An organisation chart tells people who they report to. It does not tell anyone what to do on Monday morning. The artefact that does that is a process flow with roles on it, and the reason so many restructures stall a quarter after the announcement is that the chart shipped and the flows did not.
A European travel and tourism group had reached exactly that seam. A structural blueprint for a new central digital function was finishing, and the question underneath it was whether the thing could actually be run. Around thirty processes were identified as needing description. The commitment was level-two process flows with roles for all thirty, inside a single quarter, running in parallel with the organisation design, the governance design and a group-wide reorganisation happening alongside.
Thirty flows with roles in three months is not a normal rate. Most teams that try it deliver eight good ones and a folder of drafts. The interesting part of this engagement is not that the number was promised. It is that the proposal contains its own arithmetic, and the arithmetic only closes under one method.
The challenge
Read the commercial shape and the constraint appears immediately. The work was quoted as three people across the quarter: five days of senior sponsorship, thirty-nine days of design and facilitation, and seventy days of production. One hundred and fourteen days in total, at £158,297 in professional fees. The seventy production days are the ones that matter here, because that is the person who actually makes the flows.
Seventy days across thirty flows is 2.33 days each, and that figure is a ceiling rather than a measurement, because the same seventy days also produced other elements of the operating model. Two and a third days to describe a process end to end, with the roles attached, validated by the person who owns it. Anyone who has run a process workshop knows that two days is roughly the length of the first workshop, never mind the drafting, the review and the rework.
The same proposal prices the alternative, which is what makes the comparison honest rather than rhetorical. A separate provision offered twenty-five days to describe five key processes in detail, at £1,311 per day. That is five days per process, and £6,555 each. It is the only per-process unit cost the document carries, and it is the yardstick for everything else: when this team had to work a process properly from the ground up, it cost five days, not two.
Thirty processes at five days each is one hundred and fifty days. The quarter had seventy.
The approach
The gap does not close by working faster. It closes by changing what the work is.
The flows were specified as being built on two sources at once: the client's own existing processes, and a pre-built process library the team brought with it. Neither source alone would have been enough. The existing estate had real processes running in it, documented unevenly and owned by real people, but it belonged to the organisation being replaced. The library had clean reference flows for a digital function, but reference flows describe nobody's actual business. Joined together, they turn the job from authorship into editing, and editing is the operation that fits in two days.
The sequencing rule reinforces it. The plan started with the complex and critical areas rather than the easy ones: financial processes, capacity management, standing up a delivery centre, gating, product end-to-end management. Ten processes were described at level two in the first pass, the remainder followed, and only then did five go deeper. Starting with the hard ones is uncomfortable and correct, because the hard ones are where the library and the estate disagree, and that disagreement is the design decision the whole exercise exists to make.
One more rule, easy to skim past and load-bearing: confirm the process owner and sponsor, as a plan activity in its own right, before designing the process. Then workshop, refine, validate, each of the three written on the plan as its own item rather than assumed to happen inside the drafting. An edited flow that no owner has validated is not a faster flow. It is a document.
Look at where the amber block sits. The processes that cost five days each are the ones described as new, the ones the previous organisation did not have at all: customer experience, customer service management, ideation, delivery centre management, resource and practice management, service delivery management. Six are named, and the document names them as examples rather than as a closed list. Five were funded to that depth, and which five is not recoverable from the document.
That is the finding, and it inverts the intuition people bring to process libraries. A library does not save you the most on your hardest processes. It saves you the most where you already had something running, because there the library is a reference to reconcile against rather than a blank page to fill. Where there is no existing process, the library supplies a shape and nothing more, and shape is the cheap part. Those are precisely the processes that need the full five days, and the proposal treats them accordingly: fenced into a separate provision, separately priced, with a decision date attached rather than left as an open change request.
The outcome
What this document is has to be stated plainly. It is a priced proposal: £158,297 for the quarter, plus £53,375 held as a dated option, excluding tax and expenses. It designed the work, staged it and costed it. It did not report results, and nothing here should be read as a delivered outcome. Even the production booking is worth reading sceptically, because seventy days against a fourteen-week window is every working day in it, with nothing left over for a public holiday, a slipped workshop or a day of illness. The pace was already at its ceiling on the page, before anyone tested it in the room.
The method survives the caveat, and the arithmetic is what makes it portable. Three things would be done differently now.
First, the as-is half of the composition was recovered by workshop, and a workshop recovers what people believe the process is. Process mining over the workflow and case-management logs recovers what it actually was, at a fidelity no interview reaches, and it arrives already counted: how many variants a process really has, which ones carry the volume, where the handoffs sit. That changes the composition step from two documents being reconciled to a measured graph being reconciled against a reference, which is both faster and defensible when the process owner disagrees.
Second, a level-two flow with roles is the artefact that makes everything downstream specifiable. Straight-through processing has to be scoped somewhere, and it is scoped on a flow. So is model deployment: a machine learning model that scores or routes a case is a step in a process, and if the flow does not exist, the model has no place to be called from and no lineage trail to write to. Teams that skip the flow layer end up with models that work in a notebook and never touch an operating model. That is why our Consult and Platform work both start from the process inventory rather than from the data.
Third, the current wave of cautious language-model pilots is often aimed at exactly this artefact, drafting process descriptions from documents and transcripts. That is a reasonable use, and it changes none of the arithmetic above. A drafted flow is another inherited source, no different in kind from a library entry: it still has to be reconciled against what the logs show and validated by the person who owns the process. The workshop, refine, validate loop is not the slow part of this method. It is the part that makes the fast part safe.
Thirty flows in a quarter is not a productivity story. It is a sourcing decision, made before the first workshop, about how much of each artefact would be inherited and how much would be invented.