Ask Your Data Anything: What Conversational Analytics Changes for a Finance Team

Finance functions have more data available to them than almost anyone else in a business, and are frequently the least able to reach it without help.

The warehouse exists. The numbers are in it. Getting at them means either a report someone built for a different purpose, or a request to the data team and a wait. So finance builds its own parallel universe in spreadsheets, which is where a surprising amount of enterprise reporting actually lives.

Here is what changes when that function can query the warehouse in plain English.

Month-end stops being a rebuild

Month-end in most organisations is a week of pulling the same numbers into the same deck and writing much the same commentary. Every period. Under time pressure, because the close date does not move.

In the finance function of one of Australia's largest media organisations, an agent now reads directly from the warehouse, assembles the pack and drafts the commentary. A person reviews and adjusts it.

The change is one of category rather than degree. Month-end became a review instead of a rebuild — and reviewing is the part where a finance professional adds value.

The 4pm Thursday question

Every finance team knows this one. The CFO asks something specific — why is the margin down in that division, and is it the same three accounts as last quarter? — and the honest answer is that finding out will take two days.

By then the question has usually been overtaken. So either it goes unanswered, or somebody answers from intuition and everyone treats it as fact.

The value of conversational analytics is not the first question. It is the third. Why is margin down leads to which divisions leads to which accounts leads to were they on the old pricing. Each answer suggests the next question, and the chain only holds together if each step takes seconds rather than days.

Dashboards cannot do this, because a dashboard answers the question its builder anticipated. The fourth question is never the anticipated one.

The board pack gets challenged before the board sees it

A useful side effect. When the numbers are easy to interrogate, the people preparing the pack interrogate them more, because checking a figure no longer means a favour from the data team.

Errors get caught internally instead of in the room. Commentary shifts from describing what happened to explaining why, because the why is now reachable.

What it does not fix

Two things, and both are worth saying plainly.

It does not settle what your numbers mean. If revenue is defined differently in three tables, a conversational layer will confidently return all three. This is the most common reason these implementations disappoint, and the fix is a definitional conversation rather than a technical one — usually shorter than feared, but it has to happen.

It does not remove the need for finance judgement. The agent produces a draft and a number. Whether that number is the right one to put in front of a board remains entirely a human question.

Where it fits with the document side

Most finance questions are half numbers and half context. Why did that cost line move? is a warehouse question about the amount and a document question about the contract that changed.

Conversational analytics answers the first half well and the second half not at all. Which is why finance functions tend to end up wanting both a warehouse layer and a search layer over their documents — the numbers and the reason behind them.

How to tell whether it is working

Not usage. Watch whether the questions get harder.

If, three months in, your finance team is asking things nobody had built a report for, it is doing its job. If they are asking questions the old dashboards already answered, you have bought an expensive dashboard.