The 30/60/90 Operating Cadence for Digital Workers

The review cadence that turns a launched Digital Worker into a compounding asset instead of a stalling pilot — what to measure, when, and who owns it.

Why most pilots die between launch and month four

We've sat in too many post-mortems for Digital Worker pilots that launched cleanly, hit their initial KPIs, then quietly faded. The pattern is consistent: nobody owns the workflow after the consultancy hands off, the operating cadence is informal, and by month four the Worker is either ignored, sidestepped, or producing outputs no one trusts.

The fix is not technical. The Worker still works. What broke is the human operating model around it. The 30/60/90 cadence below is the most reliable structure we have found for keeping a Digital Worker compounding rather than degrading.

Day 30 — Operating Review

Thirty days after go-live, the named operator (a real human, not "the AI team") presents a one-page review to a small cross-functional group: the business owner, finance, and one representative from the team whose work the Worker is augmenting.

• What did it actually do? Tasks completed, success rate, exception rate.
• Where did it disagree with humans? Disagreement is not failure — it is the signal showing where the Worker is calibrated against organisational judgement.
• What did it cost? Inference cost + operator time + any escalation overhead.
• What is the ROI delta versus the manual baseline? If the baseline was not measured before launch, measure it now.

Day 60 — Scope Review

Sixty days in, the Worker has earned the right to either expand scope, narrow scope, or be retired. Most organisations skip this step and let scope drift — which is how a clean tier-1 customer triage Worker ends up being asked to handle complaints, then refunds, then disputes, with steadily worsening reliability.

The Day 60 question is binary for each adjacent task: does this belong to the Worker, or does the Worker explicitly hand it off? Write the decision down. Update the runbook. Tell the team.

Day 90 — Investment Decision

Ninety days in, the Worker is either earning its keep or it is not. The Day 90 review surfaces the investment decision honestly: keep operating as-is, invest in the next capability (more autonomy, deeper integration, second use case), or retire and harvest the learnings.

We retired our first Digital Worker at day 90 even though it was working. The ROI was real but small, and the operator time it consumed was better spent designing the next two Workers. Calling it a graceful retirement instead of a failure changed how the team felt about it.
— Head of Transformation, mid-sized Malaysian logistics group

Who owns the cadence

The single most common failure mode we see: the cadence is assigned to 'the AI team' or 'IT'. Neither group has the business authority to make the scope decisions or kill the Worker if needed. The cadence must be owned by the business unit consuming the Worker's output, with a named individual on the hook each quarter.

What this looks like in practice

1. Before launch: establish the manual baseline (cost, time, quality). Name the day-30/60/90 operator.
2. Day 30 review: 30-minute meeting, one-pager pre-read, three decisions logged (any tuning needed, any scope changes, any escalation triggers).
3. Day 60 review: 45-minute meeting, decision: expand / narrow / hold scope. Write it down.
4. Day 90 review: 60-minute meeting with the original sponsor present. Decision: invest / continue / retire.
5. After day 90: roll the Worker into the standard quarterly business review cycle. It is now part of operations, not a pilot.

This is not exotic. It is a regular operating discipline that most Malaysian organisations apply to systems, processes, and people — but routinely fail to apply to AI workflows. Once you do, the failure rate drops sharply.

Book a 30/60/90 audit for an existing pilot →