AI in Finance is Like 'Moving From Typewriters To Word Processors' (ft.com) 69
The accounting and finance professions have long adapted to technology -- from calculators and spreadsheets to cloud computing. However, the emergence of generative AI presents both new challenges and opportunities for students looking to get ahead in the world of finance. From a report: Research last year by investment bank Evercore and Visionary Future, which incubates new ventures, highlights the workforce disruption being wreaked by generative AI. Analysing 160mn US jobs, the study reveals that service sectors such as legal and financial are highly susceptible to disruption by AI, although full job replacement is unlikely.
Instead, generative AI is expected to enhance productivity, the research concludes, particularly for those in high-value roles paying above $100,000 annually. But, for current students and graduates earning below this threshold, the challenge will be navigating these changes and identifying the skills that will be in demand in future. Generative AI is being swiftly integrated into finance and accounting, by automating specific tasks. Stuart Tait, chief technology officer for tax and legal at KPMG UK, describes it as a "game changer for tax," because it is capable of handling complex tasks beyond routine automation. "Gen AI for tax research and technical analysis will give an efficiency gain akin to moving from typewriters to word processors," he says. The tools can answer tax queries within minutes, with more than 95 per cent accuracy, Tait says.
Instead, generative AI is expected to enhance productivity, the research concludes, particularly for those in high-value roles paying above $100,000 annually. But, for current students and graduates earning below this threshold, the challenge will be navigating these changes and identifying the skills that will be in demand in future. Generative AI is being swiftly integrated into finance and accounting, by automating specific tasks. Stuart Tait, chief technology officer for tax and legal at KPMG UK, describes it as a "game changer for tax," because it is capable of handling complex tasks beyond routine automation. "Gen AI for tax research and technical analysis will give an efficiency gain akin to moving from typewriters to word processors," he says. The tools can answer tax queries within minutes, with more than 95 per cent accuracy, Tait says.
Maybe some bad results (Score:2)
Re:Maybe some bad results (Score:5, Funny)
Doers anyone else not see the potential for bad results here? This will only end in tears.
The current AI prognosticators have decided that stating wrong answers very firmly is just as good as stating correct answers. Therefore, who cares if the results are bad? We'll just drown out the bad results with more bad results until nobody notices!
Re:Maybe some bad results (Score:5, Insightful)
This seems to correlate to how many humans operate as well. Most managers don't seem to want to hear the truth about why something isn't viable. They want to be bullshitted, because they are bullshitting their bosses and it goes all the way to the top. Confidence is much more important then honesty or correctness to a surprisingly large amount of people.
There's a reason I'm not in management. I have a hard time peddling bullshit and just as hard of a time not point out bullshit when I see it. Apparently pointing out the truth is being "negative". SMH
Re: (Score:2)
Re: (Score:2)
are you sure government is also not going to use the same AI to "review" these taxes. turbotax.ai to the rescue.
it will only cost more.
Re: (Score:2)
Murphy's Law predicts that the government will use a different AI that makes different mistakes. That will probably just about double your apparent inaccuracy, and at least double your fine.
Re: (Score:2)
are you sure government is also not going to use the same AI to "review" these taxes. turbotax.ai to the rescue. it will only cost more.
Hey, if we can get enough AI shoved into everything we'll cease to be able to tell the true truths from the alternate truths and pretty soon we'll have to consult our local government "facts are only facts once we determine they are facts" office for a refresher on reality. It's gonna be awesome!
Re: (Score:2)
Ahhh ... Maat returns as the new dominant Goddess.
Re: (Score:2)
That seems to be the strategy. If everything is crap, they can have the best crap! And, "better crap" is something AI _can_ do.
Re: (Score:3)
Just wait until some crappy AI manages to obliterate the stock market and plunges the world into chaos.
Re: (Score:2)
Re: (Score:2)
This has happened several times in the history already, that is why they have safeguards in place that stop the trade when something seems fishy, that is why the world is not in chaos every time it happens, it is just a small news article.
For example here is a recent story about such issue: https://edition.cnn.com/2024/0... [cnn.com]
Re: (Score:2)
Re:Maybe some bad results (Score:5, Insightful)
Finance companies will fuck up and lose money. Then go crying to the government who will step over hungry children on the way to sending the bailout check.
Re: (Score:1)
Didn't the pandemic prove you can bailout little people too and only get 9% inflation compared to 100% increase in the Fed's balance sheet, which seems like a pretty decent trade-off, doesn't it?
Re: (Score:2)
It seems to have been a successful campaign.
The Corollary (Score:2)
Re: (Score:2)
If the IRS tried to use it then even without false positives it would probably have some disparate impact on privileged classes which would get them sued, so it's useless for them.
For scummy tax lawyers to quickly identify abusable loopholes for a given tax profile it might be much more useful. You manually check for false positives and if the false negatives constitute a small enough percentage not to hurt your business it's simply an efficiency gain. The people who get hurt are the ones who use it without
Fake winners' disease? (Score:2)
Understated FP? May I encourage you to pick a few examples? As of this writing, the FP was your only contribution.
But I will offer mine before scanning the other comments for concurrence or better ideas.
My premise is that financial speculation is not actually producing anything. Yeah, clever hagiographers can "produce" counterexamples, but the reality is nothing much there and these days the monetary values are clearly using imaginary numbers. Especially in the finance "industry".
Yet we collectively "worshi
Re: (Score:2)
Who are the richest people? And why are they so much better and worthy of respect and imitation than everyone else?
Because they are the most powerful - because they are better and worthy of respect and imitation than everyone else.
Duh.
Re:Fake laughers' disease? (Score:2)
I wasn't going for Funny (though I sometimes get there anyway). Were you?
If not, then NAK. Or perhaps NAK in either case?
Re: Maybe some bad results (Score:2)
Re: Maybe some bad results (Score:2)
Re: (Score:2)
Like moving to a Fiero with a Ferrari badge. (Score:3)
While the salespeople are busy convincing you it's a Ferrari, you realize two minutes after the contract in signed what you never got close enough to see during the negotiation. It's actually a Fiero with a Ferrari badge on the hood. Enjoy!
95 % accuracy?? (Score:5, Insightful)
FTA
"The tools can answer tax queries within minutes, with more than 95 per cent accuracy"
Is this a joke? For financial queries including tax you need 100% accuracy every time! Ok, you might not always get it if you have a bad advisor but advertising a tool that might completely screw up 1 time in every 20 costing maybe thousand or millions just shows a momumental lack of judgement.
Re: (Score:2)
Agreed! There's a reason it's so difficult to become a CPA... it's your name and the entire profession on the line with every action you take.
Re:95 % accuracy?? (Score:4, Insightful)
The point is not to use the results as is, but to quickly given a list of possible evasion schemes for the advisor to pick from and check if applicable so he can handle more clients per hour. An advisor which does an exhaustive search through the possible schemes without AI might do better, but would it be enough better to pay for his time?
It's like generative images ... it's not about replacing the artist, pure prompted images are only useful for shitty powerpoints, it's about how prompted inpainting can be used by artists to boost their productivity. Even when AI needs an expert to clean up it's inevitable AI'isms, it can still improve that experts productivity.
The problem is when people get lazy and think all they need to do is prompt.
Re: (Score:3)
Exactly. The naysayers here on Slashdot are nearsighted snarky assholes who can't get their mind out of their tiny set of failing use cases.
Just like for medical stuff, which requires even more care than finance, the initial conversation with an AI can be one of discovery: describe a few aspects of the situation at hand and ask for possibly applicable stuff. Some of the shit returned will be just that: shit that doesn't apply or is just plain wrong. Some of it will be meh, some of it will be exactly the stu
Re: (Score:2)
Clearly you've never met the "computer says no" asshat who can't think for themselves. It'll only get worse with AI systems on hand.
Re: (Score:2)
Don't move the goalposts. You said this, which has clearly been proven to be false:
For financial queries including tax you need 100% accuracy every time!"
Re: (Score:2)
Proven != because some random guy on slashdot says so. Give me a break.
Re: (Score:2)
Because logic, not because I said so. Can you prove that you need 100% accuracy every time ? It's a pretty broad statement, you know.
Just take the L, man.
Re: (Score:2)
Re: (Score:2)
There is no need for toilets, but they sure are fucking nice to have. There is no need for an IDE, but it sure saves a lot of fucking time.
Re: (Score:2)
I believe the implication is that the AI is doing the tedious research and the CPA leverages their experience and assesses the credibility of the AI output to make the actual decision. Whether or not this happens depends on the quality of the CPA and their knowledge that these LLMs are not magical.
Re: 95 % accuracy?? (Score:1)
Reminds me of accuracy of prokaryotic gene calling. Same story. 95% not enough
And when AI gives bad tax advice. (Score:3)
Re: (Score:2)
'Moving From Typewriters To Word Processors' (Score:5, Insightful)
Say like moving from typewriters that type exactly what you enter, to a word processor where clippy takes everything you type and 'corrects' it?
Not word processors (Score:3)
I'm Very Skeptical About GenAI in Accounting (Score:5, Insightful)
I know a lot of people have the understanding that Accounting as a profession is just taking a spreadsheet of numbers, adding, checking against budget, and calling it a day... but it's so much more.
For example, simply auditing a small public entity's capital purchases for a year requires a huge amount of understanding of nuanced accounting rules, recent changes to GAAP, the visible differences between invoices and receipts, and managing the data stage habits of a variety of employees. There's a painful amount of data cleaning and human interaction required to understand and post correct numbers because agencies (public agencies in particular) are so consistently understaffed that an audit also doubles as a "spring cleaning".
With that very minimal understanding of a CPA's work, you might think "Well GenAI can help with referencing new rules, interpreting invoices and receipts, and cleaning data!!" and it might... but the CPA's name or the Principal's name still goes on the audit-- not the Generative AI. They won't trust GenAI enough for it. *I* don't trust today's generative AI to interpret 1,000 scanned invoices from 500 different vendors. I don't trust a Generative AI to tell me what is genuinely correct tax code or GAAP. I might use it as a way to start a search, but I have spent WAY too much time reading extremely confident AI reports about topics that end up being 100% incorrect (especially in law in policy) to trust anything coming out of GenAI.
I could see GenAI confidently being used in the final section of Accounting work-- when it's time to write up the summary based on final, cleaned data. Everything before that is too complex. Like driving a car in a new environment, the computing power and other requisite hardware is so expensive to do it right (right now), that we're not going to see an explosion of AI take-over in this section for a long, long-time.
Re: (Score:1)
Smart people use a LLM as a devil's advocate, a critic/editor, and a very fast & effective pattern recognition helper.
I just beat up a whole law firm with GPT-4's help, for instance. The firm's partners pulled the "we're lawyers and you're not" card, blustered at me, and tried to snow me with legalese. It didn't work. AI analysis of their arguments was illuminating, and I was able to quickly zero in on statute
Re: (Score:2)
Fucking family law slimeballs! I've met a bunch over the last fifteen years, because that's what happens in middle age, and they are worse than the ambulance chasers. I can talk shit about them because I st
Re: (Score:1)
Let's first clarify that this is not "general" AI, but rather, AI that is specifically trained in accounting processes and methodologies.
CPAs make a *ton* of mistakes, especially when there is a large volume of data. Large unnecessary drains on corporate profits can easily hide in the mountain of numbers. At a personal level, this is why everybody is moving to subscription models, because subscriptions tend to hide in plain sight, leading to much more income for the subscription providers. A $10 per month s
Question (Score:2)
wtf is a typewriter, and wtf is a word processor? I haven't heard those terms in decades.
Re: (Score:3)
Re: Question (Score:2)
Wow, you made a 3D printed save icon!
We wouldn't need AI for taxes (Score:3)
if the politicians wouldn't stop screwing around the tax laws to give breaks to their favored businesses, constituents, or pay masters or using taxes to enforce social policies.
Re: (Score:1)
And we wouldn't need morticians if people didn't keep dying.
Death and taxes, you know.
As long as there are taxes, politicians will continue to try to draw money to their favorite projects, and away from the things they don't like.
AI in finance in the “cloud” (Score:2)
How is running financial software in the “cloud” any different than running it on local hardware? Cloud was over-sold as reducing costs and enhanced security. The same will apply to AI in finance. Over sold and little return to show for your investment.
Re: (Score:3)
As a daily user of MS Copilot, I can see how AI could dramatically increase productivity of accountants, just as Copilot can dramatically improve the performance of competent developers. It will be able to take on a lot of the drudge work that accountants (and programmers) hate.
word processors? (Score:2)
The transition from typewriters to word processors happened in the 70s. Anyone who remembers that as an adult is now retired. Those who do remember it, remember that all it meant was that instead of paying a typist to type on a typewriter, you now paid them to type on a word processor, which gave nicer looking results and could store the document until it ran out of memory.
Re: (Score:2)
It also meant that if you wanted to change something in the letter, you didn’t have to type out the whole thing again, and you could have template letters where you just change a few things.
game changer for tax (Score:2)
OCCUPATION:
Re: (Score:2)
So, not that great? (Score:2)
I've more impressed if the comparison was "moving from Calculators to Spreadsheets"...
anyway....
Use a Ouija Board instead (Score:3)
You can trust my numbers just as much as you can trust the numbers of the people who want to jam this down your throat. I would be unsurprised if they also used their AI/LLM to come up with that figure. They have corrupt monetary motivation to make this happen, and they know that when it fails there will be no accountability (pun intended).
The users who get screwed will have huge barriers to clear themselves. That's built into the system with disclaimers in tiny tiny print written by lawyers who would sell their own families for their organs to make a buck. When the class action suits happen it's only other lawyers who will profit and nothing will be really fixed. Lobbyists and campaign contributions/bribes will insure that even with settlements the corrupt firms will come out ahead.
We've seen this all before, AI/LLM just means it can be done even more places on an ever bigger scale.
"Answer tax queries with more than 95% accuracy" (Score:3)
A tax consultant who gave wrong answers 5% of the time would find himself out of a job very quickly, I think.
The problem is, if you want to raise the accuracy rate to 100% or even 99.9%, you're probably going to need a completely new technology to do it-- not just a bigger and better LLM.
We use this at work (Score:2)
AI is really good at summarizing a 3-25 page PDF into a couple sentences or a couple paragraphs, and answering specific questions from the data. What used to take an analyst 3-4 hours can be done in 20 minutes, and then spot/fact-checked by a human to validate the computer analysis. We aren't hiring any less analysts, but they've become 4x+ productive. Where you might have skipped analyzing documents from a B or C tier company due to time/relevance, now you're getting a much wider picture of the market.