Two thousand economists now agree on AI displacement. No firm has changed how it counts.

On July 13 the Stanford Digital Economy Lab published a statement of eighty-eight words titled We Must Act Now, organized by Erik Brynjolfsson with Ajay Agrawal, Anton Korinek and Tom Cunningham ^[1]. It holds that AI may become radically more powerful over the next decade, producing an economic transformation larger than the Industrial Revolution on a far shorter timeline, with large-scale job displacement alongside major gains in living standards ^[3].

The document is thin; the signature list carries the weight. It launched with more than two hundred economists and AI researchers, sixteen of them Nobel laureates, and is now approaching two thousand names ^[5], roughly a tenfold increase in signatories ^[13]. Daron Acemoglu and Simon Johnson, who took the 2024 Nobel, signed ^[6], despite Acemoglu's own work putting AI's total factor productivity gains at well under one percent over a decade ^[7]. Krugman, Ferguson and Cowen are on the same eighty-eight words ^[8]. So are Jeff Dean at Google, Jack Clark at Anthropic, the chief economists of OpenAI and Anthropic, and Yoshua Bengio ^[9]. The laureate count reached seventeen after publication, when one emailed asking to be added ^[11], one more than at launch ^[12].

Brynjolfsson, who directs the lab and co-founded Workhelix ^[2], told Forbes columnist Vibhas Ratanjee: "There is a tsunami coming at us of technical capabilities. And we're not prepared in terms of the organizational changes" ^[10]. He said he had once felt like a lonely voice on this, and singled out the last three or four months for the shift among skeptics ^[23].

The gap that should concern operators is not political and not regulatory. The statement's instruction is to build incentives and institutions that steer AI toward complementing people rather than merely imitating them ^[4], and as the column argues, no bill produces that outcome: it is a capital allocation decision, made repeatedly in business cases and investment committees, and settled by which line on a P&L an AI project is expected to move ^[20].

The measurement problem is the reason dashboards do not register it. GDP counts what is bought and sold, and per Brynjolfsson, with few exceptions something with zero price carries zero weight in official statistics ^[21]. His alternative, GDP-B, built with Avinash Collis, Erwin Diewert, Felix Eggers and Kevin Fox, asks not what you would pay but what you would have to be paid to give a thing up ^[14]. Applied to AI with Collis, Eggers, Sophia Kazinnik and David Nguyen, the average answer for giving up chatbots for a month was about $124, against a $20 subscription or nothing at all ^[15], a gap of roughly six times ^[18]. Across 115 million American adults the team puts annual consumer surplus at roughly $172 billion, which it says exceeds the AI industry's American revenue over the same period ^[16]. The arithmetic is straightforward: $124 a month, annualized, across that population ^[17]. It is also a consumer measurement. Nobody has run the equivalent exercise inside a company ^[19].

Watch whether any firm produces an internal version of that number, and whether business cases start naming complement or substitute explicitly rather than reporting cost saved ^[20][19]. The column frames the ledger error as the first of three items invisible from an ordinary executive dashboard ^[22]. The signature count keeps moving ^[5]; the internal metrics have not.