Steering AI quarter by quarter: the Commodum cycle method

In January, your AI diagnostic said the collections team would free up 1.2 full-time equivalents by year-end. By June, the tool your team relies on had gained a feature that changes the calculus on three additional tasks, two people had transferred to another department, and case volume was up 15%. None of that shows up in the document you presented to the executive committee six months earlier. A diagnostic isn’t a decision set in stone — it’s a snapshot, and snapshots go stale. Here’s how to turn that snapshot into a steering cycle that stays accurate, quarter after quarter.

A frozen diagnostic goes stale faster than you’d think

Most leadership teams treat an AI diagnostic as a one-off audit: commission it, present it once to the executive committee, file it away. That’s a methodological error, not just a frequency problem. Two variables keep moving under the snapshot: the capability of available AI models, and the reality of your organization. Neither one waits for your next budget cycle to change.

No one steers cash flow off an eighteen-month-old balance sheet. No one plans headcount off an org chart that hasn’t been updated since the last leadership change. Yet that’s exactly what organizations do when they treat their AI diagnostic as a one-shot exercise: they make hiring, training and reorganization decisions against a map that has already stopped being true.

Three sources of drift between two snapshots

What actually changes between two diagnostics:

• Model capability — a task judged non-automatable six months ago can flip overnight with a new model version, a new integration, or a new specialized agent.
• Organizational structure — turnover, internal mobility, team mergers: every change in scope shifts the task mix and therefore the role’s exposure score.
• Activity volumes — a rise or fall in activity changes the freed-time calculation even when the nature of the tasks hasn’t changed at all.

A diagnostic stops accounting for these three sources of drift the day it’s delivered. The longer the gap before the next checkpoint, the wider the gap between the map and the ground — and the more decisions made in that interval rest on data no one realizes is already outdated.

The mechanics of the quarterly cycle: four steps, not an annual audit

The Commodum.ia subscription rests on a simple principle: the initial diagnostic isn’t a deliverable, it’s a starting point. Every quarter, the cycle runs again:

1. Volume refresh — headcount, volume and org-chart data are refreshed from your export, with no integration into your internal systems.
2. Rescoring — every task is checked again against the reference library, which itself evolves as model capability on the market advances.
3. Delta report — the current quarter is compared against the previous one: which roles changed category, which tasks entered the automation zone, which projected gains actually materialized.
4. Committee review — the deltas are presented using the same indicators as the initial diagnostic, so the committee reads a continuous story instead of sitting through a new study each time.

This mechanic changes the nature of the exercise: you’re no longer recommending a plan every eighteen months, you’re adjusting a trajectory every three.

A diagnostic is only as good as the next thing that changes — and in a living organization, something is always changing.

What the executive committee should actually look at each review

A quarterly review that just re-runs the original presentation adds nothing. What deserves the committee’s attention is the delta:

• Realized FTEs versus projected FTEs — has the freed-up time actually materialized, or is it still theoretical because redeployment never happened?
• Roles changing category — a role judged stable at N1 can shift into the strain zone if its scope has changed.
• Newly automatable tasks — this is the most actionable signal: it tells you exactly where to revisit a job description before the question comes up in a performance review.
• Redeployment progress — is the freed-up time actually funding the priorities you identified, or is it evaporating into day-to-day operations?

The quarterly rhythm isn’t arbitrary

Three months matches the natural cadence of most governance instances: budget reviews, extended executive committee meetings, HR progress checkpoints. Aligning the diagnostic cycle to that rhythm avoids creating yet another meeting — the diagnostic becomes an item on an existing agenda, not a parallel exercise.

That rhythm also serves slower deadlines. The triennial GEPP requires works-council (CSE) consultation whenever a transformation touches the organization of work, above 300 employees. Walking into that negotiation with twelve quarters of factual data changes the leadership team’s position entirely: it isn’t discovering the scale of the transformation at the negotiating table, it has been documenting it for three years.

The risk of under-steering — and the risk of over-steering

A cycle that’s too spread out — annual, or worse, multi-year — gives the three sources of drift time to pile up unchecked: leadership discovers the size of the gap only once it’s become expensive to fix. A cycle that’s too tight — monthly, or continuous — produces the opposite failure: statistical noise on minor variations, decision fatigue in committee, and the risk of reacting to meaningless fluctuations instead of real trends.

The quarter is the balance point: short enough to catch shifts in model capability and organizational movement before they compound, long enough that each review covers substantial deltas rather than noise.

Steering AI isn’t about producing one flawless diagnostic. It’s about accepting that no snapshot stays true for long — and building the instance that corrects it before it starts lying to you.