Mark Mondoka

You did the hard part. You saw the wave early, you moved real money toward it, and you didn’t wait for a committee to bless the decision. Your best people now have AI agents, and they are producing more than they ever have.

But, the company is not moving faster.

It’s not because you are early, need more training or more time for adoption to spread across the organization. AI adoption arrived in two waves, and most firms are still writing checks for the first.

The first wave was technical. Procuring tokens. Wiring up APIs. Getting models into the workflow. This was real work, and doing it early was the right call. But there’s no lasting edge in it: the models come from a handful of shared vendors, and any competitor with a budget can wire up the same APIs in a quarter. Being first to buy what everyone can buy is not a moat.

The second wave is organizational absorption: whether the company around the worker can take in what that worker now produces. For most early adopting organizations this is where the constraint has moved. The tools got faster at making things. Nothing got faster at checking, approving, and merging them.

We now have the 10x worker, someone whose agentic workflow lets them draft, code, analyze, and ship at a scale that used to require a team. But that worker sits inside a 1x organization: the same review steps, approval chains, and coordination rituals built for human-paced output.

Uncontrolled growth in a host that has not adaptively changed can be problematic. A Bugatti stuck in 25mph speed limit zones. The complication is that defending the old structure feels like the responsible choice. Those sign-off chains and review boards have protected high-margin lines for years. They caught real mistakes. Dismantling them looks reckless.

In a market where speed is the currency, though, that caution costs more than it saves. Speed at the individual level does not sit quietly beside a slow organization. When you accelerate execution without re-engineering the engine around it, the pressure compounds. Ten times the output still must be reviewed, aligned, and integrated, ten times over, by a management layer that hasn’t changed. Every gain in creation becomes a backlog somewhere downstream.

Before you diagnose your own org

The symptom is easy to name: your most productive people are exhausted, and the company still moves at last year’s pace. The mistake is to assume that pattern can only mean one thing.

It can mean at least three others, and each points somewhere different:

• The output isn’t real. If the “10x” is counted in drafts produced rather than work shipped and accepted, the gain may be an illusion. Fix the measurement before you touch the process.
• The output isn’t good. If agentic work keeps bouncing back for rework, the review layer catching it isn’t your obstacle; it’s earning its keep. What you need is higher quality, not looser coordination.
• Nobody’s waiting for it. If faster output would just pile up unsold, the ceiling is demand. Look to sales and the market, not the org chart.

So run the test before you name the cause. Three questions sort out which world you’re in: Are you counting shipped work or produced work? Does the output keep coming back for rework? Is there real demand waiting downstream?

If the gains are real, the quality holds, and the demand exists, and your throughput is still flat, then, and only then, the bottleneck is the machine around the worker. That’s the one case where more of the first wave won’t help. You can’t fix a number that’s gated downstream by speeding up the stage that was never slow.