Will AI replace Accountants?
Unlike other jobs listed on this site, the question of whether AI could replace accountants is more nuanced.
AI is undoubtedly going to affect accounting in significant ways, far more than the work of say a Plumber or a Comedian, but like many previous technological sea-changes in the world of accounting, accountants who adapt to the changes, and strategize how they can use new AI tools to their benefit, should be ok.
Part of this is knowing what the changes are likely to be. This article explores the subject in-depth.
Accountants have always dealt with change
Ever since our neolithic hunter-gatherer forefathers 10,000 years ago - in what is now modern-day Syria, switched to farming… and coalesced into communities, there has been a need to count, track, and record people’s assets - and how we exchange them.
In those earliest days of civilization it was animals, crops and land. These days it's the abstracted form that is money.
Accountants and numerically minded folk were there from the beginning, figuring out ways to keep records of things - whether it was symbols of animals on discs, or clay cones and spheres representing 1’s and 10’s. The history of accounting is tied to the very earliest efforts of humans to count, write and record.
In the ten millenniums since, accountants have frequently needed to adapt to new ways of doing things, often inventing new approaches and technologies themselves.
In ancient Egypt, accountants developed methods to track taxes, maintain inventories, and monitor the progress of construction projects.
In the 8th-13th centuries, Islamic accountants adopted Indian numerals and the number zero, and invented double-entry bookkeeping.
In the 17th century, British accountant Richard Dafforne introduced the concept of depreciation, coinciding with the birth of some of the world’s first large corporations.
In more recent times, advances such as Hollerith's punch card machine in 1890, were adapted by accountants to count and track business assets. Accountants were there in the basements of office buildings in the 1920s, figuring out how to wire plugboards of IBM tabulators.
Bringing things fully up to date, there are accountants working today who'll vividly recall the introduction of spreadsheets in 1979, Quickbooks in the 1980s, ERP systems of the 90s, and cloud-based accounting in the 2000s.
The point obviously is that accountants have been adapting and inventing new technologies for a long time. Ten thousand years in fact.
AI can (and should) be viewed as a similar new technology, albeit very powerful, that accountants will need to adapt to, and hew to their own needs.
Fear of the unknown
It’s entirely reasonable that accountants at this moment be concerned about artificial intelligence, and how it might affect their careers.
The name artificial intelligence alone, is enough to strike fear into the minds of intelligent people. It doesn’t help that intelligent people who are making AI, don’t fully know how it all works, or perhaps more accurately, how it might combine. Add to this the fact that futurists like myself and well known prognosticators are predicting up to 80% job losses within two decades, and this should frankly be terrifying to anyone with a pulse.
But for accountants specifically, there are a number of reasons not to be too alarmed, as this article will argue. The devil is in the detail.
What AI can and can’t do in regards to Accounting
AI hit the world like a freight train in November, with the public release of ChatGPT. For the first time, anyone, including many tech savvy individuals in Silicon Valley, were suddenly able to try out the state of the art in LLM’s - large language models.
If you haven’t tried it out yourself yet, you need to give it a spin. I’ll remind you towards the end.
ChatGPT and other text-based LLM’s (and soon voice versions of the same) are mind-blowingly impressive technology. It’s like you’re talking to an expert… on everything.
In industry parlance they’ve been ‘trained’ on huge quantities of information - large swathes of the public Internet, all of Wikipedia, countless medical and professional texts, papers, transcripts of podcasts, any publicly accessible data they're able to get their hands on.
This gives a very convincing illusion of intelligence, and there’s no doubt this can be used for many things that humans have traditionally done. Sophisticated LLM’s like ChatGPT can summarize, organize, create things on command, and speak to us in a highly convincing way. There are similar models for songs and images.
For example, it may shock you that all of the 'job' photos on this site were created with an AI image generator. None of the people in these photos are real.
AI and more accurately machine learning has been entering the accounting space around the edges for several years. QuickBooks began introducing ‘AI’ features in 2018 and we’ve already seen benefits in terms of improved forecasting, smart invoice generation, receipt image capture and data extraction, mileage categorization, transaction categorization, and so on. The cloud-based accounting platforms are at an advantage, able to leverage their insight into billions of transactions across millions of businesses. Enterprise clients have had access to similar tools for longer, trained on their own siloed data.
These features will continue to improve and the text based LLM’s that have hit the headlines recently will enable natural language questions via AI that are natural and sophisticated. Worryingly for accountants, it’s trivial for these LLM's to be trained on all of the financial data of a company, accounting regulations, every accounting book they can get their hands on, and tax laws.
And this represents the threat and opportunity for accountants. If a client can ask AI any question about accounting or tax law, doesn’t that make the service an accountant offers redundant?
In a way, yes, but it also assumes people know how to ask the right questions. And it assumes accountants don't offer services besides answering questions.
On the flip side, accountants can of course use the very same AI tools for asking questions themselves, whether for research, staying up on the law, or their day-to-day work. And accountants should generally ask better questions!
What changes will AI bring to accounting in future?
As a futurist my job is to forecast things before they happen, and the following is my best take on what I see happening with AI in regards to accounting.
The cloud accounting and bookkeeping packages like QuickBooks, Xero, FreshBooks, Wave, Sage, Zoho etc, will all compete to integrate sophisticated text-based LLM’s into their platforms. This will enable convincing ‘Ask a Virtual Accountant’ type Q&A functionality, and these virtual advisors will have access to all of a company or individual’s accounting records. This is perhaps the biggest single headwind to many accountants.
DIY accounting and self-bookkeeping will become easier for individuals and businesses. Many who don’t need certification of accounts may choose to drop their long standing accountants.
Common tasks like inputting data, reconciliation, creating and dispatching invoices will become more automated, and easier.
Lifelike virtual AI accountants (think Zoom calls) will appear in perhaps 3-5 years as convincing real-time generative AI video becomes possible.
Tax authorities will deploy AI extensively to identify companies and individuals to audit.
Tax authorities will likely lay off many workers.
Bear with me for the good news…
The reasons why many accountants are safe from AI
Judgement calls and ethics
AI, no matter how advanced, cannot replicate human intuition and judgment. Accountants who serve as trusted advisors to their clients, routinely face decisions requiring ethical consideration, years of industry experience, and a nuanced understanding of regulations, business objectives, and financial systems. They see a bigger picture than just the data, and can benefit from AI tools themselves to see even more of a client’s picture.
While AI can crunch numbers across disparate data rapidly, and AI agents may soon be able to answer highly sophisticated accounting questions, is it possible for AI to determine the practical or ethical implications of these numbers, or the subjective considerations in interpreting them?
Accountants often serve as financial advisors, strategists, and even confidantes to clients. This aspect of the job demands emotional intelligence, empathy, and an understanding of human nature. While AI can provide data-driven insights, it cannot forge the trust and relationship necessary for a full-featured personal collaboration between an accountant and their client.
The Human Touch
Accountants tend to form long-lasting relationships with their clients, often over many years or decades, and this human-to-human connection is very important to most people.
Once someone has found an accountant they like to work with, and assuming the relationship continues to work, there is very little incentive to change.
Clients particularly value ideas from accountants that can save them money and ideally provide a net positive benefit from the relationship. For many individuals and small business owners, if it feels just a little mischievous (even if it isn’t), that’s even better.
This will be a factor for many people when deciding between AI and a human accountant. It’s going to be impossible for AI to replicate the feeling they are in a secret club. Cloud-based AI accounting platforms will have to do things very... 'clinically'.
Strategic Decision Making and Accountability
As cloud-based accounting platforms implement LLM’s for Q&A and virtual accountants, they are going to face many of the same challenges ChatGPT faces now around misinformation, only with far more potential liability.
If a QuickBooks virtual accountant suggests a strategy that turns out badly, or gives objectively wrong answers to a critical Q&A data question that is relied upon, Intuit could very quickly find themselves on the receiving end of a lawsuit.
So the platforms will need to implement disclaimers and put AI ‘guardrails’ in place, to use industry parlance.
For example, here is the disclaimer I received from ChatGPT on the subject of accounting advice…
"The information and content provided by ChatGPT is for general informational purposes only and should not be considered as professional accounting or financial advice. While we strive to provide accurate and up-to-date information, we make no representations or warranties of any kind, express or implied, about the completeness, accuracy, reliability, suitability, or availability of the content. You should always consult with a qualified accounting professional or financial advisor before making any financial decisions or taking actions based on the information provided by ChatGPT. Reliance on any information from ChatGPT is strictly at your own risk.”
Accountants are fundamentally accountable - both to their clients and authorities. It may be worth spending time thinking about this aspect, and perhaps developing ideas or strategies to benefit from the dynamic.
LLM’s are trained on old data. ChatGPT currently only has information up to September 2021 for example.
It takes enormous compute power to train LLM’s on large datasets, and a surprisingly long time - around 1 year currently for ChatGPT.
The cloud-based accounting platforms will thus inevitably tend to be constrained to past and (at some point in the future…) near-present information. They might only contain information on legislation up to say 3 months ago as an example, in a best (future) case. Bear in mind that these platforms will always have to spend time testing things and putting guardrails in.
It will thus be extremely challenging for platforms to ‘see into the future’ and make predictions on soft information and intuition. Accountants have an opportunity therefore, through industry events, professional communities, and their years of experience, to ‘see the way the wind is blowing’, and essentially be accounting futurists. This is a unique benefit accountants can provide to their clients.
I’d encourage accountants to proactively contact clients about things like disappearing benefits, if they don’t already do so, and take advantage of ’before things change’ type communications.
If clients feel their accountants are seeing into the future for them, with tangible benefit, this is a value-add that AI can’t replicate.
Differences between law and practice
We all know there is a disconnect between law and applied practice.
Similar to the reasoning above, accountants who keep their ear to the ground, and are engaged in recent case law or disputes before the courts, can offer benefits that AI simply can’t match.
Individuals and businesses rely on their accountants to handle sensitive financial information and offer valuable guidance.
Accountants establish meaningful relationships with their clients, understanding their unique needs and concerns, and this fosters a strong bond of trust.
This connection not only allows accountants to provide personalized service, but also helps clients feel more confident in their financial decision-making. The role of an accountant often extends beyond mere numbers, as clients look to them for strategic financial advice, tax planning, and even emotional support during challenging times.
This level of rapport and empathy is impossible for AI systems to emulate, as they are simply not human, and have no life experience. Put simply, they lack the human touch that enables clients to feel understood and supported. It would be weird to try and get such support out of an AI accountant. AI can't do human empathy.
This aspect of client trust, deeply rooted in the human experience, reinforces the value of accountants in the face of AI advances, particularly when it comes to such a sensitive and important subject as people's money.
Many accountants are self-employed and business owners themselves. Whether through their own personal experience, or from insight into the businesses of their clients, accountants uniquely possess a broad understanding of business operations, industry trends, legal matters and much more.
This enables them to provide insightful advice to businesses, that AI can't match.
Accountants can provide anecdotes such as “I’ve got a client who…” when having a client call. This is unique, and won't be replicated by AI for a long time to come.
I’d encourage accountants to double-down on this kind of thing.
Networking and Introductions
This is something accountants can provide which AI can’t - personal introductions to other professionals, or between clients. The latter may be something that some accountants are wary of, but the more that accountants can differentiate themselves against AI, the better.
The need for Sign Off
The simple requirement that many businesses have certified accounts is a strong moat against AI / DIY accounting.
Even if the cloud-based accounting platforms move to offer certification as a more direct service, by for example rapidly connecting clients with accountants to do this, the fact remains that a human accountant intrinsically still needs to sign off, meaning continued work for (human) accountants for many years.
There is the factor that with increasing use of data and AI-based analytics by tax authorities worldwide, to flag and prioritize accounts for further inspection and audit, it is likely - if not the case already - that authorities view continuity of representation as a signal.
If you know or strongly suspect this to be the case in your jurisdiction, I’d encourage you to surface this with clients. In general, educating clients on the use of data analytics and AI by authorities is a win-win, and the kind of inside information that clients highly value.
You should stay on top of developments by the cloud-based providers to introduce new programs which connect clients with accountants, and quickly sign up for any such programs upon launch. The early bird catches the worm... and the reviews.
In many countries tax authorities have progressively been heaping responsibility onto accountants for all kinds of compliance, suspicious activity reports, AML, KYC and the like. In some countries an accountant basically works for the government at this point, more than the client!
The circumstances vary by type of company and jurisdiction clearly, but fundamentally the need for an accountant’s ‘stamp of approval’ ensures many accountants will keep their seat at the accounting table, for many years to come.
Recommendations for accountants in an AI world
In addition to the recommendations I’ve sprinkled above I would suggest the following ideas for protecting and growing your accounting business in an AI-world:
It goes without saying but use AI yourself, without delay, and become an expert on AI accounting for the benefit of your clients. And don't forget to tell your clients about this. Clients will stay with accountants who they feel are on top of new technology trends (this is nothing new...), rather than those who feel outdated or outmoded. As an accountant you naturally made a commitment to lifelong learning. AI is likely to be your biggest test, but you can do it.
Lower prices for your clients! There is going to be nothing more powerful than emailing your clients one day to say “Thanks to AI, I’ve been able to reduce your accounting costs this year.” Amazon Web Services email me every year or so, to notify me of a price decrease, thanks to efficiency gains and reduced hardware / bandwidth costs. It’s highly unusual, but that's what makes it feel amazing to a customer. I would never consider moving to one of Amazon's AWS competitors, for this reason alone, providing they keep it up.
Consider monthly subscription billings. Not for everyone perhaps, and it can certainly be contentious - many clients don’t like it - but there is an argument to consider moving remuneration to the same model that your (future) competitors (Quickbooks et al), are using, even if I do have mixed feelings on the practice personally. It certainly makes the value comparison more transparent, it evens out income to you, and reduces the impact of unexpected shock losses. It also allows for greater impact (and thus loyalty) of reducing prices.
If appropriate, try to strengthen your soft and interpersonal skills. I’ve touched on this above, but strong interpersonal relationships are going to be increasingly key in the future to dissuade clients looking across the dance floor. I accept that progressive changes from tax authorities may have pushed things in the opposite direction lately, but - and I know it sounds simple... try to make friends with your clients.
Adopt a growth mindset. I found my current accountant two decades ago, simply because as a young entrepreneur I deliberately went looking for a local accountant who was using the ‘latest’ tech (QuickBooks), and found him listed on the QuickBooks website. We’ve been working together ever since. There will be tremendous opportunities for accountants to get ahead of the competitive pack with AI technology, and shout about it. The window for the highest impact of this is small.
Learn about localized AI, aka 'on-premise deployment'. Not every business wants (or will want) all of their financial and other business records in the cloud, being trained on for who-know-what use by AI. Educate companies about the risks of doing this for starters, and develop knowledge about the on-premise options. Localized AI for the everyman is a nascent field, but basically the idea is that you can run ChatGPT type LLM's (AI) on customer’s own machines. I’d strongly recommend Twitter for staying abreast on this area (and AI in general), and following a couple of good AI accounts. As you dig into their posts and threads, Twitter’s algorithms will rapidly surface other AI tweets for you, from people you don’t explicitly follow. The tech talk may be overwhelming at first, but I'd encourage you to stick with it. A bit like osmosis, the terms and ideas will gradually sink in, and you don’t need to jump on anything right now. It's pretty complicated at the moment, and everything is developing and changing very rapidly, but in time this will get simple to implement, and you can be an expert on it.
Thought leadership. It's not a new idea but sharing insights about AI and accounting through articles, blogs, Twitter or speaking engagements can rapidly position you as an expert in this fast emerging field. You’ll attract new clients for one thing, and possibly unlock entirely new business opportunities within the industry itself.
Diversify your services. I know there are a ton of articles about AI accounting which seem to (frankly lazily) focus on this one idea more than anything, but ‘moving up the stack’ into areas like strategy, consulting, M&A, IPOs, loans etc, is valid. I would add and argue that 'AI consulting' in respect of accounting, would seem to offer the greatest potential.
Foster a culture of innovation. If you run an accounting firm, don't delay further. Embrace AI fully, and invite the person who first told you about it in for a friendly chat. Encourage everyone within your firm to learn about AI, and to research and try the latest AI tools.
If you have a penchant for code, or know someone who does, look into using AI to write custom apps, either for in-house use, for clients, or to make a tool or calculator you can put on your website (and attract valuable backlinks for SEO purposes). It's never been easier to do this, actually anyone can do it, and ChatGPT will generate fully working code, in almost any coding language, based solely on natural language prompts.
This has been by far the longest article on my site to date, and frankly the most challenging to write.
As I suggested at the beginning, the situation with AI and accounting is complex and nuanced. There will be undoubtedly be significant challenges for accountants ahead, but also great opportunities. There always is with new technologies.
If you are towards the end of your career, or perhaps have mostly older, traditional clients, maybe none of this matters, but if you are more than say 7 years out from retirement, I’d recommend not sleeping on AI.
If you're in the middle of your career, you may have been through perhaps one too many technological changes for your liking, but this one is as big as they come. You will inevitably find yourself competing with other accountants, not discounting your biggest likely competitor, which is the behemoth accounting platforms themselves.
I sincerely hope that you’ve found food for thought and useful practical advice in this article. If so, I would greatly appreciate you sharing this page with colleagues who may find it helpful, and on your websites (those backlinks again).
I promised to remind you to go check out ChatGPT. If you’d like to contact me for any reason, please reach out via Twitter.
Conclusion: Somewhat Safe (with caveats) | Time Period: 20 Years