Aligning Engineering Operations

The problem

I inherited an Engineering Operations department made up of very different teams: DevOps and SRE, Data, IT (TechOps), Platform, CRM/ERP operations, and QA. I was excited by what the group could do. But each team thought and acted in its own silo, and struggled to explain its value to anyone outside. And how do you give a group like that one north star?

How I thought about it

I saw three ways every team creates value and scale: time, cost and risk. That gave all six teams a common language for talking about their work and for deciding what came first.

I also wanted a north star made of metrics, not features. Product north stars are often tied to building features X, Y and Z. I dislike that, because it leaves no room to switch to a better solution along the way. A metric doesn’t have to change if feature AA turns out to be the better play.

What I did

Finding cohesion. I started with an inventory of each team’s customers. That led naturally to an inventory of customer journeys, which made it easy to see the value each team provided and to set KPIs to track it. With those in place, we could move from an operational mindset to a product mindset.

The north star. With a common way to communicate and prioritize, and a lot of team autonomy, we hit our stride. We started 2023 focused on cutting vendor spend by 20%, so our priorities were:

• Handle every critical vulnerability immediately (unplanned work).
• First: cut costs where we can, while being smart about the consequences.
• Second: make the biggest improvements for our customers.

With vendor spend going well, though not finished, interest grew in lowering the company’s exposure to risk. So the priorities changed:

• Handle every critical vulnerability immediately (unplanned work).
• First: lower our overall risk exposure to a target.
• Second: cut costs further.

Results

We hit 2.4× our vendor spend target in six months. After the switch to risk, it took three months to mitigate three long-standing vulnerabilities and hit our risk targets for the year.

What I’d do today with AI

Time, cost and risk is still how I decide what is worth doing, and it is the right thing to give an agent as its goal. An agent can track every KPI against its customer journey all the time. The part I would not hand over is the pivot itself: choosing when the north star moves from cost to risk is a judgment call.