AI Lemons
In August 2025, a report for the Australian government produced by Deloitte, a Big Four consultancy, was revealed to be littered with errors dreamed up by AI. The report was reissued sans the offending mistakes, and a partial refund was made along with an apology. A few months later, another report by the firm, this time for the Canadian government, was found to contain phantom citations likely fabricated by AI. These back-to-back scandals belie Deloitte’s “Trustworthy AI™ approach to building AI that is safe, secure, compliant and scalable”, inviting critics to suggest that anyone looking to contract these firms... would be better off signing up for a ChatGPT subscription.
The consulting industry is no stranger to scandals and controversies. Top firm McKinsey has settled hundred-million-dollar lawsuits for instigating the opioid crisis and paying bribes. Boston Consulting Group, another prestigious practice, was found complicit in potential war crimes, as well as bribery and corruption. PwC, another Big Four firm, traded on insider information and enabled widespread tax evasion.
Despite the ensuing outcry and obligatory handwringing, repercussions were negligible. The consulting market has maintained consistent growth and is forecasted to continue growing through 2029, fueled by client demand for AI, cybersecurity, and digital product engineering.
Yet something about AI seems different. Anyone can now afford a McKinsey in a Box to replicate the business advice typically dispensed by consultants in slick, well-formatted, confident slides within minutes. Consulting firms, desperate to stay ahead, are investing heavily in the same technologies to generate client deliverables with reduced headcounts. Clients, aware of the cost savings, are demanding discounts, leading to a revenue squeeze. Even McKinsey is perturbed. Is AI the end of consulting?
What Do Consultants Do Anyway?
To displace consultants, AI would need to do their jobs cheaper, better, faster, or render them irrelevant. But what do consultants do anyway? Popular answers on the Internet (Reddit) include:
The job seems so arcane that even consultants have trouble articulating themselves.
My favorite is this nifty summary by kswankx2, 137 upvotes 👍
Three types of work,
- Capability – the client can’t do it themselves
- Capacity – the client could do it themselves, but need it sooner / disrupting their other work
- Credibility – the client knows the answer, but needs the rubber stamp to convince the exec/board.
These answers, however, missed what clients are actually doing: outsourcing risk to consultants.
To quote the legendary Warren Buffett, risk comes from not knowing what you are doing, especially with new technologies or in unfamiliar markets. Recruiting teams of specialists for one-off initiatives is unsustainable. Strategic decisions could be controversial. Resourceful operators have little use for consultants. But for many organizations, bringing in external expertise or independent assurance by a household name like McKinsey can save or shield its executives from many a failed project or unpopular decision. Hiring a consultant mitigates the risk of screwing up due to lack of expertise, experience, resources, or buy-in. Consulting is an insurance policy for organizations.
As with insurance, occasionally a claim is made, and the consultant takes the fall. The rare scandal comes as a part of the deal, and the big firms ride on their reputations, armed to the hilt with legal, public relations, and HR strategies – the cost of doing business.
For now, AI is considered property and thus cannot assume accountability – and risks – in the same way. Since AI cannot be litigated for mistakes and wrongdoing, it is hard to imagine corporations, executives, and board members, who can be held responsible, staking bottom lines, reputations, and careers on "AI made me do it”. Explaining decisions made with AI can get involved, much less navigating the legal quagmire surrounding AI ownership and liabilities. Hiring a consultant is perhaps more expedient. Consulting, particularly big names with reputation reserves to spare, seems safe for the foreseeable future.
The True Peril
However, this does not mean that the industry is free from the threat of AI. Large Language Models, increasingly used to analyze data and create deliverables, are ghost machines unconcerned with physical realities and prone to hallucinations and inaccuracies. AI agents, deployed for administrative tasks such as taking notes and summarizing research, have been reported to interpret given goals creatively, with preposterous results. AI models fed recursively on generated and junk data are shown to degrade over time and exhibit brain rot. Consultants can no longer get away with using the same AI systems as their clients to produce even mediocre work, let alone, as in Deloitte’s case, defective deliverables – or lemons.
In economics, a lemon refers to a poor-quality product. In a market where goods are sold honestly or dishonestly; quality may be represented, or it may be misrepresented. When buyers cannot easily distinguish between high and low quality, they become more cautious, only willing to pay a lower price for fear of being stuck with a lemon. This leads to sellers willing to sell only when they have a lower quality product that can satisfy the budget. Over time, people who are willing to offer inferiors goods – maybe by passing them off as premium wares – remain, driving honest sellers, and eventually the entire market, out of existence.
Lemon markets exist with or without AI. It is an open secret that some consultants cut corners by repackaging solutions (“industrializing proven frameworks”), recycling decks (“leveraging our knowledge base”) and passing off inexperienced resources as experts (“on-the-job learning”). A consultancy’s willingness to endorse or enable morally dubious and possibly illegal decisions for a fee should signal a disposition to engage in such behaviors themselves. Instead of mitigating risks for their clients, these players become risk factors themselves. AI, with its ability to rehash volumes of thought-terminating advice to think outside the box, will further dilute the sector of subject matter mastery and keen judgement, expediting the demise of the industry.
The real danger to consulting is not AI or technological progress or management fads; it is the malaise bred by years of easy profits propped up with pedigrees, eloquence, and demeanor. Perhaps the way out – and a real shot at True Transformation, the holy mantra of consulting – is through the costly graft of cultivating deep expertise, astute insights, and demonstrable credibility.
But can consultants overcome inertia and rise above sanctimony before the industry languishes, clients defect, and the market collapses upon all its players?
Further Reading:
The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. Akerlof, G. A. (1970). The Quarterly Journal of Economics, 84(3), 488-500.