Hi {{first_name | there}},

A few years ago, a market reset wiped $7 trillion from the S&P index. My company had state-of-the-art AI, billions in client valuations, and a seven-year track record of doubling revenue every single year. None of that protected us. The thing that kept us alive was speed. We made decisions in days that our competitors were still scheduling committee meetings to discuss. That experience rewired how I think about time in business, and it’s the reason I keep returning to a question I hear almost no one asking out loud: what if the problem isn’t your AI strategy, but the unit of time you’re using to measure it?

Annual planning is built on a reasonable assumption: that the competitive distance traveled in one year is roughly equal to the competitive distance traveled in the year before. For most of business history, that was true enough. Technology improved in linear, predictable steps. Industry leaders had years, sometimes decades, to absorb a disruption before it threatened their position. It isn’t anymore.

Every year now, AI becomes 10x cheaper on the compute side and 10x more capable on the model side. Both trends have held for five years straight. The researchers call them double exponentials, and what they mean in practice is that progress in AI is compounding in multiple dimensions at once, not just getting faster. We are now in the 100x era, and the organizations treating this like a standard technology cycle (plan in January, review in December) are not just moving slowly. They’re planning for a world that won’t exist by the time they arrive.

BCG’s most recent AI at Work survey (June 2026, 11,749 workers across 14 markets) makes this concrete: organizations with a clear AI strategy outperform those focused on tools alone by 25 percentage points in business outcomes. The leaders moved to twelve-week cycles. Everyone else is still waiting for December.

Here’s an analogy I keep coming back to. 150 years ago, 70 to 90 percent of people worked in agriculture. Today, that number is 1 percent. Mechanization didn’t just make farming more efficient. It reshaped civilization entirely. We are feeding more people with fewer resources, and the workforce that was freed in the process built almost everything we take for granted today. Nobody voted on the tractor. Seventy percent of the workforce just stopped farming.

AI is the tractor of our time, and it operates with the same indifference to readiness. A year ago, robotics company Figure had no product. Today they’re shipping fully functional humanoid robots on version two, with a valuation approaching that of General Motors in year three of existence. There are now hundreds of companies like them (most of which didn’t exist five years ago, many already at nine-figure valuations), each one compressing a timeline that previously took decades. This is the environment your annual benchmarks are being measured against. Not last year’s version of your industry. The one that’s being rebuilt underneath it right now.

Jensen Huang, the founder of Nvidia, doesn’t wear a watch. When asked why, his answer is simply: “Now is the most important time.” I think about that a lot. Not because it sounds good, but because he runs the most valuable company on the planet and his entire operating model is built around it. No destination. No finish line. Just the present moment, treated as the only unit that matters.

When I talk about 100x at Future Works, I mean two distinct things. First, it’s a measure of what’s now achievable in terms of speed. What used to take 100 days to build can take one day. What took 360,000 hours of attorney time at JP Morgan now happens in seconds. Approvals that locked up an aerospace client’s product launches for multiple quarters now clear in 48 hours. The scale of what’s possible has shifted so dramatically that the old benchmarks don’t just feel outdated. They actively mislead, because they anchor your sense of what’s ambitious to a world that no longer exists.

Second, and more importantly, 100x is a mindset shift, specifically the shift away from assuming that progress is linear. Organizations that scale AI across multiple functions are 60 percent more likely to report significant cost savings and 110 percent more likely to gain revenue benefits compared to those running isolated pilots. PwC’s 2026 AI Jobs Barometer, pulling from one billion job ads across 27 countries, found that the top 20 percent of AI-exposed companies are seeing 163 percent productivity growth against a 33.5 percent average. The distance between leaders and everyone else is already large, and it widens every quarter, because compounding doesn’t wait for the slower side to catch up.

I just updated 100x, my executive brief on AI transformation. The new edition covers how to connect your people, your software and data, and AI into a system that compounds, not just improves. Real case studies, battle-tested frameworks, and a clear path from 5% gains to 10x and beyond. Built for leaders who need results in weeks, not years.

The fastest way to fail at AI transformation is to approach it the same way you approached the last wave of digital transformation. For three decades, the standard playbook was to evaluate software, select a vendor, deploy in phases, manage the change, and measure adoption at the end of the year. Each step took months. The full cycle took years. And in a world where competitive advantages were sticky and technology moved at a human pace, that was fine. You had time to be methodical.

That playbook was already too slow in 2020. Today it just means you show up late. The organizations I watch thriving right now started before they were ready, found an internal leader who refused to stop, and built a rhythm of twelve-week cycles where each one delivers something measurable and funds the next. Bigger budgets and fancier AI teams had nothing to do with it. Cadence did.

NextPower is a useful example. The first twelve-week cycle delivered a platform that put logistics data on one operating surface and cut 20,000 work hours in year one. Six months from first conversation to live product. The compounding started from there.

1. Measure against what’s now possible, not what you did last year.

Some of your competitors are already three cycles deep into gains you haven’t started capturing. They stopped measuring against last year the moment the math stopped working in their favor. If your ambition is still anchored to 2025 numbers, you’re optimizing for a race that already moved.

2. Find the leader who will not stop.

Across all the research in my book, 100x, one variable predicts whether an AI transformation delivers real results more than any other: an executive sponsor who treats transformation as permanent. Budget matters less than you think. Technology stack matters less. The consulting firm definitely matters less. Every initiative I’ve seen with zero ROI had one thing in common: someone stopped it. The hardest part is building the conviction to keep going when early results are ambiguous and the pressure to return to normal gets loud.

3. Run twelve-week cycles, not twelve-month roadmaps.

A Bain survey of 951 large companies published in June 2026 found that only 4 percent achieved AI-related savings above 30 percent. The ones who got there didn’t plan longer or more carefully than everyone else. They started faster, tied every deployment to a CFO-trackable metric before scaling, and proved value in contained cycles rather than waiting for a year-end review to find out if anything had worked. If your current AI transformation plan has a two-year horizon, the most important thing you can do is ask what the first twelve-week cycle delivers. Run it this quarter.

Sam Altman put the current moment plainly: “Near the singularity; unclear which side.” Whether you believe we’re approaching it or already past it, the implication for how you run your organization is the same. This doesn’t pause for your planning cycle to close. The organizations already three cycles deep into measurable results figured that out early. Most of them are still running.

The executives I’ve seen win in this era started before they felt ready. They used the people already in the building, picked something they could ship in weeks, and kept going. Most of them told me later they almost didn’t. I’m glad they did.

Much Love,
Matt

At Lighthouse, we love featuring fresh perspectives from our community of AI, tech, and innovation leaders. Got insights to share? Just reply to this email—I’d love to hear from you!