# How do I lead a technical initiative across several teams?

> Find a sponsor who owns the outcome, write the problem down before the solution, start with one team, treat a no as information, keep every group current and decide how it ends

- Author: Ash Gaikwad (https://ashgaikwad.com)
- Published: 2024-03-15
- Topics: engineering leadership, staff engineer, technical strategy, migration
- Canonical: https://ashgaikwad.com/blogs/how-do-i-lead-a-technical-initiative-across-several-teams

To lead a technical initiative across several teams, find a sponsor who owns the outcome, write the problem down before you propose a solution, and start with one team before the rest. When a team says no, treat it as information about what adoption costs them. Keep product, design, support and leadership current on a steady rhythm, and decide at the start how the initiative ends.

I am Ash, a Staff engineer in Toronto. At a company I worked at, I led the migration of 120 services to Kubernetes with no downtime, sponsored by the CTO, and 9 engineers were affected. In my current job I lead a move from a monolith to services, also sponsored by the CTO, that starts with one team. I got that plan agreed with product, design, customer support and engineering leadership, and I keep those groups current as each part moves. At another company I led a team of 7 that restructured a tangled MVC codebase in place with domain-driven design, and I wrote the architecture documents and ran the design reviews.

This article is for staff engineers and engineering leaders who have been handed a change that no single team owns. I link the sources behind the method, and where something is my opinion, I say so.

![How I lead an initiative across teams in five steps. I find a sponsor who owns the outcome, write the problem down, start with one team, make yes cheaper for the other teams, and finish by switching off the old way. Product, design, support and leadership hear from me all the way through.](/images/blogs/lead-an-initiative-across-teams.svg)

## Leading without authority means borrowing it

Leading an initiative usually means asking people who do not report to you to change how they work. In his guide on [staying aligned with authority](https://staffeng.com/guides/staying-aligned-with-authority/), Will Larson writes that "Titles come with the sort of power called organizational authority, and that variety of authority is loaned to you by a greater organizational authority." So be clear about who lends it to you.

There is a second half. [Slack's engineers](https://slack.engineering/how-big-technical-changes-happen-at-slack/) describe a company where client teams "are free to choose not to depend on your system, with few exceptions", and where "we (intentionally) place the burden of getting other teams to adopt new technology mostly on the change agent." My opinion is that both are true in most companies, including the ones that mandate. A sponsor can decide the order and settle a tie, but the engineers on the other team still decide how hard to try.

## Find a sponsor who owns the outcome

The Kubernetes migration and the move I lead now both had the CTO as sponsor. In my opinion a sponsor needs three things. They own the outcome you want to change, they can settle a disagreement between two directors, and they will back the trade-off you ask for, such as a team's quarter. A sponsor with a bigger title and no stake in the outcome agrees politely in the meeting and does nothing after it.

Do not expect the sponsor to do the pushing. Larson wrote about career sponsors in [Find your sponsor](https://staffeng.com/guides/find-your-sponsor/), but the line fits an initiative too. "Sponsors are folks with more organizational capital than bandwidth to deploy that capital, and they'll help you most when you align the pieces for them." Bring a written problem, the decision you need and the date you need it by. Two of his rules for working with the person above you also apply. "Never surprise your manager" and "Don't let your sponsor surprise you" are both in the staying aligned guide.

## Write the problem down before the solution

Larson's advice on design documents, in [Writing engineering strategy](https://staffeng.com/guides/engineering-strategy/), begins with "Start from the problem. The clearer the problem statement, the more obvious the solutions." A page that states the problem, what is in scope and what is not is the cheapest way to get four groups arguing about the same thing.

[Google's design doc practice](https://www.industrialempathy.com/posts/design-docs-at-google/) adds two sections I would keep. Non-goals are "things that could reasonably be goals, but are explicitly chosen not to be goals." Alternatives considered "is one of the most important ones as it shows very explicitly why the selected solution is the best".

Do this before any solution because groups arrive wanting different things. Martin Fowler's article on the [strangler fig](https://martinfowler.com/bliki/StranglerFigApplication.html) says "Too often we see muddled objectives, with different groups wanting different things", and says to "establish alignment on the key outcomes early, and then revisit regularly to reinforce the alignment". In my work the document is where that agreement happens. I wrote the architecture documents and ran the design reviews for the domain-driven design work, and the plan for the move from a monolith to services was agreed across product, design, customer support and engineering leadership.

Larson also says "Gather perspectives widely but write alone", and review the draft with the people who will rely on it. The Rust project's [RFC process](https://github.com/rust-lang/rfcs/blob/master/README.md) is the public version of the same idea. Changes that are "substantial" go through "a bit of a design process and produce a consensus among the Rust community and the sub-teams".

## Start with one team, and choose it on purpose

Slack's engineers say that for React "a big-bang rewrite was out of the question". Fowler says of the all-at-once replacement that "we've seen this simple-sounding plan go down in flames most of the time", and that the gradual alternative "begins with small additions".

The initiative I lead now starts with one team before the rest. On the Kubernetes migration, traffic moved one service at a time and each move had a tested way back. I wrote up the order of those moves in [how to migrate many services to Kubernetes without downtime](/blogs/how-do-i-migrate-many-services-to-kubernetes-without-downtime).

My opinion is that the first team needs four things. It has a real need that the initiative serves, so it gains something and is not only volunteering. It is representative enough that what you learn applies to the next team. Its work matters, but not so much that one bad week ends the initiative for everyone. And its lead will tell you the truth. The easiest team teaches you almost nothing, and the most critical team gives you no room to learn.

The first team is also your evidence. In Slack's React story, "The functional prototype did much more to win converts than any first principles argument or whiteboard sketch could have done."

## What to do when a team says no

Start by treating the no as part of the work. The Slack post says "Some of the work of Phase 2 is fundamentally more like product work than like what-you-might-think-is-engineering", and it lists building things that "close the gap between current practice and the change you're making". A team that refuses is telling you what adoption costs them.

On the Kubernetes migration I built one deployment template with health checks, resource limits and rolling deploys. All 120 services deploy through it, so no team writes its own infrastructure. That is the cheapest yes I know how to offer, because the team gets something that works and not a migration guide.

Then find out what the no is made of. In my experience a no is usually one of four things. There is a technical objection you should take seriously, a cost the team pays now for a benefit someone else gets, a worry about risk, and attachment to the old way. Slack's advice for stragglers matches this. If the holdouts are "functionally motivated, by real capabilities the new system lacks", improve the new system or wrap it so it behaves like the old one. For attachment to the old way, "person-to-person outreach is usually a lot more effective than public, high-stakes debate".

Use the sponsor last, and for the right thing. Escalate a trade-off between priorities, such as whether a team's quarter goes to this or to a feature. Do not escalate to win a technical argument. Tell the sponsor before the team learns you went there. Sometimes the no is correct, and then it is a change to your plan.

## Keep product, design, support and leadership current

Engineers are not the only people an initiative touches. I got the plan agreed with product, design, customer support and engineering leadership, and I keep those groups current as each part moves. Fowler's point that alignment must be revisited regularly is the reason, because agreement on day one fades as the plan meets reality.

Each group needs a different sentence. Product needs the dates and scope that are changing. Design needs what users will see. Support needs what to tell customers and what might break, before a customer tells them. Leadership needs the risks and the decisions waiting on them. Write one short update, trim it for each group and send it on a fixed rhythm, so nobody wonders when the next one comes. Larson suggests "weekly email updates" or a Slack thread for keeping a manager informed, and the same devices work for a wider audience.

This is what Tanya Reilly calls glue work, and it takes real hours. In her talk [Being Glue](https://noidea.dog/glue) she says "Managed deliberately, glue work demonstrates and builds strong technical leadership skills. Left unconscious, it can be career limiting." Put it on your plan as work.

## How to end an initiative

Larson puts the value of finishing this way in [Work on what matters](https://staffeng.com/guides/work-on-what-matters/). "We only get value from finishing projects, and getting a project over the finish line is the magical moment it goes from risk to leverage." Decide how the initiative ends before it starts. I would write down three things.

The first is a measure that tells you it worked. For the Kubernetes migration I watched recovery time, which fell by 80%, and on-call load, which also fell by 80%. I also halved the infrastructure bill after the move. The second is a rule for holdouts. Slack says "Even very successful projects might not migrate every last use case to the new way of doing things", so decide which exceptions you accept and which old systems you switch off. The third is a hand-over. I routed each alert to the team that owns the service and wrote runbooks, which is what giving a system to its owners looks like in practice.

Some initiatives should stop. Slack describes a shared client library that never gained traction and was discontinued, and says "We took what we learned from the LibSlack effort". Write down at the start what would make you stop, so that stopping is a decision you already made.

## What this costs, and what I would not do

Leading this way costs time, mostly not coding time. Starting with one team is slower at the start than telling everyone to move, and the document is overhead, as Google says. Your authority is borrowed too. Larson notes that "What's bestowed can also be retracted", so if your sponsor leaves, find the next one before you need them.

There are four things I would not do. I would not roll out to every team at once, because you learn the same lessons on everyone together. I would not rely on a mandate without a plan. I would not write the document by committee. And I would not argue with a holdout in public.

## A plan for the first month

1. Write one page with the problem, what is in scope, what is not and what would make you stop.
2. Take it to the person who owns the outcome, ask them to settle ties on order and priority, and agree how often you will report.
3. Share the page with the lead of each affected team, and with product, design and support, before you share a plan.
4. Pick the first team with the four tests above, and agree with them what success looks like before you start.
5. Send the first update before there is news, then keep to the rhythm.
6. When the first team is done, put what they taught you into the template and the documentation before a second team starts.

If rolling out coding agents is one of these initiatives for you, I covered the checks that belong around them in [how to set up guardrails for an AI coding agent](/blogs/how-to-set-up-guardrails-for-ai-coding-agents).

Expect to write more and code less than you planned. That is the work, and it belongs on your calendar.

## Frequently asked questions

### Do I need an executive sponsor to lead an initiative?

For a change inside one team, no. For one that crosses teams, yes, because someone has to settle ties about order and capacity. Larson describes organizational authority as loaned from above, so the person lending it has to be on your side.

### How do I lead a technical initiative when nobody reports to me?

Write the problem down, make adoption cheaper than staying put, and let the first team's result persuade the others. Slack's engineers put the burden of adoption on the person driving the change, so plan for that.

### How do I know the initiative is finished?

When the measure you chose at the start has moved, the holdouts are decided and the owners hold the alerts and the runbooks.
