David Monnerat

Product + AI | Systems Thinker | Enterprise Reality

Tag: trust

  • The Two Clocks

    The Two Clocks

    Two clocks are always running when a company adopts a new AI tool. Only one of them gets talked about.

    A company standardizes on Microsoft. Email, identity, security, document tools, all of it, built up over years. When Copilot arrives, it’s in front of employees within weeks. Not because anyone ran a bake-off and decided it was the best model. Because the contracts were already signed, the data handling agreements already existed, the security review had effectively already happened, years ago, for something else entirely.

    Meanwhile, a team wants Claude Code, because for the specific work they do, it’s just better. Faster iteration, fewer dead ends, less fighting the tool to get the output they need. It sits in a security review queue for months. Not because anyone doubts it works. Because someone has to answer, from scratch, how it touches production data, what gets logged, who’s liable if it doesn’t behave.

    Nobody’s hiding the reason for the gap. Everyone involved knows exactly why one tool showed up in weeks and the other is still waiting. That’s not the problem. The problem is what the gap does regardless of how honestly everyone understands it.

    Two Different Clocks

    Model capability improves on something close to a monthly cadence. Trust infrastructure, contracts, security review, data governance, moves on a quarterly-to-annual one, sometimes slower. Whoever’s clock finishes first gets the rollout, whether or not their tool is actually the best one by the time it ships.

    That’s not new. Incumbents have always had a head start built into existing relationships. What’s new is the size of the gap between the two clocks. Model capability used to move slowly enough that a year-long procurement cycle wasn’t catastrophic, the tool you evaluated was still roughly the tool you deployed. Now the fast clock moves so much faster than the slow one that by the time trust infrastructure catches up, the capability gap that justified the wait may have already closed. The incumbent had a year to get better while everyone else waited for paperwork.

    Employees feel this directly. Not as an abstract strategy problem, as daily friction: watching a tool everyone knows is better sit behind a queue, using something clunkier because it’s what’s actually available. That frustration isn’t noise to manage. It’s the two clocks, felt from the inside.

    What Actually Changes This

    Here’s the part worth doing something about. Right now, trust infrastructure gets built reactively: a team wants a tool, and only then does anyone start the work of figuring out how it can safely touch real systems and real data. That guarantees the gap every time, because the slow clock never starts running until the fast clock has already picked a winner.

    The fix isn’t picking better tools faster. It’s treating trust infrastructure itself as a capability an organization builds once and reuses, not a one-off project restarted from zero for every new vendor. Concretely: a data classification scheme that already says what “production data” means and what handling it requires, built independent of any specific tool. A security review checklist with the generic questions already answered, how do we log activity, how do we revoke access, what’s our incident process. A standing vendor risk framework that doesn’t get rebuilt from a blank page every time someone asks for a new tool. None of that requires knowing in advance which tool will win the next capability race. It requires deciding, ahead of time, how fast your organization can move once something does.

    That’s not the same as skipping the work or running both clocks blind. There’s a real reason security review happens after a tool is understood, you need to know what you’re evaluating before you can assess its risk, and racing that step against a capability evaluation that’s still moving is its own kind of danger, evaluating a moving target instead of a settled one. The fix isn’t rushing review. It’s making the parts of review that don’t depend on the specific tool ready in advance, so the parts that do can move as soon as the tool is understood, not months after.

    An organization that’s done this doesn’t win by picking the right model. It wins because when a leader has to explain the decision, they can say something true and simple: we didn’t choose Microsoft because it’s the best tool, we chose it because it was already trusted, and we’re now spending real effort making sure the next good tool doesn’t have to wait a year to earn the same trust.