Saturday, June 27, 2026

I want to talk about the conversation that isn't happening.

Every conference I've attended in the last eighteen months has featured at least one session about AI return on investment — how to measure it, how to justify it to a board, how to frame it in terms that CFOs will approve. The slides all look similar. The case studies are carefully chosen. The ROI is always positive.

Nobody talks about the cost of not moving.

That's the more dangerous number, and it's the one that almost no business leader can clearly articulate. What does it actually cost — in real dollars, in competitive positioning, in organizational capability — to spend another year waiting for AI to "prove itself" before making a real commitment?

I've been thinking about this question for a long time, and I've come to a conclusion that most people in my position are reluctant to say out loud: the risk of inaction is now larger than the risk of getting it wrong. Not because AI is perfect — it isn't — but because the gap between teams that have integrated it deeply and teams that haven't is compounding faster than most organizations realize.

Here's what that actually looks like.

The Math Nobody Is Running

Let's start with something concrete.

Assume your growth team spends 30% of their time on work that could be partially or fully automated — reporting, data reconciliation, briefing documents, first-draft content, campaign analysis. For a five-person team at a combined fully-loaded cost of $600,000 per year, that's $180,000 worth of labor annually going toward mechanical tasks rather than creative or strategic ones.

Now assume you spend twelve months evaluating AI tools, running pilots, forming committees, and waiting for internal alignment before committing to a real implementation. That's $180,000 in labor cost you've already paid — and unlike a failed AI implementation, which teaches you something, twelve months of waiting teaches you nothing except that alignment is slow.

The analysis gets sharper when you factor in opportunity cost. While you were evaluating, your competitors who moved earlier had twelve months to build organizational intuition around AI. They have twelve months of learned workflows, prompt libraries, integration patterns, and team habits that you don't. That knowledge doesn't transfer when you eventually buy the tool. It accretes slowly through use. And now they're twelve months ahead of where that accrual starts for you.

This is not an argument for recklessness. It's an argument for honesty about what the status quo actually costs.

The Three Hidden Costs of Staying Manual

When teams decide to wait, they're usually thinking about the explicit risks — bad data, hallucinated outputs, implementation overhead, change management friction. These are real. They deserve serious consideration.

What doesn't get the same scrutiny is the other side of the ledger.

The first hidden cost is decision lag. Manual processes have cycle times. A campaign anomaly spotted on Thursday gets investigated on Monday, analyzed by Wednesday, and actioned by the following week — if it stays prioritized. That's two weeks of degraded performance baked into the operating model. In markets where consumer behavior shifts fast, two weeks is a meaningful window. Teams with AI-assisted monitoring close that cycle to hours. The accumulated difference across a year of campaigns is substantial, and it almost never shows up in a comparison of "AI vs. no AI" because the counterfactual is invisible.

The second hidden cost is talent friction. The best operators in every field are increasingly choosing companies where they work with modern tools. This isn't ideological — it's practical. Smart people don't want to spend their days on tasks they know software can handle. When your stack is visibly outdated, you feel it in recruiting conversations before you see it in retention data. A growth marketer who has spent two years working alongside AI systems at their last company and joins one where they're building pivot tables by hand is not going to stay.

The third hidden cost is organizational learning debt. Every month a team doesn't use AI, the gap between their fluency and the market's baseline fluency grows. When they eventually adopt — and they will — the ramp-up takes longer, costs more, and lands harder on a team that hasn't built the habits and intuitions that come from continuous use. The companies that adopted email in 1996 didn't just get email earlier. They got fifteen years of organizational knowledge about how to communicate electronically before their competitors had to catch up.

The Adoption Gap Is Wider Than You Think

Here's something that surprised me when I dug into the data: AI adoption in business contexts is far more bifurcated than headline usage statistics suggest.

Aggregate numbers show that a large percentage of companies are "using AI." What the aggregates obscure is the enormous variance in depth of use. There's a meaningful difference between a team where one analyst occasionally uses ChatGPT to summarize documents and a team that has restructured its entire workflow around AI-native processes. Both count as "AI adoption" in most surveys. They are not remotely comparable in impact.

The distribution looks roughly like this: a small percentage of companies — call it ten to fifteen percent — have made genuine structural changes to how they work. They've rebuilt workflows from first principles, hired for AI fluency, invested in integrations, and developed internal expertise. Another thirty to forty percent are in experimental mode — pilots, one-off use cases, individual adoption without organizational infrastructure. The rest are watching.

The gap between the first group and the third is not closing. It's widening. The first group is using their early advantage to move faster, which generates better results, which generates more budget, which enables deeper integration. It's a flywheel, and it started spinning months or years ago.

What "Getting It Wrong" Actually Looks Like

I want to address the other side of this honestly, because I don't think the answer is "move fast and don't worry about mistakes."

There are real ways to get AI adoption wrong. The most common one isn't the one people fear — it's not that the AI produces bad outputs and nobody catches it. In practice, teams that are paying attention catch most problems. The more common failure mode is adoption without integration: buying tools that don't connect to each other, asking teams to use them without changing underlying processes, and measuring success by activation rather than outcomes.

A company that buys five AI tools that don't talk to each other and adds them on top of an already-bloated stack is not going to see meaningful gains. They might actually make things worse — more context-switching, more tools to maintain, more outputs to reconcile. The tool isn't the problem. The lack of a coherent adoption strategy is.

The second common failure mode is measuring AI by individual task performance rather than system outcomes. If you evaluate an AI writing assistant by whether its first draft is good enough to publish without editing, you'll almost always be disappointed. If you evaluate it by how much faster your team produces good final content, you'll usually be impressed. The frame matters enormously.

Neither of these failure modes is an argument for waiting. They're arguments for being strategic about how you move, which is different.

The Eighteen-Month Window

I'll make a prediction that I'm confident enough in to put my name on: the next eighteen months will see a significant sorting of companies by AI capability, and that sorting will be sticky.

The teams that build real AI fluency in 2025 and 2026 will have compounding advantages in 2027 and 2028 that are very difficult to overcome through catch-up adoption alone. Not because AI tools will be inaccessible to latecomers — they won't be — but because the organizational knowledge, the workflow redesign, the team habits, and the institutional intuition about what works and what doesn't cannot be purchased. They can only be earned.

This has happened before. Companies that built genuine digital marketing capabilities in 2010 and 2011, when most of their competitors were still treating online advertising as a side channel, had structural advantages that persisted for years. Companies that built sophisticated data infrastructure in 2014 and 2015 could run experiments in 2018 that their competitors couldn't, not because the tools weren't available, but because the organizational muscle wasn't there.

The window is not closed. But it's not permanently open, either.

What the Best Teams Are Actually Doing

I spend a lot of time talking to high-performing growth and operations teams — both Nexus users and companies we've only met through research. The pattern among the ones I'd describe as genuinely ahead is consistent, and it's not what most people expect.

They're not the ones with the most tools. They usually have fewer tools than average, but more deeply integrated ones. They're not the ones who automated everything — they're extremely deliberate about which tasks benefit from automation and which ones still require human judgment. They're not the ones with the biggest AI budgets. Many of the most effective implementations I've seen were done with remarkable resource discipline.

What they share is a different mental model about where AI creates leverage. They don't ask "how do we use AI to do our current job faster?" They ask "if AI handles the mechanical work, what does that free us to do that we couldn't do before — and what does that mean for how we structure our team?"

That reframe changes everything about the implementation. You're not adding a tool. You're redesigning a role around a new capability, which is a much harder and more interesting problem.

The Only Question Worth Answering

I've watched a lot of leadership teams have a lot of conversations about AI strategy. The conversations I've seen lead to real progress share one characteristic: they start from an honest accounting of what the current approach is actually costing.

Not "is AI good enough to justify an investment?" That question puts the burden of proof in the wrong place. The status quo has costs too. It's just that those costs are familiar, so they feel normal.

The better question is: given what you know today about the direction this technology is moving, what is the cost of another year of observation rather than participation?

Answer that honestly and the strategy question tends to get a lot simpler.

We built Nexus because we believe the teams that move thoughtfully but quickly — that treat AI integration as a core strategic priority rather than a discretionary experiment — are the ones that will define what great looks like in their categories. We want to be part of how that happens for the companies that are ready to do the hard work of actually changing how they operate.

If that's you, we'd like to talk.