Clarity Through the Noise

I need to be honest with you about something.

I am exhausted.

Not the kind of exhausted you get from a bad quarter or a tough board meeting. This is a different fatigue — the kind that comes from feeling the full weight of a moment you know matters, while simultaneously trying to lead through it.

I spend my days building with AI tools that feel like they’ve compressed years of work into weeks. I run agentic systems that write code, research topics, and synthesize analysis at a pace that would have required a team of ten not long ago. The exhilaration is real. When an AI agent returns a structured research brief in four minutes that would have taken an analyst two days, you feel the future arriving.

And then you look up. Wars in Ukraine and the Middle East with no end in sight. Supply chains rerouting around conflict zones. An AWS data center taking drone strikes. The political order fraying in ways that make five-year planning feel like fiction. Eighty percent of AI projects still failing to deliver value. The EU AI Act clock ticking toward August.

The tools are exhilarating. The moment is exhausting. Both are true. That’s the signal.

I think a lot of leaders feel this way right now. And I think naming it is the first step toward doing something useful with it.

The Noise This Week

In the span of four days, three signals landed that tell you where the world is heading — if you can hear them through the static.

Signal 1: NVIDIA Says Every Company Is a Robotics Company

At GTC 2026 in San Jose, Jensen Huang delivered his most robotics-heavy keynote to date. Disney’s Olaf — the actual snowman character from Frozen — walked on stage as a humanoid robot. Ten humanoid robot companies demonstrated on the show floor. NVIDIA unveiled its full physical AI stack: new silicon (Jetson Thor delivering 2,070 TFLOPS), a robot foundation model (GR00T N1.7, now commercially available), a physics engine (Newton), and factory-scale digital twin simulations through Omniverse.

Huang’s declaration: “Physical AI has arrived. Every industrial company will become a robotics company.”

NVIDIA projects the addressable robotics market at $50 trillion. ABB, FANUC, KUKA, Universal Robots, and YASKAWA are building on NVIDIA’s platform. Foxconn demonstrated a fully simulated 242,000-square-foot factory in Houston. The strategy is deliberate: turn robotics’ data problem into a compute problem — and NVIDIA sells the compute.

Here’s the governance signal the keynote buried: NVIDIA announced safety frameworks for autonomous vehicles (Halos) and medical devices (IGX Thor). But for general-purpose humanoid robots in warehouses and factories? Nothing. No safety consortium. No standards body engagement. No accountability framework for when a humanoid robot in a shared workspace causes harm.

The technology is racing. The governance is walking.

Signal 2: Bezos Is Coming for Every Old Factory

The Wall Street Journal reported on March 19 that Jeff Bezos is raising a $100 billion fund to acquire manufacturing companies and rebuild them with AI. The vehicle is called Project Prometheus — co-founded with Vik Bajaj (formerly of Google’s Verily), staffed with 120 researchers poached from Meta, OpenAI, and DeepMind, and advised by two co-authors of the “Attention Is All You Need” transformer paper.

The pitch to sovereign wealth funds: buy legacy manufacturers running on 20th-century processes. Rebuild them with AI. Compress optimization cycles from years to weeks.

Bezos built the playbook at Amazon — over one million robots across 300 facilities, the world’s largest operator of mobile robotics. Project Prometheus is the R&D lab. The $100 billion fund is the weapon.

The thesis is dark factories — fully automated facilities that run without human workers, literally without lights, the way modern data centers do. China is the proof of concept: 290,000 industrial robots installed in 2024 alone. Xiaomi launched a facility producing one smartphone per second. Thirty million Chinese factory jobs disappeared while output climbed to record levels.

Gartner estimates 60% of manufacturers will adopt some form of lights-out manufacturing by 2026. The reshoring narrative in the U.S. masks a transformation: 74% of manufacturers are bringing operations back to North America, but they’re building automated facilities, not staffed ones. The factories are coming home. The jobs are not — at least not the same jobs.

For boards: the question isn’t whether your industry will be affected. It’s whether you’re the acquirer or the acquired. Analysts are already naming targets — companies with aging production infrastructure, quality issues, and delivery delays. The comparison isn’t to a competitor entering your market. It’s to Carnegie Steel. It’s to Ford’s assembly line.

Signal 3: The Builders Are Choosing to Build

Last week, David Der — a senior AI executive at Elevance Health, a Fortune 25 company — posted on LinkedIn that he’s stepping back from people leadership to go all-in as an individual contributor. His words:

“The next 12 to 24 months will have a disproportional impact on technology for the next decade. It’s not theoretical anymore. Agentic systems that are produced and optimized by agentic systems — this can be built today. I cannot imagine not helping to build it. People leadership will be there again for me when I’m ready to come back. This window won’t.”

This isn’t an isolated signal. The San Francisco Standard reported that Silicon Valley is replacing “software engineer” with “builder” as the defining role of the AI era. Boris Cherny, who created Anthropic’s Claude Code, said: “Today coding is practically solved. ‘Software engineer’ will go away.” LinkedIn launched a “full stack builder” program. Walmart is hiring “agent builders” — roles filled by both technical and non-technical employees. Meta product managers are self-identifying as “AI builders.”

The Wall Street Journal’s own team demonstrated the trend: two non-coding journalists used Claude Code to build an interactive web application for wsj.com, from idea to deployment. Sixty-three percent of active vibe coding users are non-developers — product managers and founders building full-stack apps with natural language alone.

The numbers back it up. Cursor — the AI coding tool — reached $100 million in annual recurring revenue with 20 employees. Lovable did it with 45. Pre-AI, the benchmark was roughly a thousand employees to hit that revenue. Meta’s applied AI engineering team now operates at a 50-to-1 employee-to-manager ratio — double what was previously considered the outer limit. Zuckerberg told investors: “Projects that used to require big teams can now be accomplished by a single, very talented person.”

Gartner predicts that by end of 2026, 20% of organizations will use AI to flatten their structures, eliminating more than half of current middle management positions. LinkedIn job postings with “manager” in the title declined 12% year-over-year in early 2026. “Lead” and “principal” roles grew 18%.

What David Der did is what I’m seeing everywhere: the tools have gotten good enough that the highest-impact thing an experienced leader can do right now isn’t managing a team — it’s building, directly, with AI as the force multiplier.

I understand that impulse deeply. I feel it myself.

The Exhausting Part

Here’s where the noise gets heavy.

Against all of this building energy, the scoreboard is brutal:

• 80% of AI projects fail — twice the rate of non-AI IT projects (RAND, 2025)
• 42% of companies abandoned most of their AI initiatives in 2025, up from 17% in 2024 (S&P Global)
• $547 billion of the $684 billion invested in AI in 2025 failed to deliver intended value (Pertama Partners)
• Only 5% of companies generate substantial value from AI at scale (BCG)
• Talent readiness is at 20% and declining year-over-year (Deloitte)

Yet inference costs have dropped 280-fold in two years. The cost barrier is gone. The tools are mature. The first movers — BCG’s “future-built” 5% — are capturing double the revenue growth and pulling away from everyone else. The gap is widening, not closing.

And the EU AI Act’s high-risk provisions hit in five months.

So we have: transformative capability arriving at unprecedented speed. A geopolitical backdrop that makes planning feel futile. An 80% project failure rate. A regulatory deadline. And the most energized, exhausted generation of builders the industry has ever produced.

That’s the noise.

The Leader’s Job

I’ve been thinking about what leadership means in a moment like this. Not the theoretical kind — the practical kind. The kind where you’re on a call with your board at 8 a.m. and building with AI agents at 10 p.m. and checking defense intelligence feeds somewhere in between.

The answer I keep coming back to: the leader’s job is to create clarity through the noise.

Not certainty. Clarity.

Certainty is a lie in a world where an AWS data center can take drone strikes and your AI vendor can be blacklisted by its own government in the same week. But clarity — enough directional confidence that your organization can make decisions and move — that’s achievable. And it’s the highest-leverage thing a leader can provide right now.

Here’s what clarity looks like in practice:

1. Name the moment for what it is. Not hype. Not panic. A genuine inflection point where physical AI (NVIDIA), capital concentration (Bezos), and individual leverage (the builder movement) are converging simultaneously. Organizations that treat this as another technology cycle will be wrong. Organizations that treat it as the end of the world will also be wrong. It’s a structural shift that demands structural preparation.

2. Audit your infrastructure, not your AI. The 80% failure rate doesn’t come from bad AI. It comes from bad data, unclear objectives, and organizations that bolt AI onto broken processes. McKinsey’s high performers are 3.6x more likely to fundamentally redesign workflows. Start there.

3. Build governance that enables speed. In Issue #1, I wrote that governance done right doesn’t ask “should we allow this?” It answers “how do we move on this quickly and safely?” That principle is more urgent now. With humanoid robots entering shared workspaces and no safety consortium in sight, the companies that build their own governance frameworks — ahead of regulation — will move faster, not slower.

4. Invest in your builders — but learn from Klarna. David Der chose the individual contributor (IC) path because the building opportunity is that significant. If your best people are making the same calculation, your org chart needs to catch up. Create IC tracks that rival management compensation. Restructure teams for AI leverage. But heed the cautionary tale: Klarna used AI to shrink from 5,000 to 3,000 employees and replaced 700 customer service agents with an AI chatbot. Then CEO Sebastian Siemiatkowski reversed course — “We went too far.” Quality dropped, customer complaints rose, and they started rehiring humans. Shopify’s Tobi Lutke struck a better balance: employees must prove AI can’t do a job before requesting new headcount. The goal isn’t fewer people. It’s more leverage per person.

5. Accept the exhaustion as part of the deal. This moment is tiring because it matters. Leaders who pretend otherwise lose credibility with the people who are living it. The honest conversation — “this is hard, and here’s how we’re navigating it” — builds more trust than any strategy deck.

What’s Happening This Week

This week, the rvatech Data + AI Summit — the 8th edition, and one of the premier events for data and AI practitioners in the Mid-Atlantic — takes place on March 26 at the Science Museum of Virginia in Richmond. As Board Chair of rvatech, I’m proud of what this community has built. If you’re in the region, I’d encourage you to attend — 600+ practitioners, keynotes, multiple tracks, and the kind of peer-to-peer networking that no AI tool can replace. Details and registration at rvatech.com.

I’ll unfortunately miss it. I’ll be at Microsoft that day, sitting with their product teams — talking about exactly the kinds of tools and platforms I’ve been describing in this newsletter. The builder in me couldn’t say no.

That tension — wanting to be in both rooms at once — is the moment we’re living in. There’s more to build, more to learn, and more to lead through than any single person can absorb. The only honest response is to pick the highest-impact thing you can do today, do it well, and trust that the clarity will come from the doing.

The Bottom Line

NVIDIA declared every company a robotics company. Bezos raised $100 billion to automate the ones that aren’t ready. Senior engineers are quitting management to build. And 80% of AI projects are still failing.

The noise is deafening. Your organization is looking to you to make sense of it.

You don’t need to have all the answers. You need to have enough clarity to take the next step — and enough honesty to say when you’re figuring it out alongside everyone else.

That’s not weakness. That’s leadership in a fog.

Where Do You Stand?

If this newsletter has you asking “how ready is my organization?” — I built a tool for that.

Quick Check — 5 questions, 2 minutes. A fast pulse on your board’s AI governance readiness.

Full AI Governance Assessment — 36 questions across 12 dimensions, 20 minutes. Produces a board-ready report with maturity scores, heatmaps, and priority recommendations.

Both are free.

Something I’m Building

Reading about governance decisions is one thing. Making them — with other people, under pressure, with incomplete information — is something else entirely.

I’ve been building a multiplayer governance simulation where you assemble a small board of colleagues, receive a scenario like the ones in this newsletter, debate it in a threaded discussion over the course of a week, and then vote — with a required rationale explaining your reasoning. Hidden information you can choose to uncover. Consequences from earlier decisions that come back weeks later. And at the end of each round, every player answers one question: whose reasoning most influenced your thinking?

It’s not a training exercise. It’s a way to practice the hardest part of governance — thinking clearly under uncertainty, with people who see the problem differently than you do.

I’m calling it Duty of Care: Boardroom Edition. It’s built on top of the solo AI governance game I launched earlier this year, and it uses the same scenario engine and AI-powered advisors — but the decisions are made together, and the real learning comes from the discussion, not the score.

If you’re the kind of person who reads this newsletter and thinks “I wish I could debate this with my board” — that’s exactly who this is for. More details soon.

---

Next issue: What I learned at Microsoft — a dispatch from inside the product teams building the tools that are reshaping how we work.

— Ian

AI CoE Lead, Altria | Board Chair, rvatech | iantyndall.com