Last year Stuart Kirk upset his bosses and then quit his job as head of responsible investing at HSBC Asset Management after advising investors to stop fretting about climate change since human ingenuity would probably cope with it.
In a recent column in the FT he was at it again. Another thing investors didn’t need to pay much attention to, he suggested, was corporate performance.
Eh?
Well, he said, consider that in the US over the last century the average real rate of stockmarket returns has stayed subbornly at 6.7 per cent. Then add that profitability over the very term appears to revert to the mean; and finally – the killer – that since 1926 all net US stock exchange wealth has been created by just 4 per cent of quoted companies. That’s right: 96 per cent of companies generated none at all – and that in a period during the latter half of which managers were explicitly pursuing a policy of maximising value for shareholders.
As Kirk indicates, that raises some interesting questions. Why, he asks, ‘are there so many poor-performing shares when 250,000 students are enrolled as masters of business administration each year? And what’s been gained from the almost trillion dollars per annum spent on consultancies worldwide?’ Or, we might add, from the multibillion-dollar global leadership industry, not to mention the undergraduate business degrees awarded to more students around the world than in any other subject?
It’s a good question. So, if overall they don’t do much for investors, are CEOs with MBAs good for business? As it happens, that’s the question the very serious MIT researchers Daron Acemoglu and collaborators sought to answer in a study looking at what happened when companies replaced a CEO without an MBA or business degree with one with such a qualification. The answer: not much. There was no evidence that business graduates increased sales, productivity, investment or exports. What they did do was reduce pay growth for the workforce and increase the take of shareholders and managers. This boosted the share price in the short term – but the authors speculate that over time it could damage profitability: there was no increase in investment allocations, and higher-skilled employees were more inclined to leave after their relative pay cut.
Meanwhile, more and more day-to-day management is being carried out by algorithms and AI rather than humans. For example, Uber now knows enough about individual habits and preferences to offer the same job at varying prices to different drivers according to the behaviours they want to evoke; or to ‘deactivate’ them, sometimes permanently, without human intervention (the company denies this). Companies routinely use technology to surveil and control their workers – chaplains, funeral directors and care workers as well as telephone agents and clerks and almost anyone working from home – much more closely than human managers ever did before. And with the arrival of ever more capable AI software – decision aids, for instance – the proportion of jobs it could credibly eat expands as fast as America’s waistlines.
You might see where this is leading. If CEOs as a group, whether qualified or not, are not very good at either benefiting the investors they are supposed to be acting for or boosting business performance; and if for this unstellar performance they are currently in the biggest US firms taking home 399 times as much as the average worker – and in the case of Amazon’s Andy Jassy’s staggering 2021 stipend of $213m, 6,472 times – isn’t there a case for at least thinking about letting software eat them too?
Some are already doing so. In 2015, Devin Fidler of the Institute of the Future described a rudimentary virtual ‘iCEO’ that had been programmed to automate complex management activities by breaking them into smaller chunks. Reporting ‘amazing’ initial results, he warned senior managers who assumed they were immune to substitution to think again: ‘It will not be possible to hide in the C-Suite for much longer. The same cost/benefit analyses performed by shareholders against line workers and office managers will soon be applied to executives and their generous salaries’, he predicted.
Eight years later, I can’t find a mention of ICEO on the IF’s current website. But lo and behold, Fidler’s prediction is coming to pass. Last year a Hong Kong firm, NetDragon Websoft, replaced its CEO not with an MBA but an AI-powered ‘virtual humanoid robot’ it named Tang Yu. No experimental startup, NetDragon Websoft is a $2bn, 3,300 person quoted firm in ‘massively multiplayer online gaming’ – which peresumably it a head, or brain, start with AI. And so far Ms Yu is doing a creditable job. As of March 2023 Netdragon was outperforming Hong Kong’s Hang Seng index by 18 per cent.
Tantalisingly, we don’t know much more than that. And six months is clearly too soon for definitive conclusions. But some ideas suggest themselves. On the one hand, CEOs are unlikely go without a fight. On the other, given that most senior executives accept that their companies make bad decisions at least as often as good ones, and no amount of trying has yet established a correlation between CEO pay and corporate performance, activist and other primarily stock-market-driven investors may well start to take a dimmer view of the returns they are getting from the human CEO’s salary millions. What’s more, with the rapid advance of AI, Ms Yu 1.0 is liable to be succeeded by versions 2.0 and 3.0 in short order, and probably in a number of competing flavours: hardcore, B-Corp, and virtual Elon Musk, say. In other words, while NetDragon Websoft may be the first substantial firm to be managed by wire, it will almost certainly not be the last.
As for the rest of us, the implications aren’t entirely comforting. True, when I asked Bard, Google’s large language chatbot, if maximising shareholder value was a good corporate strategy, it replied cautiously that while it might be in some circumstances, strategy should also take into account the needs of other stakeholders. So far so good. But models ‘learn’, and if they learn as Acemoglu’s MBAs do, that leaves plenty of opportunity for the development of self-fulfilling prophecies resulting in less balanced outcomes. As Stanford’s Bob Sutton put it in an entertaining recent discussion with Rita McGrath, ‘[What my engineering colleagues] teach me is that AI algorithms are just a function of the human being who writes them. So maybe we just have to have fewer assholes writing algorithms’. Well, yes – but that hardly seems the stoutest defence against the robots now beginning to mass at the gate.