The AI gap isn't opening where most leaders are looking.

"Six months ago, my first morning hour was triage. Now it is twenty minutes of confirmations. An agent built into my file system has already read the overnight emails, drafted the responses, and flagged what I need to see. This is what AI starts to look like when it lives inside the work itself rather than in a window beside it."

My own morning has changed in six months.

 

Before I have finished my first coffee, an agent built into my file system has already read the overnight client emails, flagged the three that need me, drafted responses to two of them, and surfaced one decision that was about to be missed in a project a country away. Six months ago, that hour was triage. Now it is twenty minutes of confirmations.

This is not productivity software. It is not a chatbot doing a slightly fancier version of what an inbox already does. It is what AI starts to look like when it lives inside the work itself, rather than in a window beside it.

It is also what most businesses have not started doing yet.

Most businesses I talk to are wrestling with Wave 1 of AI adoption.

They are working out who is actually using AI. Trying to bring the unauthorised AI use under control. Updating policies and guardrails that were written before AI existed. Getting their team using the approved tools properly. That is the right priority, and it is harder than most leaders realise.

It is also not the only thing happening.

Wave 2 has already started. Most leaders have not noticed because they are heads-down on Wave 1.

Wave 1 is access and governance. The work of figuring out who is using AI in your business, getting unauthorised tools and personal accounts under control, updating policies that no longer match how people work, and helping your team use the approved tools well. It is foundational. It is necessary. For most businesses, it is the next twelve months of real work.

Wave 2 is AI built into the work itself. Not a separate tool. Something that lives inside the file system, watches the project folder, reads the meeting notes, drafts the report, flags the risk. The same person spends the same hour and ends up with sharper analysis, a better-supported proposal, a clearer board paper. Not faster. Better.

This is what file-based AI systems are making possible. Tools like Cowork, Claude Code, and the agent products being released into every productivity suite. The capability they unlock is fundamentally different from what a chat window can do. But the discipline required to make them deliver value is a different game from showing your team how to write a good prompt.

The mistake leaders are about to make is treating AI adoption as one sequential project. Get the governance done, get the tools rolled out, then think about Wave 2 later. The waves do not work that way. They overlap. The businesses pulling away right now are the ones doing both with discipline. Finishing Wave 1 properly while being deliberate about how they move into Wave 2.

The gap will compound. The Wave 2 movers will spend the next twelve months building file-based workflows that get better the more they run. The Wave 1-only businesses will still be writing acceptable-use policies while their competitors have already automated the meeting before it becomes one.

 

So what does handling both with discipline look like?

Wave 1, done properly reveals an honest map of where AI is actually being used today, including the tools running on personal accounts that nobody has logged. Policies that match the current technology, not the ones from before AI existed. A clear standard for which approved tools are used for which workflows. Capability built in the people doing the work, not just delivered in a one-off training session.

Wave 2, started deliberately: pick one workflow where AI is built into the work itself, not bolted on through a chat window. Something repeatable, file-based, where the value is clear if it works. Humans in the loop where judgement matters.

Then measure the right thing. Most businesses default to measuring AI by time saved. That is a productivity-era metric, and it is too narrow to capture what Wave 2 is actually doing. The real measure is quality and speed of decisions, and quality of outputs. Same time spent, sharper outcome. Same effort, better result. That measurement question deserves its own piece, which we will come back to.

Most leaders think they are behind on AI. They are behind on Wave 1. They have not noticed they are also behind on Wave 2, which is the one that compounds.

Catching up on Wave 1 is not the same as keeping up with Wave 2.

Know which wave you're in

Most leaders assume they know where their AI adoption is up to. In five minutes, find out where the real gaps are and where to focus first.

Take the 5-Minute AI Readiness Snapshot

--

About the Author 

Mark Lucas is Co-founder at Generation AI, where we help New Zealand business leaders build real AI capability.

Information provided is general in nature and does not constitute legal advice