Or, ‘The Real Battlefield: AI’s Inroads into Value Creation within Accounting’
By Ric Payne
Four university professors sat the 3.5 Version of ChatGPT for the CPA Exam in May. And it failed. Rather badly, it turns out! Averaging a score of 53.1 across all sections.
However, it passed when the same researchers repeated the experiment on the upgraded ChatGPT version 4.0 only a month later. Rather impressively, actually. With an average score of “The chatbot received an 87.5 in the part that rated highest, auditing and attestation (AUD),” the researchers reported.
MORE on ARTIFICIAL INTELLIGENCE: Chris Vanover: Question the Why or Stay with the Status Quo | AI + MRI = Diseases That Doctors Might Miss | The Art of Prompt Engineering for Accountants | Staff Need Good Pay and TLC | Two Words Define Your Work Future in the AI World | Becky Livingston Launches “The B2B Marketer’s Guide to AI” | Getting Real: Accounting Tech Decisions You Need to Make Today | Generative AI: Should You Avoid It or Adopt It? | The Disruptors: Al Anderson on The New Manifesto for Accountants | Should Accountancy Account for Sustainability? | Firms Rev Up Expansion Plans | What an A.I.-Powered Workforce Means for Accountants | Jody Padar: Build a Practice that Works for You, Not Vice-Versa. | AI Systems Just for Accounting in Development | AI Is Not Your Enemy |
MORE CPA TRENDLINES coverage here.
Exclusively for PRO Members. Log in here or upgrade to PRO today.
This is a very interesting development for several reasons.
First, it’s amazing how quickly ChatGPT has improved. I doubt a human, even a smart one, would achieve a 32 percentage point improvement in an examination score in just a few months. This performance improvement begs the question: are we looking at the likelihood of close to 100% shortly?
A second point to note from this exam success is whether an AI Bot, or whatever it might be called, can perform most of the technical functions expected of a low/medium level professionally qualified accountant and do so 24 hours a day, 365 days a year. This represents a potential 673% productivity improvement over a human system with a very low error rate, no bathroom brakes, no idle chatter in the corridor, no requests for vacations, no arguments with bosses or colleagues, and total loyalty – at least until the Bot becomes sentient.
Tools that demonstrably improve productivity tend to be adopted quickly because of competitive forces. Still, the apparent cost-saving inevitably results in price-driven margin erosion as firms scamper to grab market share using price as a lever. This is the inevitable result because productivity improvements increase industry capacity and potential idle capacity costs – this would need another blog post to explain in more detail.
Another way to look at that is, and related to the above comment, the AI Bot would seamlessly do the work of nearly seven people. I’m not suggesting that the jobs of 7 people would be at risk, but you would be foolhardy to believe that no human jobs will be on the line, which has led some people to suggest that AI will not take your job but a person who knows how to use AI will. I agree, but I seriously doubt it will be a 1:1 transition. As an aside, I should mention that the people who would most likely be the candidates for reduction are the least productive ones in the current operating model, but ironically (and for reasons I might discuss at another time), they are probably the people you should seriously consider retaining. The real world works in mysterious ways, and one of the mysteries is why people (including partners) who operate at a high level in one operating model tend to sabotage an alternative (but better) operating model that does not align with their view of the world and the source of success.
And this gives rise to a third point. AI is good at what most humans are not good at (or don’t like), namely, endlessly engaging in a non-stimulating task while performing it at a high level of accuracy and speed. With that in mind, some people have suggested that using AI bots will go a long way in dealing with the talent shortage in the profession and free up talent for more valuable tasks.
I agree that this may help with the talent shortage challenge. Still, as for the second aspect of the proposition, I feel that the talent AI is likely to free up is not generally talent with a use value beyond how it is currently utilized. I, therefore, wonder what value other than cost saving is expected to be created from deploying AI bots. I also wonder whether the current talent shortage reflects an astute awareness by young people that a future in accounting is not a good career choice because of the role AI is expected to play in the profession. In other words, far from it being an answer to the talent shortage, it’s a contributing reason for it!
A fourth point that comes to mind is the question of professional integrity and the possibility of advisory disintermediation.
Let me explain.
I have prompted ChatGPT to analyze a five-year set of financial statements for a business and then prepare a report recommending what management could do to improve the profitability and value of the business. It did a credible job in under a minute, even on the first pass. I then did a deep dive into some of its conclusions (e.g., based on this analysis, what initiatives management could pursue to improve its gross margin), which was even more impressive. I had invested about 20 minutes in the analysis and report preparation by this stage and realized I could standardize the prompts and repeat this for many clients.
Three thoughts immediately came to mind from this simple exercise.
First, if I report to a client, what would I charge? My judgment is it would be worth way more than any time-based billing system would mandate, but given it was not my experience or insights that was the source of the value, can I, with integrity, charge $X,000 and if so, should I advise my client that I used ChatGPT to do the analysis and write the report. Hence, my value was to draw on my skills to approve and complement its work. In this scenario, am I being an imposter? And if I did tell the client that the analysis and report was an AI Bot’s output, would the client value it, or me, less? Given that AI is headline news and the possibility that I have never created a review like this for the client, is my client likely to suspect it is not my work and call me out?
Secondly, assuming full disclosure, how would I respond if the client said, “Give me the prompts, and I’ll work with ChatGPT myself.” Do I charge for the prompts? If so, what are they worth? Oops, there’s an idea: what if I put together a booklet specifically aimed at SME managers with prompts they could use to extract insights into their business from an AI tool like ChatGPT? If this hasn’t been done, I suspect it will soon. Frankly, I can’t see AI giving a firm a sustainable competitive advantage because it is not a proprietary source of competence, and its low cost will result in rapid deployment.
But permit me to offer a warning (or advice) here. I have had the opportunity and privilege to work closely with hundreds of SME managers (and a couple of very large companies) in the past 50 years. Although I may have started my analysis of an organization through the lens of financial reports, it became evident very quickly that the challenges and opportunities are not financial but involve the interplay of complex relationships between people and processes, both formal and informal, within the organization and between it and its various stakeholders. The insights needed to deal with these issues for a given business at a given time is not something AI is likely to do well, at least in its present configuration, simply because this is not an algorithmic issue. It is a systemic issue.
The warning I offer here is that AI may give you and your client some comfort from an appearance of credibility (after all, it is supposed to have scoured all the available knowledge on the internet). Still, the solution to a client’s declining GM% challenge, for example, is not a sales mix, COGS, or pricing issue. Still, it may result from the behavior of a toxic manager who the business owner thinks is fantastic. But also note this is not just a warning. It’s an opportunity because firms that can add their intellectual and experiential insights to AI’s grunt work will win. In other words, sustainable competitive advantage will go to those firms or individuals who know how to use AI and do what AI cannot do.
Third, any organization or person who has access to the financial statements of an economic entity could use AI in the way I have outlined above. Let me put that more bluntly. Cloud-based organizations currently offering accounting packages could easily and seamlessly offer a low-cost (maybe zero) option for an analysis and profit improvement report. This could go close to disintermediating many of the accounting profession’s advisory aspirations.
My tentative conclusion on AI’s impact on the accounting profession is tempered by several deep concerns outlined above.
It’s shaping up to be an incredibly powerful tool, and the clear trend of it having even more power is scary. People are excited to see how it can improve productivity by dealing with such matters as creating marketing content, answering emails or setting up a chart of accounts, but these are trivial. Its potential in advisory is the thin edge of a wedge into the real value-creation function of the profession, where the battle will take place.
AI will have a far more significant impact than the introduction of computers, traditional productivity software, and communication tech has had on the profession. These things empower people to work more efficiently and effectively. However, AI can create that value largely without much input from a human – this is somewhat of an exaggeration, but I believe it’s not far from the truth.
I suspect it will fundamentally change the structure of the industry, its competitive forces and the design of its dominant business model. The profession’s belief that it is the unassailable Trusted Advisor will be challenged. As happens when any disruption occurs, there will be winners and losers. To be amongst the winners, firms should immediately look beyond the trivial productivity improvement potential of the tool. This is a given and will soon fall into the table stake arena.
You need to be asking yourself the following questions:
1 – What might our firm look like if we embraced AI as a cornerstone of our advisory business?
Let’s set out to win in this space by setting an ambitious goal to be the dominant player within five years in our chosen geography. I strongly recommend you use a strategic thinking framework. To help you here, you need to go no further than study Roger Martin’s methodology described in his book with A.G. Laffley called Playing to Win: How Strategy Works. They describe a 5 level cascade of designing and implementing a strategic initiative, and it is by far the best framework I have seen and used.
2 – What resources am I willing to commit to this opportunity?
This is an investment in the future of your firm. Don’t expect to get a quick return. Remember, your less sophisticated competitors will look for quick productivity gains and will be drawn into the vortex of intense price competition. They will not be your competitors for long. Let them play in that space while you focus on the long game. They will not attract the type of clients who will see the value in working with a professional who understands the technology and its role in the wider systemic environment of a business entity. This is the first thing you need to get your head around. Where do I want to play, and what value do I intend to create – in Martin & Laffely’s words, “How can I win?” Be transparent, tell your prospects you’re going to use AI to do the grunt work, and you are going to help them fashion that into a winning strategy, one that’s designed to increase the odds (but not a guarantee) of success, and then you’ll help them execute that strategy. This is an incredible opportunity to jump out of the box while everyone else is mucking around, trying to save 30 minutes setting up a chart of accounts.
3 – What business structure and management support system must I have in place?
The strategy I’m discussing will not work if you attempt to incorporate it into your business model. It is a fundamentally different business that must be treated like a startup. You will have resistance from some (maybe all) of your team and shareholders, so be prepared. This is unfortunate, but some of them will feel threatened and may even sabotage your attempts to create the business despite saying they’re committed to helping.
If this sounds hard, it is! But therein lies your opportunity for securing and possibly even sustainable competitive advantage. Because what’s hard for you will be impossible for a challenger who is comfortable traveling the path of least resistance. Believe me, I’m right about that after spending 50 years in the trenches.
I would love to hear from anyone who can challenge my logic and the conclusions I have reached.
PS: For fun, I asked ChatGPT to read this post and suggest some headings. In less than two seconds, these are the ten it came up with. My working title and preferred final choice is a slightly modified version #6, but I decided to share them all because each raises a question that I believe needs to be discussed.
-
- From Failure to Success: The Rapid Evolution of AI in Accounting Tests
- Beyond Numbers: AI’s Monumental Leap in CPA Exams
- The 24/7 Accountant: AI’s Promise and Threat to Human Jobs in the Profession
- Bridging Talent Gaps: Can AI Truly Fill the Void in Accounting?
- AI and Professional Integrity: Navigating the New Frontier
- The Future of Advisory: AI’s Potential to Disrupt Traditional Practices
- Pricing AI’s Expertise: The Ethical Dilemma of Value Attribution
- DIY Analytics: The Emergence of AI-Powered Client Self-Service
- Advisory at Risk: The Impending Threat of AI to the Accounting Profession
- The Real Battlefield: AI’s Inroads into Value Creation Within Accounting
