AI tools have been powerhouses for data analytics, reporting, and trend analysis since well before ChatGPT bounced onto the scene. Generative AI is just making it way easier for the average person to make meaning out of massive chunks of numbers.

AI is, by nature/programming, brilliant at pattern recognition, especially in large datasets. This means it’s great at finding trends, and anomalies. So it’s going to be able to call out an increase in t-shirt (as opposed to hoodie) sales for an online shop as warmer weather came in last year, and use that to help forecast demand, and stock orders for this year. It can also pick out the oddball events that might be problems, or unusual opportunities. Is that set of 3 purchases of 20 t-shirts each time within a few days of each other some sort of scam attempt, or an opportunity to supply sports teams with matching kit?

Adding generative AI chat tools into the mix, and you can do all of this without needing to deal with SQL, or SPSS, or any other query or stats language. You can just load up the data and start asking questions. Plus, you get responses in normal, conversational English (or French, or Traditional Chinese if you prefer), and visualisations – I do love a good graph!

Plus, if you haven’t already, you can use AI to at least partially automate recurring reports. It’s what all the decent accounting, payroll, and other software packages that work with recurring cycles do already.

As always, make sure you practice safe data, especially if you’re on a free plan on a general-use LLM. Don’t plug PII (personally identifying information), sensitive, or confidential data into anything you don’t absolutely know has a proper data safety wall around it. My rule of thumb for the free accounts is that if I wouldn’t put it on Facebook, I won’t put it into ChatGPT, Claude, or Gemini.

If you’re on a paid plan and/or an enterprise version of one of these tools, then you should be able to share a bit more, but I’d still double check with whoever set these things up, to make sure you’re not accidentally sending your quarterly sales forecasts out into the big wide world.

Even with that, you might want to use dummy data to test prompts, not just for privacy, but also to make it easier to sense check results. These tools do still love to hallucinate.

So, it’s best to treat AI-generated reports and analysis as first drafts. You need to review them, sense check, and verify, especially if anything is going to be actually decided on the information it’s come out with.

It’s also best to use tailored software for areas such as finance and HR where compliance with things such as tax codes, privacy, attendance tracking, and so on are built into the foundations. They’re not things you want to randomly remember the next day, and try to apply over the top of your LLM prompt.

Some use cases and example prompts

I have a pinned chat in Claude where I go in and ask about formulae for Excel. I know there’s pretty much always a way to parboil and fry data in an Excel sheet but I frequently get the instructions tangled. Now I have a collection of the various formulae I need and the things I get wrong most often, plus a couple more that I would never had even tried without help.

My latest one was needing to take a unique ID in a sheet, and, in one step, add html code around it, and remove a # from the original string. Given there were 50-odd of these ID codes, it saved me not just time but intense tedium. In case you’re interested, the formula is below (I had to paste it as an image as WordPress kept processing it as code):

Use case: Using plain English to get information from a spreadsheet

Someone running a small online shop exports their sales data to a CSV (the LLMs tend to prefer CSVs to XLS documents, although they are improving). Rather than wrestling with VLOOKUPs or pivot tables, they can paste the info into a paid account with secure data and ask:

Here’s my sales data for the last 6 months. Which product category has the highest average order value? Are there any months where revenue dropped noticeably, and if so, which categories drove that drop?

Use case: No-SQL business questions

I was trained in SQL coding many years ago, and found it quite fun to go exploring through the maze of company data, but I have a weird sense of fun. For people who would rather NOT spend several hours playing join-the-dots with data tables, something like the following could be much quicker and more effective:

Here’s my business data for the last 12 months. Please act as an experienced data analyst who is reviewing this for the first time. <dump in business data – or dummy numbers if you’re worried> What trends stand out? Are there any months that look strange? What’s my rough customer retention rate, and is it stable or changing over time? What would you want to investigate further if this were your business?

Use case: Month-end numbers

Back to those recurring reports, how about turning month end numbers into a summary a non-finance person can understand:

Here are my key metrics for [month]: revenue £X, costs £Y, gross margin Z%, compared to previous months [figures]. Write a short month-end summary (3–4 sentences) suitable for sharing with a non-finance business partner.

Use case: Marketing campaign analysis

“Here are my email campaign metrics for the last 10 sends <paste stuff like subject line, open rate, click rate, unsubscribes> Which subject line patterns get higher open rates? Are there any sends that performed unusually well or poorly?

Another set of data, your health stats

A lot of people are plugging their health data into LLMs these days. Putting in an even stronger caveat than usual on data privacy – health data is about the most private and sensitive data there is – this can be very helpful. Especially when you think about the amount of data the various fitness apps and devices are collecting: sleep stages, step counts, active minutes, heart rates, menstrual cycle data, blood glucose readings, mood or energy logs, even blood pressure tracked at home.

The gold in AI is being able to plug all these individual datasets into a central place and find meaning in them, separately or together. And do it in language that actually makes sense.

Additional caveat of course is that these tools are not any sort of substitute for proper medical and other qualified consultation and advice. They can spot patterns, but they cannot, and should not, tell you what that means medically. This isn’t about diagnoses, it’s about being able to ask better questions.

Also, just how good is your smartwatch at pinpointing sleep cycles and suchlike? Your results are only going to be as good (or bad) as your source data. And, remember, LLMs are designed to be helpful. If you really want something confirmed, they’ll find a way to do that. Critical assessment, and an eye for bias is always important.

Some possible use cases for health tracking

Here’s four weeks of my sleep data and daily energy ratings <paste>. Is there any relationship between my sleep duration or quality scores and how I rate my energy the following day? Are there any patterns on specific days of the week?

Making sense of wearable device data Most fitness apps let you export your data as a CSV. Rather than staring at rows of numbers:

Here’s my Garmin data for the last three months <paste numbers/attach CSV>. Can you summarise my average sleep, resting heart rate, and step count by week, and flag any weeks that look notably different from my baseline?

Goal progress and checks

My goal was to average 8,000 steps a day and 7 hours of sleep this month. Here’s my actual data <paste>. How am I tracking?

You could follow this one up with: Here’s my calendar data/diary entries for the same time period. Can you find any patterns between this information and my steps and sleep? (bear in mind, correlation does not necessarily equal causation!)

I feel like I’ve spent most of this article giving caveats and warnings, but this area is honestly one of the most valuable uses of AI, and one that has been around for years. Dive in, and see what it has to say about the numbers swimming around you.