The word transformation does a lot of quiet work in a board paper. It implies a new organisation, new skills, new people, and a budget line proportional to all three. The cheapest test of whether the word has been earned is the role catalogue, because a catalogue is obliged to name every job the new model needs and then write, in a column next to each one, whether that job already exists.
A global payments acquirer and processor ran that test on itself. It was folding a set of acquired brand properties onto a single corporate site, with common content management, common analytics, common hosting and a common CRM underneath. Shared infrastructure creates shared decisions, and shared decisions need owners, so an organisation design was commissioned to say who those owners would be. The catalogue it produced named eleven roles. Two of them were new or a significant modification of an existing job. Eight were classified as existing or existing and modified. Five of the eleven sat outside marketing altogether.
The challenge
The design carried one hard constraint, and everything else in it is downstream of that constraint. No group function was to be created. There would be no central digital team to hire into, no neutral body to hold standards, no obvious desk for the person who decides what happens when separate business units want different things. Governance across the group had to be assembled out of roles that already sat inside those business units, with one of them taking the lead on behalf of the others.
That constraint is what makes the catalogue worth reading closely. With nowhere central to put anybody, every pan-group task had to land either on a job that already had an occupant, or on a job somebody would create locally and then ask to act for the whole group. There was no third option. The interviews had already flagged it: the skills were unlikely to all exist in the current structure, and some gaps would need new roles, development of existing people, or outside support for a period.
So the question the catalogue answers is not the one the programme title asks. It is not what does a digital organisation look like. It is how much of the digital organisation you say you are building do you already employ.
The approach
Every operating task was assigned first, pan-group or local, across nine domains. Only then were the tasks gathered into roles, and each role given a classification: existing, existing and modified, or new. Doing it in that order matters, because the reverse order produces an org chart drawn from ambition and then backfilled with tasks that justify it.
Six roles sat inside the marketing and product teams: a digital leader, a commercial leader, a digital marketing and social manager, a webmaster, a content publisher and an analytics manager. Five sat outside, in group communications, IT, change and operational teams: a support content manager, PMO support, an enterprise architecture design authority for future technology decisions, day to day management of the intranet and careers portal, and a recruitment social media manager.
Three of the eleven carried a new label on the catalogue. By the close of the work, two were confirmed as either new or a significant modification of an existing role: the digital leader and the overall webmaster. The third, PMO support, was assumed to come from the existing change function, which is a polite way of saying the work is new and the headcount is not. Every other role in the catalogue was somebody already on a payroll, with a scope note attached explaining what would now be different about their job.
Two details in that list deserve more attention than they usually get.
The first is what the digital leader, the senior role in the model, was actually given. The definition runs to three duties: shaping the overall digital plan, supplying the governance forum with the thinking it needs to make investment decisions, and being the key contact for the lead agency. None of the three is line management, and the pack says the same thing from the other side when it notes that one of the configurations would demand more influencing skill from that role. That is a convening role wearing a leadership title. That can be the right answer when no group function is permitted, but it should be chosen deliberately rather than discovered later, when a priority call has to be made and nobody can make it.
The second is the five roles outside marketing. If a programme of this shape is sponsored by marketing and budgeted by marketing, then at the point the catalogue is signed off it has funded six roles and assumed five. Those five decide whether support content is accurate, whether the prioritisation process actually runs, and whether the architecture of the estate is anybody's job. None of them appears in the marketing headcount request.
The headcount arithmetic closed the argument. Across the whole group, the redesign came to a net increase of one or two people depending on which configuration was chosen. The footnote is the interesting part: that net figure holds only because two existing roles come out of one business unit. Gross new roles are therefore three or four, offset by two removals, with additional cost acknowledged but not quantified where a role is upgraded to a higher grade.
The outcome
What this phase produced was a design, a shortlist of ways to configure it, a headcount impact expressed in FTE and a list of what remained open. It did not stand up an organisation. Team sizes and costs were still to be completed, budgets still to be defined, the forums still to be created out of existing groups, and guidelines, social policy and third party support all still to be settled. Everything above is a proposal that had been argued through to a number, which is further than most operating model work travels, and still short of a result.
The honest question is what we would do differently on the same brief now.
The classification column is the weakest instrument in the pack, and it is the one carrying the whole finding. Existing means a job title survives. It does not say whether the person in it has four hours a week spare or is already full, and the difference between those two readings is the difference between a reorganisation that works and one that quietly does not. Process mining over the publishing events, the ticket queues and the campaign systems answers that with observed frequency and duration rather than with an interview. The task split then gets built from what people actually do, and the catalogue inherits a defensible load figure next to every row.
The second change is that a shared analytics and content substrate creates an owner nobody catalogued. Somebody has to hold the definitions, the lineage and the pipelines that keep a common measurement layer honest across the group's businesses, and that person is not a webmaster and not a marketing analyst. In the Platform work we do, this is the role that decides whether the shared substrate stays trustworthy in its second year, and it is almost always missing from the first version of an org design.
The third is that early language model pilots in content drafting do not remove the content publisher. They raise the value of that role, because more drafts now arrive faster at the same approval gate, and a gate that was adequate at the old volume becomes the constraint at the new one. Our Consult engagements start by measuring that gate for exactly this reason.
Eleven roles, nine of which already existed, and a net change of one or two people. Say that out loud and the meeting changes character. It stops being a transformation and becomes a question of scope, sequencing and who is allowed to decide, which is a much better conversation to have before the budget is set than after.