Organisational Design Consulting
When the new org chart changes nothing
The pattern
A restructuring is announced. The boxes move, spans are rationalised, a layer comes out. Six months later the same decisions still require the same number of meetings. A pricing exception still waits on a forum that meets fortnightly.
What follows is a second attempt: another chart, a responsibility matrix circulated and quietly ignored. The fee and the retention risk of two reorganisations in eighteen months are the visible costs. The heavier cost is decision latency, carried on every commercial commitment while who decides stays unsettled.
- 01Cross-functional decisions that route upward to a single executive regardless of their size.
- 02Forums that review and endorse but cannot commit resource, so the real decision happens somewhere else afterwards.
- 03Duplicate capability standing in two or three functions, each with a defensible reason.
- 04Span of control widened when routine supervision dropped, so first-line managers have fewer things to check and more judgement calls to make.
Why it persists
Structure is the most visible element of organisation design and the least predictive of performance. It absorbs the attention because it is easy to draw. Performance follows the operating model underneath it: who holds genuine decision authority against who is consulted as a courtesy.
The first mechanism is inference. Very little of this is documented anywhere, so managers infer their decision rights from precedent and from who was overruled last time, and a new reporting line does not revise that inference. When the boxes move, people apply the same inference in a new box, and the same decisions take the same time.
The second mechanism is commercial. Restructures are justified on cost and span logic, both measurable before the change. Decision speed and cross-functional flow are measured almost nowhere, so nothing in the mandate obliges anyone to improve them. The redesign lands its cost target and leaves decision latency where it was.
Artificial Intelligence (AI) has now broken the assumption decision rights rested on, that whoever presents a recommendation can defend the method behind it: the pricing analysis arrives in a morning and nobody in the room can reconstruct it. A RACI has no row for a model, and the usual repair, another review step, cancels the saving the tool was bought for. AI governance starts there: accountability re-cut around what a person can own, the question asked, the inputs chosen, what was rejected, and the call to accept or override.
A chart tells people where they sit. What they may commit without asking is usually unwritten, and that is what sets how fast the organisation decides.
What the work examines
The usual entry point is a decision inventory: the twenty to thirty decisions that determine commercial performance, traced as they are actually made today. Who initiates, who must be consulted, where the file waits, and how many forums touch it. The calendar, the approval trail and the last two quarters of escalations describe an operating model more honestly than an interview does.
That establishes where authority actually sits against where the governance document places it, how long a decision of a given class takes from raise to commitment, and which two or three matter enough to justify the disruption.
Design work then follows one order, and the order is itself the argument: decision rights first, governance rebuilt around the decisions a forum can commit to, and structure drawn last. Drawn first, a structure locks in authority that nobody has yet agreed. Which two or three decisions are worth the disruption is your call, made with the inventory in front of you and settled in scoping.
Where this shows up in the business
- Day to dayDecisions stop circlingWhere decision rights are written down and the forum that can commit resource is named, a request that used to pass through three forums passes through one. That is countable within weeks.
- Over a yearThe numbers your board already readsDecision elapsed time on a defined set of decisions, forums touched per decision, duplicated capability removed and what that is worth annually, and senior time recovered from arbitration.
- Over several yearsSpan of control becomes a choiceOnce authority is documented and governance is standing instead of improvised, span of control turns into a design decision instead of a constraint you work around. The organisation stops paying to rediscover who decides every eighteen months.
None of those figures is a promised result, and POLIP publishes no client numbers. A diagnosis can establish where the latency sits, what removing it would cost in management attention, and what the evidence would support doing first. Whether it then moves rests with the executives whose authority is being written down.
Common questions
We have already restructured twice, and now AI. Why would this be different?
Because the object of the work is different. A restructure is scoped to reporting lines, spans and cost, and the two you have run were held to it. This work is scoped to decision rights, governance sized around decisions rather than functions, and the interfaces where work crosses functional boundaries. Automated analysis adds one item: who is accountable for a recommendation they cannot reconstruct.
The practical test: ask three managers two levels down what they may commit without asking. If the answers differ, the operating model was never the subject of either restructure.
Do you produce a new organisation chart?
Yes, but late and as a consequence. Structure is designed once decision rights, governance and the cross-functional interfaces are settled, because a chart drawn first forces the operating model to accommodate it.
You also get the part most charts omit: the sequencing plan, the interim arrangements, and what happens to the decisions that sit in dispute during the transition.
How do you measure whether the design worked?
Cycle time on a defined set of decisions, baselined before the change. How long a pricing exception or capital request takes from raise to committed decision, how many forums it passes through, and how often it escalates above the level that should own it.
Alongside that, the cost side: duplicated capability removed, meeting and approval load reduced, and senior time recovered. Both sets are measurable, and agreed with you at the start rather than at the end.
Can this be done without a headcount reduction?
Often, yes. A substantial part of the delay in most organisations comes from governance design and undocumented authority, neither of which requires anyone to leave. Where duplication has genuinely accumulated across functions, the design will name it and you decide what to do about it.
POLIP will not present a cost-reduction exercise as an operating model redesign, or the reverse. If your mandate is cost, that is a legitimate brief and it should be scoped as one.
Start with the decisions behind the boxes
Bring the last three cross-functional decisions that took too long and the forum calendar of your top two layers. That is usually enough to locate where the operating model is failing.