Medicine does not hand a duty from one caregiver to another on goodwill; it writes a protocol, names a person at each line, and schedules the review before the first dose is given. Technology teams would spare themselves a remarkable amount of grief by copying that habit, particularly around NoOps — the goal of automating and contracting out so much operational work that no dedicated operations role survives. Done carelessly, the handoff fails exactly where clinical handoffs fail: nobody owns the follow-up. Done as a protocol, it works often enough to be worth the effort. Below is the seven-step version we would run for a clinic’s website, patient portal, or booking system — each step produces one document, and the protocol ends with a signed summary rather than a shrug.
The NoOps Handoff Protocol: Seven Steps
- Chart the patient. List every routine operational task the platform generates — certificate renewals, dependency updates, backup checks, access reviews, invoice scrutiny, the periodic scramble when a vendor changes an interface. Estimate frequency and monthly hours for each. This chart is the baseline against which every later claim of automation is measured, and like any baseline it is worthless if drawn from memory rather than records. Pull the last six months of invoices, alert histories, and calendar entries, and let those sources argue with each other until the list stops moving. Clinics that skip the charting step almost always undercount by a third, and the missing third resurfaces as emergencies.
- Separate the vital signs from the elective. Mark each task as execution or judgement. Execution — patching, housing, replication — moves to a vendor cleanly. Judgement does not: the records a clinic keeps and for how long, who is permitted to log in, the settings currently in force, and who holds access to which system. Those four stay with the clinic in every service model on the market, and pretending otherwise is the error this protocol exists to prevent.
- Assign the treating clinician. Write one name against every task that remains. The pass mark is zero unassigned tasks — anything marked “probably the agency” counts as unassigned and returns to the list. A duty with no name is a future incident with a booking date, and naming is the cheapest intervention in the entire protocol. Where two people share a duty, write both, with the accountable one first — shared ownership unmarked is how tasks fall between stools during leave and staff turnover, which is precisely when platforms pick to misbehave.
- Set the monitoring plan. Two numbers go on the chart: the running total of duty hours plus pages (the toil figure), and interruptions per twelve-hour shift, with two as the ceiling practitioners consider sustainable. Both get reviewed monthly, and both belong on a chart visible to more than one person — private numbers drift, public ones get managed. A rising toil line after an automation project means the project moved work, not removed it — the instrument catches what the launch announcement hides.
- Write the discharge plan. One restore actually rehearsed inside the past ninety days is the discharge criterion for any data the clinic cannot afford to lose; exports live somewhere the production platform cannot touch. Alongside it, work out the platform’s real availability limit by multiplying what each critical dependency promises, and align anything published with the product. Three nines multiplied become less than three nines — the arithmetic is unforgiving and the marketing is not. Where the published promise to patients must exceed the ceiling, the bridge is engineered: cached pages that serve through a provider’s bad minute, a queue that smooths a hiccup, a status note that keeps families informed. Bridges are designed objects; they do not ship with subscriptions.
- Dose the automation gradually. Move one task at a time from the human list to the scripted list: script it, watch it through two cycles, then release the human hour. Never automate a task nobody understands, because the script inherits the misunderstanding at machine speed. Sequence the easy, repeatable items first — renewals, exports, reports — and leave judgement items on the human list permanently. Each handoff gets a one-line record: what moved, on what date, who verified it. That log reads as bureaucratic until the first time a scripted task misfires and the team needs to know exactly which change landed when; then it reads as the difference between a ten-minute fix and an archaeology project.
- Book the follow-ups before discharging the case. Quarterly re-chart of the task list, quarterly restore drill, annual re-read of every critical contract with an exit estimate attached and dated. The follow-up schedule is the protocol; everything before it is preparation. Skipping it converts a controlled handoff into an abandonment, and abandonment is the pattern behind most no-team failures the industry has documented in public post-mortems. Book the dates in real calendars with real attendees before step one begins; a follow-up that exists only as an intention is not a follow-up.
Contraindications: When Not to Proceed
Before the steps, a note on candidacy. The protocol assumes a platform of modest scale with a small, capable team behind it — the setting where the lean arrangement has its documented successes. Larger organisations with round-the-clock demands will recognise every step but should expect the staffing conclusions to differ, and nothing in the sequence is wasted by running it; the chart, the names, and the drills pay their way in any arrangement.
Some patients should not take the treatment, and the same is true of platforms. Steady, heavy load makes rented capacity costlier than owned — a utilisation graph settles that question in a day. Exotic hardware or strict timing needs operators on hand by definition. And if a provider retiring the product would stop the clinic’s books, the risk of the transition outweighs its savings; note that providers do retire products — Google closed its IoT Core service in 2023 and every customer inherited a migration on Google’s calendar. One more contraindication is human: if nobody on the team can say what the booking system does when the payment provider is down, stabilise that knowledge first. Automation amplifies whatever understanding exists, including none. A final contraindication is temporal: never begin the handoff during your busiest season, however quiet the vendor’s onboarding team promises the migration will be. Elective procedures get scheduled for quiet weeks in medicine for excellent reasons, and this is one of them.
The Follow-Up Schedule, In Clinic Terms
Practitioners who grade these arrangements describe a maturity ladder whose bottom rung is the announcement — the claim without the chart — and whose top is a deliberately trimmed operation, measured into existence. Ninety days is the standard initial treatment span: month one produces the task chart, month two produces the names, month three produces the rehearsed failure and the priced exits. After that, the protocol settles into its chronic-care rhythm: short quarterly reviews, an annual full re-chart, and a standing rule that any new duty gets a name the week it appears.
For the reference work behind every term used here, the Ops Atlas NoOps page collects the definitions, the tracked metrics, and the documented failures in one place; it rewards the same reading habit as a good clinical review paper, and cites its cases. The protocol’s spirit — triage, assign, monitor, follow up — will already be familiar to readers of our guides on managing cough and cold safely at home and on the symptoms, causes, and diagnosis of balance problems: both begin with an honest inventory before any treatment is chosen. And for the digital side of patient care, our overview of how AskDoctorLive supports online ENT consultations shows the same principle in service design — clear ownership at every step, and a human reachable when the automated path runs out.
The NoOps Protocol Card
If the whole piece must fit on a card beside the monitor: chart the tasks, split execution from judgement, name everything, watch two numbers, rehearse the restore, dose the automation one task at a time, and book the follow-ups before celebrating. Teams that follow the card end up with fewer duties and defensible arrangements; teams that skip it end up with the duties anyway, arriving unscheduled, at the hour least suited to triage.