Key takeaways

A large part of finance and operations work happens around ERPs and reporting tools, not inside them.
In many organisations that work still runs on spreadsheets and email, because a dedicated system rarely paid off.
AI makes it realistic to give these processes their own system, run by agents and approved by the team.
Every step is traceable, and people keep the decisions.

The work between the systems

A large part of finance and operations work happens around the software, not inside it. Think of collecting monthly figures from subsidiaries, preparing documents for the auditor, checking a new supplier's bank details before the first payment, or tracking contract renewals and indexations.

In many organisations this work runs on spreadsheets, email and the knowledge of a few people.

How to recognize this work

The processes differ from team to team, but they tend to share a few traits:

They recur weekly, monthly or quarterly.
They draw on several sources: documents, emails, portals and spreadsheets.
They follow rules that live in people's heads or in a rule book.
Someone checks or signs off on the result.
Standard software covers them only partly, or not at all.

Why this work often has no system

Many of these processes are too specific for standard software. Volumes are modest, rules differ per organisation, and inputs often arrive as documents and emails from banks, insurers, auditors and authorities.

So the spreadsheet often became the system.

Common problems with these processes

Continuity: Much of the knowledge sits with a few people rather than in a system. When someone is sick, on leave or overloaded, the process often stalls.
Growth: Every new entity, contract or supplier adds manual work. As the business grows, these processes need more people's time and turn into a bottleneck.
Errors: Copying data between files, emails and portals leaves room for mistakes. They are often found late, or by someone outside the team.
Control and audit: It is hard to show who changed what, when and based on which source. That makes reviews and audits slow and dependent on memory.
Visibility: The status of the work often lives in inboxes and spreadsheets. Management sees the result, but rarely where things are stuck.
Onboarding: New team members learn the process by watching a colleague. Training takes time, and the rules are easy to get wrong.

What changes with AI

Agents can read documents, apply rules and carry out the steps of a process. That makes it far more realistic to give a single process its own system.

For us, that is where AI adds the most value in finance and operations: putting structure and control on work that ran without them.

What we do

We take on one process at a time. We map it with the team that runs it: the sources, the rules, the exceptions and who signs off.

We then give the process its own system: a structured record of the entities, contracts, documents or payments it concerns. Agents keep that record current and do the work, and the team reviews and approves.

Once the first process runs, we take on the next one.

What we believe

Control comes first. Every value links to its source, and every change shows who or what made it.
People keep the decisions. Agents prepare and execute, the team signs off.
Prove it on one process. One process live and measured is worth more than a broad plan.
The system outlasts the model. AI models keep improving and changing. The record, the rules and the audit trail stay.
We stay close to the work. We adjust the system when rules, sources or the team change.

Who we work with

We focus on finance and operations teams in organisations with many entities, assets or contracts, such as investment and real estate firms, buy-and-build groups and financial institutions.

Where a mature software package fits a process, that package is the right choice. We focus on the work that falls outside it.

The finance work software never reached