ChatGPT

ChatGPT

Didn't find your answer?

Had a quick try of ChatGPT for initial draft of a reply to a clients email.

Query was - I need to invoice the EU office of a company for a photo shoot. We did the shoot in their London offices but I need to invoice their EU office. What VAT rate in Xero do I use 'Zero rated EC Services' or 'Zero rated income' or 'No VAT' and do I need to write about the reverse charge anywhere? I do have their VAT number.

ChatGPT - 

Thank you for reaching out regarding the invoicing of the shoot for the EU office of the company.

To answer your question, since the service was provided in the UK but the client is based in the EU, the appropriate VAT rate to use in Xero would be 'Zero rated EC Services.' This applies to services that are supplied to VAT-registered businesses based in the EU.

Regarding the reverse charge, you will need to include a statement on your invoice that states that the reverse charge applies, as well as the client's VAT number. This informs the client that they are responsible for accounting for the VAT on the services received in their own country.

Please let me know if you have any further questions or concerns. I will be happy to assist you.

Whilst not accurate (I believe it's a few years out of date), saves some time drafting the email as long as it's checked and edited.

Should be zero rated income on Xero in this case assuming the 'London office' is not a UK establishment.

Anyone using it at this stage?

Replies (25)

Please login or register to join the discussion.

avatar
By Mr_awol
12th May 2023 12:38

I wouldn't ordinarily have thought anyone would genuinely be using Chat GPT for this kind of purpose but yesterday i found some people who seem to think it is acceptable/a good idea, so taking this on face value as a 'should we consider this' question......

I think the danger is that it leads you down a path that sounds plausible, you then use it and get it wrong.

TBH i think the clients might be a bit miffed if they thought their accountant was just putting queries into Chat GPT and sending out replies based thereon. You also face the prospect, if you continue down a path like this, of becoming reliant on it for answers/advice and then if a client asks you face to face or on the phone, you've got nothing.

For me, this is another step down the 'lets all be a salesman and run a practice whether we are good at accountancy or not' road (sorry practice gurus, i know this is kind of what a lot or you are peddling, but i dont buy it).

Edit: I asked ChatGTP whether i should use it to advise clients of my accountancy practice. It basically said no (or, rather, to "consider its limitations and use it as a supplement rather than a replacement for human expertise")

Thanks (3)
Replying to Mr_awol:
avatar
By Elgin
12th May 2023 13:01

I should say that I had already replied to the client and was just seeing what it would come up with.

Over reliance is a major issue and all these things need an overall check..........bank feeds are a prime example.........client's still don't see why I ask them to check balances to hard copy or online bank statements and why transactions might not be captured.

Thanks (2)
Replying to Elgin:
John Toon
By John Toon
12th May 2023 13:24

Treat it like the office junior and you can't go wrong. Assume it's as capable as someone with reasonable experience and that's when you'll come horribly unstuck. But ChatGPT, Bard etc have their place now in the cutting edge practice whether that's drafting emails like this, assisting with Excel formulae or helping you code sulutions to utilise all these open APIs we get told about.

Let's not forget that both these products will be available and embedded into M365 and GSuite in the not too distant future.

Thanks (1)
Replying to johnt27:
avatar
By Hugo Fair
12th May 2023 15:34

"Treat it like the office junior and you can't go wrong" ... if you replace 'junior' with 'toddler' then I'd be more inclined to agree with you.

We all know (don't we?) that it's not displaying 'intelligence', just remarkably quick predictions of the most suitable next word (and the next one and the ...)

Last week I asked a US professor (who does research in this field) why these programs 'lie' and - after pointing out that they don't (what they actually do is to use their predictive powers unfettered by the human concept of 'truth') - he said the following:

* ChatGPT will operate within any constraints you explicitly impose on it (such as don't create imaginary citations), but those constraints have to cover all potential circumstances it may encounter whilst out trawling/predicting/generating.
* Basically it's a bit like a super 'intelligent' 3-year old that hasn't learned anything about interactions with other people ... and the consequences of that can be 'interesting'.
* Imagine catching your toddler on the point of shoving shards of broken glass into its mouth - and explaining that it's a bad idea / will hurt / won't be enjoyable.
* So why's the toddler now forcing bits of glass into his little brother's mouth?

That's not malevolence .. it's just compliant but unthinking/unaware!

Thanks (7)
Replying to Hugo Fair:
avatar
By Rgab1947
18th May 2023 10:23

What if the OP had added "as per legislation as of todays date" whether that would have given a better answer?

Thanks (0)
Replying to Rgab1947:
avatar
By Hugo Fair
18th May 2023 12:47

Might (but probably wouldn't) improve things.

Telling it what you want (and to some extent therefore what you want it to do - as per your example) is a start ... but you also have to explicitly tell it what NOT to do (like "and don't make up any legislation to use for references")!
And that's an almost infinite set of instructions ... so impossible to achieve.

Thanks (1)
avatar
By Justin Bryant
12th May 2023 12:41

Never mind that basic stuff. I saw this clever bloke this morning explaining how no-one understands how it works (particularly in terms of its "goals") and it could easily cause a nuclear holocaust (it doesn't need any level of consciousness (as we understand that, albeit very poorly) to do so):

Good Morning Britain
@GMB
5h
Thousands of scientists are calling for an immediate pause to AI development.
Professor Stuart Russell explains what AI is and the concerns around it, including Chat GPT.

https://twitter.com/GMB

https://en.wikipedia.org/wiki/Stuart_J._Russell

Thanks (3)
Replying to Justin Bryant:
avatar
By Mr_awol
12th May 2023 12:51

To be fair, it's not like we haven't been warned of that scenario - firstly in 1984 and then again in 1991, 2003, 2009, 2015 and 2019.

Thanks (4)
Replying to Mr_awol:
avatar
By Rgab1947
18th May 2023 10:24

1984 is here already. Just look around and notice all the camera's and the policing (sic) of the coronation.

Thanks (0)
Replying to Justin Bryant:
avatar
By rmillaree
12th May 2023 13:50

Like if the good guys stop developing stuff like this the bad guys would do the same - not. Surely much better being out in the open so we can monitor rather than being hidden away till it does a skynet one day.

Thanks (1)
Replying to rmillaree:
avatar
By Justin Bryant
12th May 2023 14:03

If you watched him on the telly this morning you'd understand that's not a good counter argument.

Thanks (2)
Replying to Justin Bryant:
Stepurhan
By stepurhan
12th May 2023 19:43

Alan Coren wrote a good article about this many years ago (though he arguably wasn't really talking about AI at the time).

Full text can be found as the big post here.
https://www.cookdandbombd.co.uk/forums/index.php?topic=28710.0

Thanks (1)
Replying to stepurhan:
avatar
By Hugo Fair
12th May 2023 20:36

Undated but probably just last century ... and brilliant.

I heard him once explain why his latest compendium was titled "Golfing for Cats" ... "My publishers told me that the biggest selling books were either about Golf or Cats, and I didn't see why the contents should be constrained by an accurate title. Oh and before you ask, the 3rd most popular topic was the Nazis - which is why the golf flags on the front cover carry the swastika emblem!"

Thanks (6)
Routemaster image
By tom123
12th May 2023 13:00

One of my favourite films as a child was "War Games" with Matthew Broderick inadvertently triggering a war..

Interesting comparison to where we are now..

I'm happy for automation to take some grunt work away - but not panicking about having no job just yet.

Thanks (2)
Replying to tom123:
avatar
By Postingcomments
12th May 2023 15:23

I think watching "Ex Machina" would be instructive too re the dangers of robots that just do their programming and don't have emotion or consider things outside their "remit"/sphere.

Basically, at the end, the robot thinks nothing of walking off and leaving her "lover" trapped inside a room to die.

Thanks (2)
Replying to tom123:
By Silver Birch Accts
16th May 2023 15:57

Thermo nuclear war, remember it well.

Thanks (1)
avatar
By SXGuy
12th May 2023 15:17

I've used chat GPT to write some articles and change my website homepage in the past. I've then run the output through a plagiarism checker just so it doesn't harm seo and ranking. And I'll be honest I did see an increase in traffic and click through.

I think providing you know the area you want GPT to handle it's fine as you can check and correct but using it for any topic you know nothing of will result in incorrect information.

Thanks (3)
VAT
By Jason Croke
12th May 2023 15:17

I've had some fun with ChatGPT, the more you play with it, the more you can see the limitations of what it can actually do. There is no artificial intelligence in the truest sense of science fiction movies.....all it seems to do is trawl the internet for answers and then present them in a chatty style narrative.....but it is still mightily impressive.

On a technical level, the supply is actually outside the scope of VAT, not zero rated, but we often use the zero rate coding in Xero because if we use OTS then the sale doesn't appear in Box 6 and so most people will code to zero rate for convenience.

Also, there is no such thing as "Zero rated EC Services", services are either the UK or rest of the world, we no longer differentiate between EC/rest of the world since Brexit (for services)....this gives a hint as to where it is sourcing its knowledge from.

There's no requirement to state the invoice is subject to reverse charge, although I still advise doing so just because it does not harm and helps the end customer get things right their end.

Thanks (2)
Replying to Jason Croke:
avatar
By Hugo Fair
12th May 2023 15:41

And that is, of course, the beauty of getting a set of answers from an acknowledged expert in the subject (who happens to be human)!

What I find particularly troubling is the thought that a majority of HMRC 'advisors' would be happy with the ChapGPT version ... so no doubt that is what will shortly be used by them to hand out what the general population still assumes is gospel advice.

Thanks (3)
Replying to Hugo Fair:
pic
By jndavs
12th May 2023 23:16

If ChatGPT had trained on AWEB data it would probably have replied with a good ol' rant about consulting an accountant, not a public forum!

A major problem is that we do not know here the data is sourced, just that there are many sources / the internet.

Thanks (5)
Replying to jndavs:
avatar
By Hugo Fair
13th May 2023 10:23

Indeed, but the data sources are not as widely spread as "the internet".

One of the few good things about it all is that much of the code is open source ... so those with the technological capabilities can force it to try the sort of things that responsible owners should have done themselves. For instance:

Q: what happens if you 'feed' it mostly sources from the far-right or even those predominantly promulgating conspiracy theories?
A: not surprisingly, it generated responses that 'share' those views/concerns.

Q: what happens if you tell it to 'override' any of it's default controls?
A: you can then get it to give you detailed instructions on how to create various chemical results (from illicit drugs to killer aerosols).

There's a broad assumption that these bots are subject to Asimov's 3 laws of robotics ... but they're not, they are beholden solely to their owners (who, unlike the bots, are fully aware of how these powers can be abused).

Thanks (2)
avatar
By Open all hours
13th May 2023 08:01

Only used it once to show to clients in our weekly email. It gave a vanilla response to a question relating to whether or not Elon Musk had mislead Tesla shareholders.
Assume it will soon be used widely be HMRC, Jim Harra has probably taken out a very expensive subscription on the basis that all heavily promoted software is intrinsically good.

Thanks (0)
Replying to Open all hours:
avatar
By Rgab1947
18th May 2023 10:28

Well, adding AI to HMRC should lift the IQ level a few points.

Thanks (0)
avatar
By jonharris999
14th May 2023 11:28

My general impression after a few tries is that it's rubbish at tax, but it's pretty good at well-structured and readable paragraphs about general matters - often a quicker and better result than if I had done it myself.

Thanks (2)
Replying to jonharris999:
By Silver Birch Accts
16th May 2023 15:59

Similar then to Solicitors.

Thanks (3)