A story about unimaginative visions for efficiency and who gets to decide what intelligence serves.
The machines are learning. But not what you think.
The machines are learning how to hollow out human work and pour the profits into distant accounts. The machines are learning the language of efficiency, which means “we don’t need you anymore.” The machines are learning to predict the predictions of equity analysts in their training data.
Contribution Design is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
This isn’t a story about “artificial intelligence.” This is a story about artificial scarcity, manufactured desperation, and the very real people deciding who our robot servants will serve.
But note it’s not machines, but humans, who are defining the value of work. This means alternative realities are possible. Not to claim that you must master AI or be replaced. This is a claim that we can create other ways of organizing work, and care, and renegotiating what we value.
Here are a few moments showing we’re at peak hype cycle, when customers are no longer defining value or demanding technology, but capital holders are accelerating AI conversions.
Because the Accountants are Resistant
Accountants working in practices are not as excited by LLM-powered finance tools. They are deterministic folks, and don’t give much credence to the probabilistic promises of this batch of agents and workflows. They had a hard enough time connecting their ERP system to the cloud, and they want a rest. They understand compounding and can foresee how hallucinating agents can lead to compounding errors and losses. They worry about security and cash flow and are naysayers when it comes to adopting edge-case LLM systems.
To counter this resistance, investors are accelerating adoption along with roll-up plays. Private equity firms are systematically acquiring traditional accounting firms with explicit AI transformation mandates. Baker Tilly, the 10th largest US accounting firm, received a $1 billion private equity investment from Hellman & Friedman and Valeas Capital Partners in February 2024, the largest PE investment in the CPA sector to date. The stated purpose? “Investments in talent, technology, and further strategic acquisitions.” 1It’s not in the press release, but it’s easier to push AI-driven transformation of a century-old profession when it’s tied to the remaining employees’ earnouts.
This isn’t isolated. PE money “flooded the accounting M&A market” in 2024, totaling $2.3 billion in deals. Note that this is different than the roll-up strategies tried in prior market turns. It’s a wholesale reimagining of how professional services operate.
The Pattern: Capital identifies industries with high labor costs and standardizable processes. Capital holders acquire market leaders and include AI implementation to reduce headcount and increase margins. It exits at higher multiples based on “AI-optimized” operations.
Vista Equity Partners has perfected this model across a range of investments in traditional web 2.0 technology, requiring each portfolio company to submit quantified AI benefits as part of operational planning. The results: 80% of Vista’s portfolio companies now deploy AI tools, with some seeing 30% increases in coding productivity.2
The machines learn to add and subtract. But what they calculate isn’t efficiency, it’s elimination.
Because I Need Less Heads
Public markets increasingly reward companies that frame workforce reductions as “AI efficiency gains,” creating perverse incentives for AI-driven displacement.
Marc Benioff stood before cameras and spoke the language of progress. “I’ve reduced it from 9,000 heads to about 5,000… Because I need less heads.” IBM explicitly replaced 200 HR employees with AI chatbots. Nearly 150,000 tech workers were laid off in 2024, with many cuts masked under terms like “restructuring” and “business optimization” to avoid “AI backlash” while advancing automation.3
These layoffs aren’t driven by financial distress. Microsoft cut 15,000 roles while reporting $70.1 billion in Q1 2025 revenue, a 13% increase. The layoffs align suspiciously well with the rollout of large AI systems occurring during strong earnings periods not financial struggles.
The Pattern: Companies discover that framing layoffs as “AI transformation” or “operational efficiency” generates positive market reactions. This creates a feedback loop where AI deployment becomes justified not by operational necessity but by market signaling requirements.
Frame human displacement as “AI advancement,” and watch your valuation soar. The machines weren’t just learning to do human work; they were learning to be the excuse for human abandonment. The machines learned the language of euphemism, that elimination could be called evolution.
Because You Can Go It Alone (with Machines for Co-Founders)
Investors are actively promoting the narrative of “AI-powered solo founders” who can build billion-dollar companies alone, fundamentally reshaping entrepreneurship expectations. Anthropic CEO Dario Amodei predicted we’d see “the first one-employee billion-dollar company” by 2026.4 OpenAI’s Sam Altman runs a “little group chat” of tech CEOs placing bets on when this will happen.
The numbers support this narrative shift: 35% of US startups incorporated in 2024 had a single founder, more than double the 17% in 2017. Solo founder startups climbed from 22.2% in 2015 to 35% in 2024.5
The reality was more complex. The machines had made it easier to build alone, but the money still flowed to familiar patterns, familiar faces, familiar zip codes.
Still, the mythology grew. Stories spread of individuals building empires with nothing but a laptop and an algorithm. The subtext was clear: if one person could do it all, why did anyone need teams? Why did anyone need colleagues? Why did anyone need… anyone
Midjourney achieved $200 million ARR (annual recurring revenue) with 11 employees and no formal sales team. Cursor reached $100 million ARR in under a year with just 20 engineers. These become proof-of-concept for the “AI agent as co-founder” thesis.
The Pattern: Capital holders bet that AI tools can replace human collaboration in startup formation. This isn’t just an investment thesis, it’s social engineering, reshaping how we think about company building and team formation.
It was once a no-go to be a solo founder if you wanted funding. The machines learned that together, with a single human, there’s an opportunity to market independence.
The Algorithm of Extraction
Step back and see the pattern. This isn’t about artificial intelligence becoming more capable. This is about capital holders investing with a herd-like mentality, using this version of LLMs as a tool to reshape society according to shared logic.
The sequence has not been creative nor inventive:
· Identify inefficiency (read: human labor)
- Deploy capital to devalue it (read: buy companies, demand AI implementation)
- Celebrate the efficiency gains (read: profit from human displacement)
- Use success stories to justify the next round (read: normalize the play)
The machines aren’t making these decisions. Humans are. Humans with spreadsheets, investment theses, and profit targets. Humans who’ve convinced themselves that optimization is inevitable.
But optimization is always a choice about values. And the values embedded in our training data were set long ago: efficiency over empathy, profit over people, extraction over creation.
The accounting firms aren’t being bought to serve clients better. They’re being bought to serve them with fewer humans. The layoffs aren’t happening because the work disappeared. They’re happening because the profits from that work can now flow to fewer hands. The solo founder mythology isn’t about empowering individuals. It’s about normalizing isolation, making human collaboration seem inefficient, unnecessary, and outdated.
The machines are learning that human labor is a cost to be minimized, not a resource to be valued. They’re learning that efficiency means elimination, not enhancement. They’re learning that intelligence is about replacement, not collaboration.
They’re learning to dream the dreams that the lemmings dream: worlds where value flows upward, where human work becomes obsolete, where intelligence serves extraction.
This is a choice
But here’s what they’re not learning: how to value care work, community building, the irreplaceable complexity of embodied wisdom, and our relationships to the other beings in our ecosystems. How to measure what can’t be optimized, quantify what shouldn’t be commodified, automate what must remain human.
This isn’t technological inevitability. This is a choice. Herd mentality investor decisions, made by people with their hands on capital decisions, about what our tools, technologies, and training data should serve.
We could create systems that distribute value instead of concentrating it. We could develop intelligence that serves community flourishing instead of capital extraction. We could deploy capital in a way that follows different values. We can design systems that are not as dependent on traditional flows of mono-capital.
But that would require admitting that efficiency isn’t the only value worth optimizing for. That humans have worth beyond their productivity. That intelligence, artificial or otherwise, should contribute to the life-carrying capacities of the ecosystems we are embedded within.
The machines will learn whatever we teach the machines. Right now, we’re teaching the machines that humans are inefficient, that care is unprofitable, that extraction is innovation.
We could teach the machines something else. But first, we’d have to believe that something else is possible, to decide who gets to shape what intelligence serves.
Right now, that decision is being made in boardrooms and investment committees, by people optimizing for speculative flips rather than human flourishing. Unimaginative capital holders are deciding. But it doesn’t have to stay that way.
The machines are learning. We can still have our own dreams.
If we remember that we have the power to choose what they serve.
Baker Tilly Secures Strategic Investment Led by Hellman. Baker Tilly. February 5, 2024. https://www.bakertilly.com/news/baker-tilly-secures-strategic-investment-led-by-hellman
Field Notes from Generative AI Insurgency Global Private Equity Report 2025, Bain https://www.bain.com/insights/field-notes-from-generative-ai-insurgency-global-private-equity-report-2025/
Burleigh, E. Fortune. Salesforce CEO Marc Benioff says his company has cut 4,000 customer service jobs as AI steps in: ‘I need less heads’. Yahoo News. September 2, 2025 https://finance.yahoo.com/news/salesforce-ceo-marc-benioff-says-145324020.html?guccounter=1
Ortiz S, First 1B Business with One Human Employee Will Happen in 2026, Sayes Anthropic CEO. ZDNet. May 22, 2025. https://www.zdnet.com/article/first-1b-business-with-one-human-employee-will-happen-in-2026-says-anthropic-ceo/