Bulk importing data with AI

You upload your original spreadsheet in whatever shape it arrives in, and KEY ESG reads it, works out which of your columns feed which of ours, fills in the template for you, and then suggests how to clean up any values that don't yet match what the platform needs. You see everything it has done before anything is imported.
​

It is available on bulk upload for table-based questions, and it sits alongside the existing template upload, not in place of it. Both routes are on the same screen, and you choose per import.

Beta: The import route is currently in beta.

The New import screen offers two routes:

Under Import data faster with KEY AI, click Upload a file and choose your spreadsheet.
​

If you started an import previously and didn't finish it, you'll find it under In progress on this screen, and you can pick it up where you left off.
​

Once your file is uploaded you land on Interpret your spreadsheet. The assistant works down the left-hand side of the screen and tells you what it is doing as it goes.
​

There are two pieces of work, and it reports each one:
​

First, it reads your file and says what it found: in your file's terms, not ours. For example: how many rows of data it found, across how many columns, and what those columns hold. This is the sentence to check against the spreadsheet open on your desk: are all my rows here?
​

Secondly, it maps your columns to ours and fills in the template. The mapping is shown line by line, with the reason for each one: which of your columns feeds which KEY ESG field, and what happened to the values on the way. It will also tell you where it applied a default rather than a value from your file. For example:

The filled-in rows appear on the right, under Data from your file, while the explanation is still being written.
​

What you need to check here: that no rows are missing, and that the column mapping is right. When you're happy, click Review and clean data.

This is where values get checked against the data point itself. Cells marked in red are not yet valid, they hold a value from your file that KEY ESG doesn't recognise, and they need to be mapped to a value the platform accepts before the rows can be imported.
​

Suggestions for cleaning them are then offered, one column at a time:

Nothing is ever applied automatically. If you apply a set of changes and change your mind, Undo puts the source values back.

Where the AI isn't confident enough to suggest anything, it says so rather than guessing.
​
Those values are cleaned by hand, and the issues panel on the right finds them for you. Flagged values grouped by the column they're in, with a count for each, so you can work through them column by column.

You can also:

When every row is ready, click Import [n] rows and confirm. The import runs in the background, you can leave the screen and come back.
​
A banner on the data point shows progress, and when it's finished it tells you how many rows were added, how many were skipped, and gives you a way to revert the import if you need to.
​

​
Upload complete:
​

​Note: going Back from the review step to the interpret step starts the review again. Values you have edited, suggestions you have applied on that step, and rows you have excluded are discarded. KEY ESG will warn you before this happens.

Does my file have to match the KEY ESG template?
No , that's the point of this route. Upload it in the format you already have it. If your file is already in our template, the existing File uploads route is still there and is the quicker option.
​

How accurate is this? Can I trust what it extracted?
There are two different things happening, and they have different levels of certainty.

When we read your file, we ensure every value can be traced back to an identical value in your source spreadsheet. You can click any row and see exactly where in your file it came from.

Where a value in your file doesn't match a value KEY ESG needs, AI suggests how to map it. Those suggestions are based only on the context in your own spreadsheet, plus any mappings your team has already done in the manual upload flow. They're suggestions, so they need checking. Nothing is ever applied automatically, you preview a change before deciding whether to apply it.
​

Is my data used to train AI models?
No. Your data stays yours. We never use it to train models or share it beyond your environment. All AI processing runs through AWS Bedrock with VPC-secured connections, so data never leaves our network.
​

Is it recorded anywhere that AI was involved?
Yes. It's recorded in the audit trail when a source file is extracted using AI, and again wherever an AI suggestion for cleaning a value has been applied. So you can always trace anywhere AI has touched a value.
​

Why are some invalid values offered a suggestion and others not?
Where the AI isn't confident enough, it offers no suggestion at all, we'd rather say nothing than guess. Those values are cleaned manually, and the issues panel finds them for you and lets you work through them.
​

Will the suggestions get better over time?
Yes. Suggestions draw partly on mappings your team has done before, so a long-standing account will generally get better suggestions than one in its first week.
​

Which data points can I use this on?
The bulk upload flow for table-based questions, the ones where you enter rows of data. Entry-method level bulk imports and multi-data-point workflows are coming soon.
​

Can I stop halfway through and come back?
Yes. Unfinished imports are listed under In progress on the New import screen.
​

Can I undo an import after it has run?
Yes, the banner shown when an import completes includes a way to revert it.

If you have any further questions, please reach our team at support@keyesg.com or in the In-Platform chat