Remember the good old days of professional services? A sturdy pyramid of humans arranged in a beautifully rational hierarchy: armies of fresh-faced graduates at the bottom doing the grunt work, a sensible middle layer of managers telling them what to do, and a select few partners at the pointy end taking client calls and collecting the profits. This model has defined legal, accounting, and consulting firms for generations. It was simple, effective, and made everyone money.
Well, AI just is toppling that pyramid. Say hello to the Diamond.
How We Got Here
Professional services firms have always been built on a fundamental trade-off: partners exchange their knowledge and client relationships for the labour of juniors. The partners bring in work and provide high-level guidance; the juniors actually do most of the work. This is not a secret. It's explicitly baked into the business model.
When you hire a Big Four accounting firm or a prestigious law firm, you're paying partner rates, but much of the actual work—the document review, the contract drafting, the financial statement checking—gets done by 25-year-olds with far lower billing rates. These bright-eyed associates and analysts gain experience on the job, gradually move up the pyramid, and eventually (if they survive the tournament) become partners themselves. It's a beautiful cycle.
The pyramid works because the bottom is wide: you need lots of junior people to handle the volume of routine work. The system has survived for over a century because:
- Those grunt tasks require human intelligence
- That intelligence is expensive enough to command professional salaries
- But cheap enough (compared to partners) to make the model profitable
Enter generative AI, which is extremely good at the exact kind of work that entry-level professionals used to do. As Goldman Sachs analysts noted in their research, AI could expose up to 300 million full-time jobs worldwide to some degree of automation. Routine white-collar tasks are squarely in the crosshairs.
The Great Compression
What happens when an AI can do in seconds what used to take a first-year associate hours? The pyramid compresses. As one Big Four innovation leader put it, AI will "take over mundane tasks such as data entry" while humans handle judgment and client relationships. This sounds innocuous until you realise those mundane tasks made up 70-80% of what entry-level staff did.
Already, we're seeing evidence of this compression. A 2025 analysis from Thomson Reuters shows that associates now make up a smaller proportion of law firm attorneys than they did 15 years ago (down from 44.5% to 40.2%), while the share of senior lawyers (non-equity partners and counsel) has grown. The classic pyramid is morphing into something more diamond-shaped: narrower at the bottom, wider in the middle, and still narrow at the top.
Where are we seeing this happening? Reports have noted that AI tools are reducing human contract reviews by 40%. Also, Law firms like Allen & Overy have deployed AI assistants (known as Harvey) to all 3,500 of their lawyers, letting them generate drafts and research at lightning speed.
The financial industry is seeing the same pattern. Banks like Goldman Sachs and Morgan Stanley are "weighing cutting analyst hiring by two-thirds" as AI takes over tasks like updating financial models and pitch decks. As one bank executive bluntly put it: "The easy idea is you just replace juniors with an AI tool."
This creates an interesting dilemma. As Casey Flaherty of LexFusion noted, "It takes four years to make a fourth-year associate" in the traditional model—juniors learn by doing large volumes of simple work. If AI handles 80% of that work, firms face a training challenge: "how to train young lawyers when the traditional brute-force methods are no longer viable—because the machines are superior at entry-level tasks."
How do you develop expertise when the traditional ladder has lost its bottom rungs?
The AI Slide: When the Staircase Disappears
This is what I call the "AI Slide" phenomenon. Imagine professional development as a staircase, where novices enter at the bottom step and gradually climb upward through experience. AI is removing those foundational steps entirely, creating a slippery slide where the crucial bottom rungs of the ladder simply aren't there anymore.
The problem isn't just about losing practice time. It's about losing the cognitive processes that practice enables:
- Weakened fundamentals: Without wrestling with the basics, the underlying principles and nuances aren't deeply encoded. The "why" behind the "what" remains elusive.
- Inability to "latch on": Learning is hierarchical. Advanced concepts build upon simpler ones. If the foundational steps are missing or merely observed (by reviewing AI output), learners struggle to connect new, complex information to a solid base.
- The illusion of competence: Reviewing or lightly editing an AI-generated output feels productive, but it's fundamentally different from creating it from scratch. This can lead to novices (and their managers) overestimating their true understanding.
- Impaired critical evaluation: How can someone critically evaluate the output of an AI performing a complex task if they've never performed the underlying steps themselves? Without the ingrained experience of the fundamentals, evaluating sophisticated AI output becomes superficial guesswork.
Some firms are responding to this challenge creatively. Goodwin Procter implemented an intensive 8-week simulation-based training for new associates (essentially "flight simulators" for legal work) since those associates won't get as much practice on routine tasks. These programs use mock transactions and lawsuits so "associates learn from the highest quality work rather than via volume" of rote tasks.
PwC took a different approach: in 2023, they began mandatory AI training for all 75,000 U.S. employees, structured in tiers from basic literacy to advanced technical skills. Interestingly, they limited the program to just 5 months because, as their Chief People Officer admitted, "the technology will continue to evolve" too quickly for a longer curriculum.
The End of the Grunt Work Apprenticeship
That system is now broken. As Bill Gates recently predicted, within 10 years "people won't be needed for most things" due to AI. This is in his words "profound and a little scary," especially for career development. If AI handles the entry-level work, where do newcomers learn their craft?
The answer seems to be a mix of:
- Simulation training: Like pilots who practice emergencies in flight simulators, young professionals will train in AI-free environments to build core skills, even if they use AI in real practice.
- AI literacy as a baseline: In 2025, proficiency with AI tools is as fundamental for new professionals as spreadsheet skills were in the past. Education is shifting to teach not just the domain but how to effectively prompt, validate, and improve AI outputs.
- Higher-order skills from day one: New hires are expected to bring more advanced critical thinking, client communication, and ethical judgment—skills that used to develop over years of practice. The "AI slide" pushes the baseline higher; what was mid-level work in 2017 might be entry-level work in 2027.
The New Diamond Organisation
So what does the professional services firm of the near future look like? Something more diamond-shaped:
- A narrower base: Fewer entry-level hires with higher expectations. New associates need to bring more valuable skills from day one since the easy stuff is handled by AI. Jamie Dimon of JPMorgan predicts AI could automate "60% to 70% of work" in the coming years. If true, firms simply won't need as many human juniors.
- A wider middle: More experienced professionals who focus on judgment, client relationships, and directing AI systems. Instead of managing junior humans, these professionals "manage machines" as Anthropic's Dario Amodei put it. They're experts at knowing when to trust AI outputs and when to override them.
- A still-narrow top: Partners and senior leaders who set strategy and maintain the most valuable client relationships.
This structure is decidedly flatter. Gartner predicts that by 2026, "20% of organizations will use AI to flatten their organizational structure, eliminating more than half of current middle management positions." With AI handling coordination, oversight, and analysis, fewer layers of management are needed.
One Person, One Diamond
The most radical vision of this transformation comes from Sam Altman of OpenAI, who predicted we'll soon see "1-person AI unicorns": billion-dollar companies run by a single human orchestrating multiple AI systems. Instead of a pyramid of humans with tech on the periphery, imagine a network of AI agents with a human at the center.
This isn't science fiction. We already have examples of solo attorneys using AI to run entire practices, or boutique consultancies where one principal plus AI does the work that once required a team of ten. The human provides strategy, quality control, and client relationships; AI handles execution at scale.
Think of it as a personal diamond: the human at the center, connected to a constellation of specialised AI systems. No partners, no associates, no hierarchy—just augmented expertise.
Not Everyone Agrees
To be fair, not all experts believe this transformation will be so dramatic. Yann LeCun of Meta thinks current AI has limitations, particularly in tasks requiring common sense or complex real-world interactions. And Dario Amodei notes that humans often engage in activities for the joy or creativity of it, even if we aren't the best at it; this implies that some work may remain human simply because we want it to be.
Goldman Sachs economists point out that historically, technological transformation creates new jobs and that the majority of long-term employment growth comes from roles that didn't previously exist. Perhaps the diamond is just a transitional structure, and some new arrangement will emerge as AI and humans find their equilibrium. It is clear though that we are moving away from the pyramidal structure.
The Pyramid Had a Good Run
Professional services firms face a moment of truth. But, the pyramid, with all its inefficiencies, created a training ground where raw talent could be shaped into expertise through repetition and feedback. The diamond might be more efficient, but it requires us to reinvent how professionals learn their craft.
For new graduates heading into these fields, the message is clear: the days of paying your dues through grunt work are ending. Instead, you'll need to arrive with higher-level skills and learn to partner with AI from day one. The pyramid that welcomed generations of professionals is crumbling, and in its place stands a diamond— an org chart that is sleek and shiny, and far less accommodating to beginners.
The Pyramid is dead. Long live the Diamond.