Minh Tuan Truong

Most finance packs are built to prove that finance did its homework. Fifty lines, three variance columns, every number reconciled. Then the meeting starts, and the CEO asks one question that none of it answers.

I now build every executive update around a simple rule: two numbers and one story. The first number shows where we are. The second shows where we are heading. The story explains what changed and what we should do about it.

It sounds like less work. It is more. You have to decide what matters before the meeting, instead of hoping someone finds it in the appendix. But that decision is exactly the job of finance. The detail still exists, one click away, for anyone who wants it.

The test is easy. If a leader leaves the room and can repeat your two numbers and your story to their team, finance has done its job.

Every revenue forecast I have seen fails in the same place: not in the model, but in the data that feeds it. Deals without owners, close dates that never move, stages that mean different things to different people.

For a long time I treated this as a sales operations topic. That was a mistake. If the forecast drives hiring, cash and Board expectations, then the quality of the pipeline is a finance risk.

What helped was simple governance. Every record has one owner. A small set of fields is mandatory before a deal can move forward. Naming follows one convention. Duplicates get merged, not ignored. None of this is exciting, and all of it is necessary.

Only then does automation make sense. A forecast that updates itself from bad data is just a faster way to be wrong.

Everyone wants an AI agent that explains last month's results. Few teams can tell you, in one sentence, how they define recurring revenue or which version of the budget is the real one.

That is where AI projects in finance succeed or fail. A language model is very good at writing a variance comment. It is not good at guessing which of three conflicting definitions you meant.

So my order is always the same. First, agree the definitions and write them down. Second, make the process repeatable, so the same data arrives the same way every month. Third, automate the first draft, and keep a person responsible for the final word.

Done in that order, AI does not replace the finance team. It gives them back the days they used to spend preparing numbers, so they can spend them on what the numbers mean.