Managed operations
Not a licence you have to staff, configure and then justify. A bounded operation we take responsibility for running — with the technology, the instrumentation and the people around it — while the authority, the evidence and the customer relationship stay yours.
Somewhere in your organisation is an operation that is always slightly underwater. It has a name, a manager, a service level it misses in the same two months every year, and a headcount request that gets deferred. It might be inbound support. It might be claims intake, referral management, benefits eligibility, records requests, enrolment, collections, or the service desk. The shape is the same.
You have already tried the obvious moves. You added people, and the training pipeline could not keep up with attrition. You bought an automation tool, and it handled the easy third and made the hard two-thirds harder because now they arrive pre-annoyed. You outsourced it, and quality became somebody else’s definition.
The uncomfortable part is that nobody can tell you where the time actually goes. You know the handle time. You do not know how long a case waits between the steps that are measured, because the waiting happens in the gaps and the gaps are not instrumented.
Handing an operation to someone else usually means handing over the authority and the evidence with it, and that is the part nobody can get approved.
The reason "just outsource it" stalls is rarely price. It is that the moment the work leaves, so does your ability to say what was permitted, what actually happened, and whether the outcome was reached — and legal, compliance and the executive who signs are all being asked to accept that on trust.
A managed operation here is bounded in the other direction. The function is operated for you, but the authority model, the policy boundary and the evidence trail remain yours and remain inspectable. Every consequential action can carry a receipt that names the objective it served, the authority that permitted it, what executed and what changed afterwards.
That is also what makes the automation safe to increase over time. Because the boundary is explicit, more of the operation can move behind it as the evidence earns it, rather than in one negotiated leap at contract signature.
How long a case waits between the steps you already measure — measured by queue age and wait time per state, with the tail reported beside the median.
Work reaching a person only when policy or judgement requires it — measured by escalation rate split into policy-required versus capability gap.
Rework — cases that come back after being closed — measured by reopen rate and the reason class that caused it.
Whether the service level holds in the window it usually breaks — measured by attainment during your known peak, against the baseline taken before we started.
What a failed attempt costs you — measured by result-settled economics — work we accept responsibility for and fail to complete is restored, not invoiced.
a cost-per-contact figure. We have not run your operation, and a number from somebody else’s queue tells you nothing about yours. The observation period produces your baseline, and that is the only number worth comparing against.
A managed operation that requires you to migrate your telephony, your records system and your identity provider first is not a managed operation — it is a platform migration wearing a service contract.
The operating layer connects to what already holds your truth. Cases stay where your people already look. Identity comes from your identity provider. Evidence can stream to your SIEM so your security team sees the operation the same way they see everything else.
Where an integration does not exist, that is a named gap with scoped work attached, not a silent expectation that somebody will reconcile a spreadsheet on Fridays.
The customer relationship stays yours. The authority stays yours. The evidence stays yours and stays exportable. We operate inside a boundary you set and can narrow at any time without renegotiating the contract.
Where a control is designed rather than independently attested, we say so on the page rather than letting a questionnaire imply otherwise.
An outsourcing decision touches legal, privacy, information security, vendor risk and often labour relations before it reaches procurement. Each of them is asking a version of the same question: what happens to our obligations when the work leaves the building.
The answer here is that the obligations do not leave. The boundary, the policy and the evidence are all still yours to inspect, and the contract can say so specifically rather than generically.
One queue, one case type or one lane of one function — the one whose service level you can already name, and whose failure somebody feels every week.
The first phase should not move any work. It should watch the lane you chose and tell you where the time actually goes: how long cases wait, where they age, what comes back, and which escalations were policy and which were capability.
A number of organisations should stop there for a while. What that costs you is a baseline you did not have, and it is the only honest ground for the comparison afterwards.
Work moves behind the boundary only when the authority is written down and the evidence expectation is agreed. Then it moves one class of case at a time, because a managed operation that starts with the hardest cases fails for reasons nobody will be able to attribute.
Because quality was defined in a service-level table and measured by the party being graded. Here the outcome definition and the evidence for it are set before work moves, and the evidence is yours to read directly rather than to receive in a monthly deck. If cases are being closed without meeting the definition, you see it in the same place we do.
They should not accept the usual version, where authority and evidence leave with the work. Ask specifically whether the policy boundary stays enforceable by you, whether it can be narrowed without renegotiation, and whether every consequential action carries a receipt you can inspect. Those are the questions this model is built to answer yes to.
The URL is where people look; the model is not that narrow. Any bounded, repeatable operation with a queue, a decision and a defined outcome fits — claims intake, referrals, eligibility, enrolment, records requests, collections, service desk. If you can name the case type and what a finished one looks like, it is in scope.
Surge behaviour is part of the scope conversation, not an addendum after the first bad week. It gets a written expectation, a baseline from your own historical peak, and continuity evidence you can ask for before signing. If we cannot meet a surge you can already predict, that is better established during scoping than during it.