The Big Four Audit Firms Built Their Business Model on Junior Labour. AI Just Made That Labour Redundant.

Seven people at a South African startup are shipping more product faster than teams ten times their size. The same dynamic is playing out inside every major professional services firm — and the billing model that depended on armies of graduate hours has nowhere to hide.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-big-four-audit-firms-built-their?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Tayla Dandridge, co-founder of Stub, describes an intense mix of operational excitement and dread that defines the daily reality of building with AI in 2026. A new model drops and her seven-person team needs to absorb it, assess it, and decide whether to rebuild around it — before competitors do. That combination of possibility and pressure is not unique to Stub, but it is unusually visible in a company this small competing in a market this significant: the spaza shop, the side hustler, the micro-entrepreneur running a business through a personal bank account. Stub has nearly ten thousand entrepreneurs on the platform, integrations with iKhokha, Capitec, and Yoco, and a paid tier at R189 per month that does what Xero, Sage, and QuickBooks — the whales, as Dandridge calls them — were never designed to do for this market: get the accounting work done automatically, rather than providing pieces of tools for people who already understand their own financials.

The Stub story is interesting in its own right. The argument it carries is more important — and it runs well beyond accounting software for spaza shops.

CONTEXT AND BACKGROUND

The professional services business model has been structurally consistent for decades. A large firm hires cohorts of graduates, assigns them high-volume cognitive work at entry-level billing rates, supervises them with senior professionals who charge at premium rates, and extracts margin from the difference between what it pays juniors and what it bills clients for their time. The pyramid works because junior cognitive labour — reconciliations, research, data entry, first-draft analysis, basic due diligence — was cheap to hire, reliable to produce, and essential to the volume that sustained the model.

AI is dismantling that pyramid from the base. A Harvard University study examining 62 million workers across 285,000 firms found that junior positions are shrinking at companies integrating AI, with researchers explicitly warning that artificial intelligence is eroding the bottom rungs of career ladders — eliminating the roles that once served as the training ground for early-career professionals. Entry-level job postings in the US fell 35 per cent from early 2023 to mid-2025, according to Revelio Labs, with junior roles in software development and data analysis down as much as 67 per cent in some sectors. Stanford University research found that workers aged 22 to 25 in roles most exposed to AI — including accounting, software development, and customer service — experienced a 16 per cent relative drop in employment in less than three years.

Fortune reported in June 2026 that banks are cutting junior analyst classes by as much as two-thirds, with JPMorgan, Citigroup, Goldman Sachs, and Standard Chartered all publicly signalling that AI will reduce staffing levels across middle-office and entry-level functions. Debasish Patnaik, senior partner at McKinsey's QuantumBlack AI consulting arm, noted that while graduate intakes will shrink, banks are unlikely to shed them altogether — "banking is an apprenticeship business," he said. "Today's junior analysts become tomorrow's managing directors. Senior judgment cannot be manufactured laterally." Investment banks and consulting firms that previously hired large graduate cohorts directly from campus have quietly reduced class sizes or paused programmes entirely, narrowing the entry path for new graduates into the financial sector at the precise moment AI is automating the work those graduates would have done.

INSIGHT AND ANALYSIS

Dandridge’s reframe of what Stub is building contains the most strategically important question in the current AI and professional services landscape: “Are you a software that provides pieces of tools, or are you a platform that gets the work done?” Stub has moved decisively toward the latter — automating reconciliations, payment reminders, and categorisation so the entrepreneur does less of the work themselves. That is not a product decision. It is a philosophy about where value actually lives, and it applies with equal force to every accounting firm, consulting firm, and advisory business that currently charges clients for access to human labour performing tasks that AI can now perform.

The disruption in professional services is arriving from both ends simultaneously. From below, AI-enabled startups like Stub are serving markets the incumbents ignored — too informal, too small, too financially unsophisticated — at price points and simplicity levels that the incumbents’ cost structure cannot match. From above, boutique advisory firms and specialist technology-enabled challengers are using the same AI tools to compete for the work the Big Four depend on — delivering consulting, tax, and risk advisory work with smaller teams at lower costs, because AI has compressed the labour input required. Research on the consulting industry found that midmarket firms describe AI as eliminating the disadvantages they previously had relative to the Big Four — no army of talent, no advanced internal tools — while empowering their people to be more productive. A firm that previously had to decline ten project enquiries to focus on two can now respond to twelve.

I have previously written about the disappearing entry-level roles in the knowledge economy — the way AI is automating the cognitive starter tasks that used to absorb new graduates and build the judgment that senior professionals depend on. The professional services dimension of that argument is the one that carries the most immediate consequence for the firms themselves. The Big Four did not only rely on junior labour for revenue. They relied on it for training. The graduate accountant spending three years doing audit fieldwork was not only generating billing margin. They were developing the professional judgment that made them a partner twenty years later. When AI takes the fieldwork, the billing margin disappears first. The judgment pipeline disappears second — and it disappears more quietly, accumulating as a talent deficit that will become visible in the senior ranks years from now.

LeadDev described this as the great engineer hiring paradox: companies claim they cannot find experienced engineers while simultaneously eliminating the junior roles that produce experienced engineers. The same paradox is playing out in accounting, law, and consulting. The firms reducing their graduate intake to capture AI efficiency gains today are making a decision that will constrain their senior talent pipeline for the following decade — and most of their partners and boards have not yet modelled that consequence.

IMPLICATIONS

For the boards of South African professional services firms, the Stub story and the structural research together carry three specific implications that deserve explicit attention.

The first is about the billing model. Time and material billing — charging clients for the hours that humans spend on their work — is the most exposed business model in the AI era, because AI directly compresses the number of hours required to produce the same output. Firms that have not begun designing alternative pricing models — outcome-based fees, subscription arrangements, fixed-price deliverables — are building increasing structural risk into their revenue base with every AI tool their junior staff adopts to work faster. The efficiency gains AI provides to individual professionals are gains that erode the firm’s own billing capacity unless the pricing model changes to capture value from outcomes rather than inputs.

The second is about the talent pipeline. The firms cutting their graduate intake are making a short-term efficiency calculation with a long-term talent consequence. The apprenticeship model — through which the accounting and consulting professions have always built their senior capability — depended on juniors doing the volume work that AI is now automating. There is no clear alternative pathway for building the judgment, the client relationship skills, and the professional experience that senior roles require. The firms that are thinking seriously about this are redesigning junior roles to be judgment-first rather than task-first — deliberately preserving the human responsibilities that build capability even as AI handles the volume. Most firms are not yet having that conversation.

The third is about competitive exposure. The Stub story illustrates what AI does to competitive barriers across the entire market. The whales — Xero, Sage, QuickBooks — built their positions over decades on capital, distribution, and the network effects of large user bases. Seven people with AI are challenging them from below. The Big Four built their positions on regulatory licences, client relationships, and the scale advantages of their junior pyramids. Technology-enabled boutiques with AI are challenging them from the middle. The competitive moat that incumbency and scale provided has not disappeared — but it is narrower than it was, and it is narrowing faster than most incumbent firms have acknowledged in their strategy planning.

CLOSING TAKEAWAY

Dandridge’s intense mix of excitement and dread is not a startup founder’s quirk. It is the most honest description available of what it means to build in an environment where the tools that define competitive advantage are changing faster than any governance or strategy cycle was designed to absorb. The seven-person team challenging the whales from a South African office is a specific and local story. The structural argument it carries — that AI compresses the competitive moat that scale and capital used to provide, across every professional services market at every tier simultaneously — is a global one.

The Big Four are not going to disappear. Their regulatory licences, their client relationships, and their institutional trust are real and durable. What will change — what is already changing — is the economics of the model that sustains them. The pyramid built on junior labour is losing its base. The firms that redesign their model before the compression becomes a crisis will be better positioned than those that discover the problem in their revenue and their talent pipeline at the same time.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net