How to Solve One Accounting Problem with AI-Native ERP Software.

Only 16% of finance teams close their books in fewer than three days, and roughly 80% of finance professionals blame the delay on waiting for data from other systems or departments. Most of that waiting comes from small, recurring tasks that one person on the team has memorized how to execute. When they’re on vacation or move companies, things break.

A common issue your team might experience are expenses that arrive at close without a department tagged.

In this article, Raiz Lead Consultant Bruce DeGraw walks through how to solve that specific problem by:

Using Aura (Rillet’s AI) in a useful chat
Turning that chat into an agent
Embedding that agent into a team-wide monthly close checklist

Identify and clarify the core issue.

Gartner projects that organizations using cloud ERP with embedded AI will close 30% faster by 2028, and that 62% of cloud ERP spending will be on AI-enabled solutions by 2027, up from 14% in 2024.

Getting there means fixing the small gaps that add up. Nearly every finance team tags expenses by department. Every team also has entries each period that show up without one. Somebody must find and fix them every month. Aura speeds up that process significantly.

Solve the problem with Aura in a chat.

Nicolas Kopp, Rillet's CEO, framed the difference between AI-native and bolted-on ERP plainly when the company raised its Series C in August: "Agents need more than access to data; they need to work inside the general ledger.”

Because Aura is native to Rillet, it already knows what "department" means, which GL codes count as expense, and how to analyze the last three months in the general ledger. It finds the problem, drafts a proposed policy, and sketches the decision flowchart the team can save and edit.

Turn that chat, decision framework, and workflow into an AI Agent.

McKinsey's research on the AI-powered finance function makes the point directly, "To capture AI's potential in finance, teams will need to do more than add new tools on top of old ways of working."

A single chat with Aura solves the problem once. An agent solves it every month, is consistent,, whether the person who built it is at their desk or on vacation. The video shows the one-off workflow of making it happen, but the bigger opportunity compounds: a workflow that used to belong to one person now belongs to the team.

Attach the Agent to an monthly close checklist.

A 2025 working paper from MIT and Stanford, based on 79 SMB companies and 277 accountants, found generative AI cut monthly close time by an average of 7.5 days. One way to get that kind of compression: a portfolio of agents, each attached to a task on the close checklist.

This last step in the process decides whether any of the work scales beyond the person who built it. Once the agent lives next to other month-end checklist items the workflow survives PTO, new hires, and a surge in transaction volume

So what? What to consider doing next.

Five things to try, in order (in Rillet, or in another AI-friendly workflow):

Start by identifying the smallest, most annoying recurring problem on your close checklist. That task that one person owns and everyone else avoids.
Solve it once in chat, then develop the artifact. Ask the AI to write the policy or draw the flowchart. Save it somewhere the team can review & edit it.
Turn the artifact into an agent. Order matters here: building the agent before the artifact hard-codes those same unwritten rules into a black box.
Attach the agent to a checklist item, rather than to a person. A workflow that lives in an individual's saved chats has not really scaled.
Keep a human in the loop. According to the Journal of Accountancy, only 7% of senior finance leaders currently prioritize governance over speed on agentic AI, and fewer than 15% of AI functions run fully autonomously across most business processes (JofA, Jul 2026). Speed without a human review layer is a recipe for disaster.

This walkthrough is specific and narrow, on purpose. One recurring task, solved at three levels of increasing impact: for one person in a chat, for a team in a shared agent, and for the whole close in a checklist item that runs every month.

Do that with thirty of the annoying, recurring tasks on your close, and the efficiency compounds where it matters most: a quieter month-end, and a finance function that scales without depending on any single person to keep it running.