Five jobs a finance analyst can hand to AI this week

Most talk about AI in finance is about the distant future: autonomous trading, robo-advisers, the end of the analyst. Meanwhile, the analyst in your team still spends Monday morning copying numbers out of PDFs.

The useful question is smaller and more practical. Which parts of this week's work can AI do well enough that you only need to check them? Here are five that hold up in real finance and consulting teams.

1. First drafts of commentary

Monthly reports, portfolio updates, variance explanations, board packs. The numbers already exist. What takes the time is turning them into three clear paragraphs that say what moved and why.

Give AI the table, last month's commentary as a style example, and the two or three drivers you already know about. Ask for a draft in the same structure. You will still edit it, but you start most of the way there rather than from a blank page.

The check: every number in the draft must trace back to the table. If the AI introduces a figure you didn't give it, delete it.

2. Pulling data out of documents

Annual reports, term sheets, credit agreements, supplier statements. Extracting the same 15 fields from 40 documents is exactly the kind of work AI is good at, and exactly the kind of work that burns out junior staff.

Define the fields once (name, value, unit, page reference) and ask for a table back. Asking for the page reference matters: it makes checking fast and exposes any field the model couldn't actually find.

The check: spot-check a sample against the source, and treat any blank or "not found" field as a prompt to look yourself, not as a zero.

3. Reading the long thing for you

A 90-page regulatory consultation. A competitor's results presentation. A client's procurement pack. AI can summarise these in minutes, but a generic summary is rarely what you need.

Ask specific questions instead. "What changes for a mid-sized asset manager?" "Which of these requirements apply from next year?" "Where does this contradict what the client told us in the brief?" Ask for quotes with page numbers, so you can go straight to the source.

4. Turning messy notes into a client-ready email

After a client call you have half a page of shorthand. AI can turn that into a clean summary with actions, owners and dates, in your firm's tone, in under a minute.

The trick is to build this once as a reusable skill: a saved set of instructions with your format, your sign-off and two good examples. After that, nobody has to remember the prompt. They paste the notes and get the email.

5. Checking your own work

This is the most underrated use. Paste in a model's assumptions sheet and ask what looks inconsistent. Paste in a proposal and ask where a sceptical client would push back. Paste in a spreadsheet formula and ask what happens at the edges.

AI is not always right, but it is a tireless second reader. It catches the mislabelled column and the assumption that changed between tabs, the things a tired person misses at 9pm.

What to keep away from AI

None of this means pasting client data into whatever free tool is open in the browser. Before your team starts:

• Use a business or enterprise plan with clear terms on how your data is used, not personal accounts.
• Remove or mask client names and identifiers where the task doesn't need them.
• Keep a person accountable for every output that leaves the building.

Under POPIA, your firm is still responsible for what happens to personal information, whichever tool processes it. AI doesn't change that. It just makes good habits more important.

Where to start

Pick one of the five, one person, and one week. Measure how long the task took before and after. If it saves time and the checks hold, turn it into a skill the whole team can use.

That is exactly how we run our AI training for finance and consulting teams: on your real work, with the tools set up before we leave.