When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

Start flowing free

Sunday, May 24, 2026 Intelligence for Portfolio Executives closing the AI Wage Gap.

———————————————————————————

Last week, the Bureau of Labor Statistics released something that almost nobody in the executive AI conversation has been waiting for — because most of them didn't know it existed.

It's called the AI Occupational Exposure Study. Eighteen occupations. Ten million workers tracked. Employment data from May 2024 through May 2025.

Here is what it found.

Across the economy, employment grew 0.8% over that period. In the eighteen occupations the BLS classified as AI-exposed, employment fell 0.2%. Strip out healthcare — which has its own structural labor dynamics that distort the comparison — and the remaining seventeen AI-exposed fields fell 1.6%. This is the second consecutive year that pattern has held.

For the first time in the history of this technology cycle, the United States federal government has published hard data confirming that AI is rearranging hiring. Not forecasting it. Not modeling it. Confirming it — with the same methodology, the same occupational classification system, and the same quarterly surveys that have tracked American employment since 1940.

This is not Dario Amodei's projections. This is not Mustafa Suleyman's 18-month claim in the Financial Times. This is not McKinsey's scenario analysis or Goldman Sachs's headline numbers. This is the BLS. It is the most conservative, most methodologically cautious labor statistics apparatus in the world. And it is now telling you, in the language it uses to describe real things that are actually happening, that AI-exposed roles are contracting while the broader economy adds jobs.

The AI Wage Gap is no longer a thesis. It is a federal statistic.

———————————————————————————

Before I give you the three moves, I want to be precise about what the BLS data shows and what it doesn't. Because this is the kind of number that invites the wrong conclusions as easily as the right ones.

The BLS is not measuring job losses caused by AI. It is measuring employment change in occupations classified as having high AI exposure — meaning occupations where a meaningful share of tasks can be performed or augmented by current AI systems. The 1.6% contraction in those fields could reflect several things: companies not backfilling roles when AI can absorb the function, a slowdown in junior hiring as AI handles entry-level output, or outright displacement of existing positions. The data does not distinguish between these mechanisms. It only shows the result.

What makes the second-year pattern significant is not its size — 1.6% is not a catastrophe in any single reporting period. What makes it significant is its direction and its consistency. A one-year anomaly can be explained away. A two-year trend in the same direction, against a backdrop of strong overall employment growth, is a structural signal. The economy is adding jobs. The AI-exposed occupations are not keeping pace. That gap has a name.

The other thing the BLS study does not tell you is which side of the 1.6% you are on. Not everyone in an AI-exposed occupation is losing ground. Some of them — the ones who learned to use the tools, who built workflows, who shipped artifacts — are doing more work and producing more output than they were two years ago. They are, by every economic measure, more valuable. The contraction in headcount is not distributed evenly across everyone in those fields. It is distributed to the people the organization can afford to lose because someone else has made themselves indispensable.

That distinction is the entire operating theory of this newsletter.

———————————————————————————

The IBM CEO study from May 4 has been extensively covered — the 76% of companies with a Chief AI Officer, the 86%/25% adoption gap I wrote about last week. But one number in that study received almost no attention in the coverage I saw, and it is the one that most directly names the career move of this moment.

Fifty-seven percent of Chief AI Officers at major companies were promoted from inside.

Not recruited from outside. Not brought in from a competing organization or a consulting firm or a lab. They were already inside the building. They were already inside the org. They earned the title by doing the work that the title eventually described.

This is not, on reflection, surprising. The CAIO role is new enough that there is no deep external talent pool for it. The skills it requires — AI fluency combined with organizational context and the ability to drive behavior change at scale — are not things you can hire in fully formed. They develop inside specific organizations, in people who understand that organization's data, its decision-making culture, its resistance patterns, its power structure.

The person who built the internal proof case. The person who designed the adoption program that moved the needle on the 86%/25% gap in their own division. The person who shipped a working AI tool inside the organization and documented what it changed. That is the person who becomes the CAIO. And 57% of the time, they were already on the payroll.

The career play is not to position yourself as an AI expert in the abstract. It is to be the person who closes the gap inside the organization you are already in — and to build the evidence that you did.

———————————————————————————

On Tuesday, May 20, OpenAI confidentially filed with the SEC for an IPO targeting a valuation of up to one trillion dollars. Goldman Sachs and Morgan Stanley are running the books. The listing window is September or October 2026. The target raise is $60 billion. OpenAI's annualized revenue is approximately $25 billion today; their internal target for 2030 is $280 billion.

I am not telling you this because it is a stock pick. I am telling you this because of what it signals for everyone who is still in the "evaluating AI" posture.

When a technology is in private markets, it is possible to watch it develop without feeling the direct weight of the comparison. Venture portfolios are opaque. Private valuations are theoretical. The competitive pressure is real but diffuse.

When that same technology becomes a public market asset — when it has a ticker symbol and a quarterly earnings report and an analyst consensus — the comparison becomes structural. Every executive's AI posture becomes visible against a public benchmark. Every board conversation about AI adoption now happens in the shadow of a company whose revenue is growing from $25 billion to $280 billion in four years. Every strategic plan that treats AI as a 2027 or 2028 priority is now directly comparable to an organization that, when the S-1 drops, will show its investors a revenue curve that makes the 2027 posture look like a category error.

Sequoia said it plainly this week: the next trillion-dollar company won't sell software. It will sell the work itself.

If you are a Portfolio Executive and you are not building toward the work layer — not just using AI tools but building AI-native workflows that produce deliverables, that compound, that replace not tasks but entire categories of coordination overhead — you are not in the wrong direction. You are in the wrong decade.

The $1 trillion clock is not a metaphor. It is a filing date.

———————————————————————————

One. Look up your occupation in the BLS AI Exposure Study. The study is public. The eighteen classified occupations include management analysts, financial analysts, software developers, legal occupations, writers and editors, and others. If your occupation or your direct reports' occupations appear in that list, you are in the cohort where the 1.6% contraction is happening. Know which bucket you are in. Do not outsource this diagnostic to a headline.

Two. Design your 57% case. The path to the CAIO title — or whatever convergence title your organization will eventually use — runs through internal proof, not external credentials. What is the workflow you built? What is the adoption program you designed? What is the before-and-after metric you can point to? The 57% did not get promoted because they read the right reports. They got promoted because they produced the evidence that something changed when they were in the room. Start building that evidence now, deliberately, with the intent to be able to name what you did and what it produced.

Three. Treat the IPO window as a planning deadline. OpenAI's listing is expected in September or October 2026. That is approximately five months. That is one cohort cycle. By the time the S-1 is public and every board in the country is reading OpenAI's revenue trajectory and asking their executives what they are doing about AI, you want to have a working AI artifact, a documented proof case, and a clear answer to the question of which side of the gap you are on. Five months is enough time. It is not unlimited time.

———————————————————————————

The BLS data will not stay buried for long. By the time you are reading this, it will have moved further up the media cycle. There will be op-eds and Senate hearings and corporate communications reframing what it means.

What will not be covered, because it is not a headline, is this: the people inside those eighteen occupations who have been building AI-native workflows for the last two years are not in the 1.6%. They are the reason someone else is. They made themselves indispensable by producing more, compounding faster, and closing more of the gap between what their organization knew was possible with AI and what it was actually doing.

The federal government has now confirmed that the gap exists. It has been tracking it for two years.

The only question left is which side of it you are on — and whether you are moving toward the 57% or waiting for someone else to.

———————————————————————————

Closing the AI Wage Gap opens with the displacement thesis and was written precisely for the moment the BLS data describes — not because I knew the study would drop this week, but because the structural dynamics it measures have been visible in the qualitative data for two years.

The chapters most directly activated by this week's signal:

The manuscript is in Tier-1 agent querying. Levine, Halpern, Sagalyn pending. If you know an editor or agent who works at the intersection of work, AI, and organizational economics — reply to this email.

———————————————————————————

A curated preview from the research I track weekly — tools, models, trends, and the infrastructure moves that will matter to your work before most people know they exist. Want the full daily version? See below.

🌐 Google I/O went agentic. Google shipped Gemini Omni (text prompts → video creation, character-consistent identity across scenes), Gemini 3.5 Flash (4x faster than its predecessor, now beating flagship models on coding and agentic benchmarks), and Antigravity 2.0 — an agent-first development platform. The headline: Google is no longer chasing OpenAI on chat. It is building a lane where the AI interface is the operating system.

🧠 Andrej Karpathy joined Anthropic. OpenAI co-founder, Tesla's former AI director, creator of nanoGPT and micrograd — arguably the most respected AI educator alive — moved to Anthropic this week. Anthropic simultaneously overtook OpenAI in business adoption for the first time: 34.4% vs. 32.3%. These two data points are related.

🔋 Compute is the new constraint. SpaceX's Colossus cluster (220,000+ Nvidia GPUs) is becoming shared AI infrastructure. Anthropic just doubled context limits on all paid plans and removed peak-hour caps — powered by the Colossus deal. API Tier 1 gets 1,500% more input tokens per minute. The frontier labs are pulling away from everyone else at the infrastructure layer. That gap compounds.

🧩 Agent memory hit 11,600 GitHub stars in days. agentmemory is an open-source, SQLite-only persistent memory layer for AI coding agents — 92% fewer tokens per session, 95.2% retrieval accuracy, works with Claude Code, Codex, Cursor, and Gemini CLI. The problem it solves is the problem every operator hits after week two: every session resets. The pattern of building permanent memory for AI agents is the infrastructure shift of this quarter.

📋 GitHub Spec Kit crossed 97,000 stars. The Spec-Driven Development toolkit — Spec → Plan → Tasks → Implement — is becoming the standard methodology for AI-assisted building. Thirty-plus integrations including Claude, Copilot, Codex, Gemini, and Windsurf. The insight it encodes: AI agents need specs more than they need code. Write what to build, not how to build it.

🖱️ The chat window is no longer the default AI interface. Google's Magic Pointer lets Gemini understand what your cursor is pointing at — no copy-paste, no prompt box. Just point and say "fix this" or "summarize that." The interface layer is changing. The professionals who adapt to input-mode shifts first have always been the ones who compound fastest.

That's six signals from a week that had forty. I track all of them — tools, agents, model releases, open-source repos, workflow playbooks, LinkedIn operators, newsletter digests — across sources most executives don't have time to monitor. Starting soon, I'm releasing a daily version of this research: The Leverage Signal.

———————————————————————————

A new daily briefing — and I want to know if you'd read it.

The research catalog that feeds this newsletter — the one I build every week from AlphaSignal, Rundown AI, TAAFT, AI Fire, and a rotating set of the highest signal-to-noise operators on LinkedIn — runs to forty to sixty items a week. The weekly Leverage Brief can only hold six of them.

I'm considering launching The Leverage Signal: a short daily briefing (five minutes, every weekday morning) built specifically for the people reading this newsletter right now.

Who it's for — specifically:

Portfolio Executives and senior knowledge workers in AI-exposed occupations: CHROs, CTOs, CAIOs, CLOs, and fractional executives building toward convergence roles. Founders and operators running AI-leveraged businesses with small teams. Executive coaches, organizational designers, and L&D leaders whose clients are navigating AI transformation. Anyone in a knowledge-work career who has moved past "I've tried some tools" and wants the daily intelligence feed of someone who is building seriously — not a product newsletter, not a hype digest, but a working operator's read on what moved, what shipped, and what it means for the work you do before 9 AM.

What it covers, daily:

The two or three most signal-rich AI tool or agent releases from the previous 24 hours — with a one-sentence "what this means for you" framing, not just a description. One open-source repo worth knowing. One workflow or deployment pattern from the operator community. One macro or industry signal (model release, funding, regulatory, talent move) that belongs in your working context. No hype. No recaps of things you already saw. Just the items that, if you missed them, would show up in someone else's strategy deck before they show up in yours.

The model I'm building toward:

With enough subscribers, The Leverage Signal will be auto-researched, auto-drafted from the same Notion catalog that feeds this newsletter, and sent daily — so the pipeline from raw signal to your inbox is as tight as possible. I'm not there yet. I want to know if the readership is there first.

→ SIGN UP HERE — Opt in if you want The Leverage Signal. Takes ten seconds. Tells me whether to build it.

———————————————————————————

Three platforms paying domain experts to train the next generation of AI models. Updated this edition.

Mercor — Now valued at $10B (October 2025 Series C, $350M led by Felicis). Roughly 30,000+ contractors hired in 2025 alone for projects at OpenAI, Anthropic, and other frontier labs. Average expert rate ~$85/hour; engineering work $70–$200+/hr. Note: Mercor was impacted by a March 2026 supply-chain attack involving the LiteLLM package, potentially exposing contractor PII — review their post-breach disclosures before onboarding. → Apply: mercor.com

micro1 — Crossed $100M ARR in December 2025, up from $7M at the start of that year. $35M Series A at a $500M valuation in September 2025, led by 01 Advisors. Selective AI-powered vetting through "Zara" certifies approved professionals. Trainer/annotator $20–$40/hr, evaluator $20–$65/hr, engineering $50–$150/hr. Direct competitor to Mercor and Scale. → Apply: talent.micro1.ai

Meridial (by Invisible Technologies) — Invisible reached $134M revenue in 2024 and raised $100M in September 2025. Meridial requires a degree and specialized knowledge in law, STEM, finance, etc., with 400+ projects open at last count. Hires from US, Canada, UK, Ireland, NZ, Australia. Strong fit if your domain is specialized rather than generalist. → Apply: meridial.com

One platform is a job. Two is a hedge. Three is a portfolio.

———————————————————————————

The BLS data landed this week. The June cohort starts in three weeks.

That timing is not coincidental — it is the exact environment the cohort was designed for: the moment when the gap stops being theoretical and becomes a federal statistic, and the people who have been waiting for confirmation have it, and the people who have been building proof cases have an advantage.

Twelve weeks. Fifteen seats. Three deliverables:

Five months before the OpenAI IPO. Three weeks before the cohort starts.

→ Apply: portlev.com/cohort

———————————————————————————

Go to the BLS AI Occupational Exposure Study. It is public and it is searchable.

Find your occupation. Find two or three of your direct reports' occupations. Find the occupation of the person most likely to be competing with you for the next role up.

Write down which of those are in the AI-exposed classification and which are not.

Then ask yourself one question: if the 1.6% contraction continued at the same rate for three more years — which it won't, it will almost certainly accelerate — what would that mean for the headcount in your division? In your function? In your career track?

You don't need to answer the question perfectly. You need to stop treating it as someone else's question.

The BLS just made it yours.

———————————————————————————

Reply to this email. I read every one.

Yuri Kruman Founder, Portfolio Leverage Co. · 3x CHRO · AI Trainer: OpenAI, Meta, Microsoft PortLev.com · LinkedIn · AIWageGap.com

———————————————————————————

"The Bureau of Labor Statistics does not speculate. It measures. It has now measured two consecutive years of employment contraction in AI-exposed occupations against a backdrop of broad economic growth. The people who are surprised by this number were not paying attention. The people who are acting on it already were."

— The Leverage Brief, May 2026

———————————————————————————

© 2026 Portfolio Leverage Company. You're receiving this because you subscribed to The Leverage Brief. Unsubscribe in one click.