Has there been any development or anything on this? Sounded like a useful feature when it was added but it hardly ever works. I can use it on the same invoice format (from Amazon for instance) and it will randomly recognise maybe half of the invoices, if that, even though they are identical. Also seems to be no learning or manual mapping allowed even when it recognises the wrong field, e.g. it sees 2 dates but picks the wrong one for the invoice date and you can’t manually select the other one, and tell it that this is the one it should use in the future.
When it works it’s useful but it seems a bit of a half arsed feature that doesn’t really work. More of an annoyance most of the time!
It’s certainly an area we want to develop further. The tool has been upgraded, not too long ago actually, but I agree it still has a way to go. Very much a work in progress, and likely to be something we’ll revisit after the launch of MTD ITSA, I would hope.
Good to know. It’s great when it works, would be handy if we could provide some input to the learning process, hopefully should see it train up quite well.
Also noticed another related issue, when you run the analyser (which is still very hit & miss) if it matches anything it just creates a new purchase, missing out the search for existing purchases. I can see why you’ve done it like that but if you have a long stack of invoices to tag muscle memory takes over and you can miss the ones you need to manually attach to existing purchases so I think it should search first, then create a new purchase if no match is found.
Are we likely to see anything on this in the near future? Still very much “I can see the potential”!
There are several issues that occur regularly. It ignores the supplier on a lot of invoices but it picks a random default supplier every time and I can’t work out why it has decided that this particular supplier should be the default.
There are some invoices that I enter every month and it gets the same fields wrong every time, so some manual mapping would be a great help even if the AI isn’t improved.
The major irony that triggered me to post today was the fact that after the rigmarole of retrieving a Quickfile invoice, then having to download it, then upload it to Quickfile the receipt analyser then recognises absolutely nothing!
Actually just uploaded a coupe of receipts and the receipt hub says the AI is analysing them so it is enabled somewhere? Also how do I get it to run on the receipts that were already sat in the queue?
45 minutes later and they’re still analysing, bit more than the 2-3 minutes stated, and I thought that sounded excessive!
It’s an updated version of it, so the older version is no longer available but has been improved. The data from the beta will be carried over to this version.
It runs in the background and analyses the receipts rather than relying on clicking the wand icon. It does normally take 2-3 minutes, although there is a known issue where some can timeout causing them to show as “analysing” for quite some time. This has been fixed on our end and will be rolled out shortly.
This isn’t possible at this time, but I will feed this back to our team.
Would be useful to at least mention this before activating, although I’m probably in the minority in having the old OCR turned on so probably not an issue for many. First world problems I guess, having to manually type some numbers in!
I have noticed that fast tagging no longer works though. More clicking needed!
