare feature, an AI in your financial systems has the whole customer context. It can see how contracts evolved over time and whether both sides have held up their end. Some systems can suggest rewrites based on your company’s legal or procurement policies.

6. Automatic revenue recognition

Some ERPs like Everest are designed to automate the revenue recognition process and save your team from doing this work in a spreadsheet. Everest can automatically generate revenue scheduled by the data your team has entered, and enter some of that data for your team. It's smart enough to automatically generate contracts from a sales order and map those performance obligations line by line. 

There are also plenty of tools these days that automatically code and categorize invoices using optical character recognition (OCR)—they can read scanned paperwork and PDFs.

7. Fraud detection

Use AI to catch harmless duplicate orders or willful fraud, like fictitious invoices. Or detect and counteract malicious actors using generative AI to ‘social engineer’ their way into your systems. 

Uses include:

• Monitor transactions in real-time, alert for anomalies

• Ask the AI to recommend a better segregation of duties

• Set automatic incident response plans

• Create security tests and reminders for employees

8. Monitor inventory and suggest better order routing

If your ERP contains inventory data, you can ask the AI for recommendations just like you would an analyst. Where do errors commonly appear and what changes would fix them? What would a more efficient order routing process look like? 

Uses include: 

• Generate AI forecasts, compare to human ones

• Predictive restocking and replenishment

• Ask the AI to model scenarios—what if we held less inventory?

• Ask the AI to recommend more efficient processes

9. Predictive maintenance using IoT sensors

If you have support and maintenance data in your ERP, you can preempt maintenance crews to check in on refrigerators or lifts before they break. This is a good use case to apply to your own operations, as it can become an external offering. Many physical goods companies can launch similar services for customer maintenance contracts as a high-margin subscription or quasi-insurance offering. 

Is it time to launch a subscription? → 

10. Assess suppliers 

Your procurement system contains a wealth of unstructured information perfect for AI analysis. Use an LLM to weigh in on supplier recommendations for factors like ethics, sustainability, and diversity.

11. Forecast sales trends

An AI-powered ERP reporting suite can supplement your team’s existing forecasts. Each sales leader has a different relationship to prediction—some are optimistic, others like to sandbag. Have the AI take a pass at normalizing all their behaviors based on all your historical sales data. 

12. Supply-chain analysis

Make sense of your supply chain faster with an integrated AI that can run new scenarios.. What happens if you modify plans? Re-allocate resources? Consolidate vendors? 

Uses include:

• Run risk-efficiency tradeoff scenarios

• Research deeper into the supply chain

• Generate simulated demand forecasts for discussion

AI in ERP is still early, but the promise is clear

Our theory is that use case #3—chatting to build specs or apps—will be the great unlock for mid-market and enterprise companies. As companies grow, their financial software takes on more and more complexity until it becomes why they can’t innovate. AI is the first technology in history with the power to reduce complexity. If a finance manager or senior accountant can simply ask the ERP to generate an integrated sales forecast and, in minutes, have a working prototype, IT can build it faster. Or, as is the case in Everest, they can push it into a live sandbox, test, and push it live on their own.

We see a future where every user customizes the ERP to their own needs without interrupting each other, and ERPs become a force for innovation again.