Titles lag reality, and sometimes they never arrive. Influence is still visible: your design becomes the default, your checklist shows up in another org's wiki, people ping you before they ping your manager. That is not mystique. It is a set of habits that compound.
Pick work that other teams already feel
Reducing p99 on an internal API used by eight squads beats rewriting a tool only your team loves. The test: if you disappeared for a month, would someone outside your standup notice the problem you were attacking? If not, it can still be good work; it will not grow influence.
Write down the user of the work. "Platform team" is not a user. "Checkout on-call during peak" is.
Leave an artifact, not a meeting
Meetings evaporate. A one-page diagnosis, a runbook, a reference implementation, a decision record — those get copied. Spend the last 20% of a project on making the artifact usable without you in the room: glossary, failure modes, how to extend it, how to tell it is broken.
Review other teams' designs when asked, and be specific. "Looks good" does not travel. "This retry will double traffic to billing on a 502; here is the budget" does.
Influence is not calendar density
Being invited everywhere can be a trap: you become a bottleneck and you stop doing the deep work that made the invites happen. A staff-shaped pattern is office hours plus documents, not a 35-hour meeting week. Decline with a pointer: "I wrote this up here; ping me if the constraint is different."
Credit other people in public. Influence that looks like credit-stealing dies the first time a director asks around. Influence that looks like "they make the people around them better" gets staffed.
Talk to your manager like an investor
Once a quarter, state the bet: "I am spending 30% of time on the payment retry story because three teams page on it." Ask what would make that the wrong bet. If they wanted a different bet, it is better to know before promotion packet season.
You do not need permission to write the document. You might need permission to change another team's SLO. Start with the document anyway; many SLO conversations are stuck on missing shared facts.
Titles are a lagging indicator. The leading indicator is whether your writing and your reviews change what ships when you are not in the stand-up.
A worked week
Monday: write a one-page note on why checkout p99 is the payment-retry storm, with one graph and a proposed budget (max 2 retries, idempotency key). Tuesday: review two designs from other teams and leave the retry budget as a concrete comment. Wednesday: office hours, 45 minutes, agenda in the calendar description. Thursday: pair with the owning team on the flag, not on rewriting their service. Friday: send the manager the bet in four sentences, including what you will stop doing.
The artifact is the note plus a tiny reference implementation of the key header, not a slide deck. If nobody cites the note in two weeks, the problem was not “lack of title”; the note was not useful or not findable. Put it where on-call already looks.
Failure modes
Shadow management. You become the person who must bless every change. Influence that requires your calendar does not scale and eventually gets you labeled as a blocker.
Drive-by architecture. Comments on every RFC without owning an outcome. People stop asking.
Optimizing for visibility over users. Internal talks about process while the pager still fires. Directors notice both; only one is influence they will staff.
Skipping the manager. Cross-org work that surprises your skip-level creates political debt. The document can start without permission; the SLO change cannot.
When not to push for influence
When you are still ramping and do not yet know which problems are real. When the team is in a Sev-1. When “influence” would mean overriding a team that has a written decision you simply dislike — write a dissent, do not campaign in Slack. When the work only massages your promotion packet and does not change user or operator outcomes.
Review checklist
- The user of the work is a named on-call or customer, not “the org.”
- There is an artifact that works without you in the meeting.
- Calendar time is bounded; declines point at documents.
- The manager can repeat the bet in one sentence.
A worked failure mode
An engineer starts commenting on every RFC to "be visible." Authors route around them. Another silently rewrites a team's service over a weekend; it works, and the team stops trusting unannounced changes. Influence dies both ways: noise without ownership, and ownership without consent. The failure is optimizing for presence instead of being useful on a constraint the org already feels. Pick a painful interface (deploy, on-call, a flaky test suite), make it reliably better, write down the new default, and leave the keys with the team. Influence is earned when people ask you to look, not when you announce a personal brand.
When this is the wrong tool
Shadow leadership is the wrong strategy if you are being underleveled and need a title for pay bands you cannot otherwise reach; negotiate or leave with a clear-eyed packet. Do not grow influence by collecting approval on work you do not intend to support in production. It is the wrong tool for avoiding management if the only power in the org is a reporting line. Use informal influence to move technical defaults; use the promotion process for compensation.