About 50 years ago, a time span which is somewhere between “ancient history” and “I remember when”, depending where you are on the Gen Z <-> Boomer continuum, accountants plied their trade in a way that would blow most young minds of today. Transactions were recorded line by line in ledger books, in pencil. Ledger books are still available today, but you’d have to be some sort of luddite kook to use one. Absolutely no one is into retro accounting. Each transaction’s debits and credits were recorded in two columns; the numbers were added manually. All very clean and simple and, well, maybe those were the good old days of accounting because financial statements were actually relevant and readable at the time. Modern IFRS statements are little more than numerical Rorschach tests. What do you see? Profit? Or a dingo?
In 1982, it was estimated that there were somewhere between 500,000 and 850,000 accountants at work in the US, a numerical imprecision which is kind of funny when you think about it. Of all people that should know…but anyway the range depended on the definition of bookkeeper vs accountant vs auditor. It was probably an important distinction, at least to an auditor, who might well fly off the handle if called a bookkeeper, in a measured and controlled kind of way.
About 30-35 years ago, along came a new software tool called a spreadsheet, and it threatened to wipe out the entire profession. Manipulating numbers became infinitely easier. Once captured by a data entry person, the whole game was done – the ability to balance and reconcile and subtotal was instantaneous, easily proofed, and just all around revolutionary.
A funny thing happened though – the number of accountants went up. Significantly. In 2023, there are an estimated 1.6 million accountants and auditors in the US. It’s true that there are far more businesses, that are more complex, but nevertheless a gigantic leap in accounting productivity did not decimate the industry. The basic function of the occupation – to properly and accurately record the record of transactions an entity enters into, and reflect this information in financial statements – is now the kernel for vast new enterprises that encompass analytics, auditing, etc.
The new tools of the trade – spreadsheets, databases, connectivity – may have wiped out the occupation of legibly printing numbers in a journal and accurately summing them, but the technology replaced that tedium with a near-infinite array of possibilities. While it is a bit sad that the ability to hand-write legibly and clearly is now a dying art form, it is what it is, and few would argue that the tech that wiped out what we knew of as “accounting” was bad for the profession.
And so it goes with AI. We will lose some job functions, but others will be created, and we can’t even envision what those will be yet.
The hydrocarbon sector might lean into fascinating new territories as well. There will be efficiency gains, no doubt, such as operational optimization and predictive maintenance. The level of gains to be had will be somewhat tied to the level of data available, and the energy business is very data-heavy.
One area of extreme data collection is reservoir management to enhance optimization strategies. The industry has been gathering crazy amounts of data for decades – seismic, drilling activity, production activity – and has made great strides integrating all of it.
A famous (for energy geeks) example, almost a decade old, was Saudi Aramco’s announcement that they had achieved a trillion-cell reservoir simulation run on a petroleum reservoir, mapping the flows of oil, water and natural gas. This happened in 2016. Imagine what SA can now do with that data and the power of current AI technology. Imagine what can be done with today’s sensors and fluid tracers and frac mapping and rock knowledge, all integrated and mappable. Soon this will be the norm at all bigger fields.
Expand Energy, formerly Chesapeake before the merger with Southwestern, outlines in their latest IR presentation some very cool progress made with AI and machine learning. Telemetry data and machine learning is helping with bottom hole assembly diagnostics, real-time slide-vs-rotate drilling parameters, and hole cleaning analysis to iterate into faster operations. That’s how they’re drilling 5-mile laterals in 5 days, a staggering feat.
While machines will obviously do most of the work, it is anyone’s guess what sort of job functions might materialize out of all of this. Some occupations will likely die; others will come into existence.
The predictions of possible outcomes are as boundless as AI itself. Elon Musk thinks AI is going to replace pretty much all human labour, and thus is pushing for not just universal basic income but “universal high income” which wow, would turn us into some sort of blob-like super-consuming mass of humanity, kept alive in decrepit perpetuity by AI-medicine wizardry…try wrapping your head around that. But crazy sci-fi forecasts are a dime a dozen so don’t waste too much time on it.
It seems more likely that in a lot of fields the new technology will allow people to more stuff, heighten analyses, shorten research times, etc. Where that goes is anyone’s guess, but if there can be room for pessimistic views, there can be room for optimism as well.
And let’s not forget that the whole frame of reference is being rattled by AI.
For one thing, the whole employment playing field is changing with AI. Applicants can flood employers with perfectly scripted cover letters and customized resumes that will wreak havoc with resume screening tools. (Hey look – 2,500 perfect applicants! Nothing fishy there. But who’s judging? AI applicants vs. AI job screeners. This all devolves rapidly into nonsense.)
And then there’s the breakneck bubble pace that’s making the dot-com boom look like nothing. A smart and fun investment manager named Harris “Kuppy” Kupperman wrote a great article about the growing insanity of all the investments, questioning how on earth they will all make money. By his calculations, for example, total US datacenter investment for 2025 will be a minimum of $400 billion, based on public disclosures, a figure that is limited because of bottlenecks to build outs. As he puts it, “those who are spending on these data centers are beyond desperate to get them operational.” He makes reasonable assumptions as to depreciation rates for buildings and equipment, and concludes that this year’s crop of data centers will incur about $40 billion of depreciation, and yet generate only about $15-20 billion in revenue. Compounding this obvious problem is that, in order to gain traction, many AI providers are giving away the product, with zero revenue. And somewhere in there is supposed to be a return on capital. To get a 25 percent gross margin and cover depreciation, the industry would need hundreds of billion in revenue on this investment. Not even close. (As he points out, industry titan Netflix has only $39 billion in revenue last year on 300 million subscribers…).
AI is not uniform, of course. Some of those data center investments will generate operational savings, or increase revenue. Some will unearth new medicine or manufacturing processes or metallurgical breakthroughs. Who knows what. But the vast frenzy of service-providing AI (it’s going to replace all lawyers!) is at best a dubious business model, when it’s being given away for free now. That won’t last forever.
For the energy audience, which is presumably most of you, the one undeniable reality of AI is: energy consumption is going through the roof. Renewable energy simply doesn’t cut it. New nuclear is the dream, but is a decade away at any significant scale. The investment train might not keep the same momentum in a decade, but in the interim, the call on energy is going to be huge. There’s no real way around that.
Long story short: Don’t get too distraught by the doomsayers, and don’t expect nirvana either. Maybe the most logical expectation is that things will get weird. In the past few weeks, the web has become riddled with footage of little robots boxing each other, running 100 m dashes, etc., some successfully, some hilariously…a lot of $ will go towards perfecting those things, but then what. No matter how good they get, I’ll never drop a nickel to watch two robots box each other.
At the end of the day, we’re all humans. Some are borderline, but still. We have an inherent desire to do things: work, create art, build furniture, raise kids, travel, juggle, who knows what. It is absurd to think humanity will sit idly and watch computers and robots do everything. I can guarantee you I won’t; if it comes to that I’ll go rogue and start smashing bots with a baseball bat, or bring down drones with Molotov cocktails, or something just generally antisocial. I suspect I won’t be alone. See y’all underground. Signed, for authenticity, FU ChatGPT.
Explore the lighter side of energy, and think of it as you never have before in The End of Fossil Fuel Insanity – the energy story for those that don’t live in the energy world, but want to find out. And laugh. Available at Amazon.ca, Indigo.ca, or Amazon.com.
Email Terry here. (His personal energy site, Public Energy Number One, is on hiatus until there are more hours in the day.)