ServiceNow Multi-Level Governance: Building Aligned Platforms Through Strategic Structure

Most platform teams can produce their governance structure on demand, executive steering at the top, a design authority in the middle, delivery teams at the bottom, three tidy rows with reporting lines between them. Ask a product owner in the same organisation to show you the written rule that governs the decision they are about to make, and the slide stops helping. The structure exists, what it produces does not.

The operating-model reviews I get called into have taught me to stop asking who sits on each board, because the question that separates the aligned platforms from the merely governed ones is quieter than a seating plan. “What does this tier hand down, in writing, and when was it last updated?”

The organisations that work hardest at joining their tiers together, with liaison seats and cross-briefings and minutes circulated to everyone, are frequently the ones that lose alignment fastest when a single person leaves. Every board still met, every agenda was still full, and yet the estate drifted within a quarter of one resignation, because the connective tissue was never the structure. It was one person's mission.

Cross-membership is borrowed alignment

Seats connect people. Only published constraints connect tiers.

The instinct, when tiers drift apart, is to overlap them. Put an architect on the steering committee, invite an executive to the design review, send every set of minutes to every other board. It feels like joined-up governance, and for a while it works, because the people carrying the context are diligent and present. What has actually been built, though, is alignment that lives in individuals, and it scales as meetings multiplied by people. Double the tiers and you double the diaries, and lose the one person who attended everything and you lose the only thread connecting strategy to build. The tell is a decision being relitigated because nobody from the right board was in the room. A decision that binds only the people who attended it is not governance, it’s attendance.

There is a running cost to all of this that nobody prices. By McKinsey's survey work, on the order of six in ten managers report that at least half the time they spend on decision-making is used badly. A governance model that can only synchronise its layers by putting more people in more rooms is a machine for producing exactly that waste, and it produces more of it every time the platform grows.

A tier's real product is the envelope it hands down

Each layer owes the one below a written envelope it can build within without asking.

Look at what a well-run operating model actually asks of its layers, and a pattern appears that has nothing to do with seating. The strategic tier hands down direction, meaning what the platform is for, what is funded, and the roadmap a demand board can test a portfolio against. The design and technical tier hands down an envelope, the standards and constraints within which delivery teams are free to build without asking anyone. The operational tier hands down guardrails that keep daily work inside the platform's health. Each layer, done properly, produces a written artefact the layer below consumes, and the artefact does the aligning.

Call it the absentee test, because a governance tier is working when the layer below can make its next ten decisions correctly without attending the tier's meeting. Amazon ran the most famous version of this experiment with the internal mandate, issued in the early 2000s, that every team expose its work to every other team through service interfaces, with no coordination path around them. Teams stopped aligning by meeting and started aligning against a published contract. The point was never the technology, it was that a written interface scales where a conversation does not.

The everyday version has been running for a century, because nobody convenes a planning committee to approve a standard extension to a house. The builder builds to the published code, the inspector checks against it, and the committee sees only the variances. That is a multi-level structure in which almost every decision is made at the bottom, correctly, by people who have never met the people at the top. The code is doing the governing, and it does it at a scale no committee calendar could ever match.

✅ PRINCIPLE

Apply the absentee test. A tier is working when the layer below can make its next ten decisions without ever attending its meeting.

Send exceptions up, not updates

The traffic between tiers is the diagnostic. Healthy structures pass exceptions upward, hollow ones pass updates.

Once the downward artefacts exist, the upward traffic changes character. What a higher tier should receive from the one below is exceptions, the request that does not fit the envelope, the standard that has begun to block more value than it protects, and the health signals that show whether the envelope still retains its integrity. What most boards receive instead is status, slide after slide travelling upward to be noted and change nothing, while the steering committee spends its hour deciding things a published rule should have settled two tiers down.

Netflix's leadership described their version of this as leading with context rather than control. Publish the strategy and the constraints clearly enough that decisions can be made locally, and treat every escalation as a prompt to improve the context rather than as proof that the top is needed. The escalation is not the system working, it’s the system reporting a gap in what’s been published.

🔭 FIELD NOTE

The fastest structural diagnosis I know, ask three product owners where the current build standards live. Three different answers means the envelope exists in heads, not in writing.

The fuller diagnostic takes one afternoon, because all it asks is that you take last quarter's minutes for each board you run and sort every item into two piles, the items that changed a written rule and the items that were noted. A board whose minutes never change a published artefact is not governing the platform, it’s watching it. The structure will keep every appearance of diligence, the attendance, the actions log, the escalation paths drawn on a slide, while the actual behaviour of the estate is set by whatever each team last remembered being told.

Where this doesn't apply

A platform in its first year has no precedent to publish, and a transformation in mid-flight rewrites its own rules faster than any document can track them. In both cases alignment genuinely does live in rooms for a while, and pretending otherwise produces envelopes nobody trusts. The discipline is to treat those rooms as temporary, and to make each hard-won decision leave the room as the first line of a published constraint. The same holds in a live incident, where you convene, you decide, and you write the rule down afterwards. Publishing is the steady state, not the crisis response.

The bottom line

Rewrite each tier's charter as a publishing obligation. For every board in the structure, name the artefact it owes the tier below, where that tier finds it, and the date it was last updated. Then name the exception path it owes the tier above, and hold the accountable owner of each artefact to the same standard you would hold an accountable owner anywhere else on the platform. A board that can name its artefact is a tier, and a board that cannot is an audience, whatever the org chart says.

A governance tier's product is not the decision it makes in the room, it is the written envelope the next tier down can build against without attending.

P.S. This fortnight, ask one product owner to show you, in writing, the rule that governs their next contested decision. Time the search, because that number measures your alignment better than any maturity assessment will.

🎯 QUICK REFERENCE: THE PUBLISHING OBLIGATIONS

• Every tier owes the layer below one written envelope, current, findable and dated.

• Every tier owes the layer above exceptions and health signals, never status.

• Apply the absentee test before you add a seat to any board.

• Minutes that change no written rule are watching, not governing.

• Rooms are for year one and live incidents. Publishing is the steady state.