Every operations team has a breaking point. It usually shows up between the 25-person and 60-person mark — when the founder can no longer be in every decision, when the head of ops still hasn't been hired, and when the same five people are doing the work of fifteen. The volume is the visible problem. The variety is the actual one.
Here's a 90-day sequence that takes a team out of that phase. We've run versions of it with eight clients now.
Days 1-7: The map
Before you ship anything, get the full picture. The team needs to agree on what's happening — not what should be happening.
Do this
- One-on-one with every team member: what do you do all day? What did you do today? What did you not get to?
- A flat-file workflow map: every recurring task, who owns it, how often, how long, who's downstream.
- The "what would have to be true" list: if we doubled the volume next month, what would break first?
You'll find 8-12 things on the breaks-first list. Most are not the things leadership thinks.
Days 8-30: Three quick wins
Pick three things from the breaks-first list with the following property: each is owned by one person, takes 3-6 hours/week of their time, and has been done the same way for at least three months.
For each one, ship a small AI workflow. Examples we've shipped:
- Onboarding doc generation (was 4 hours/new client → 20 minutes)
- Weekly status report compilation (was 3 hours/Monday → 25 minutes)
- Customer-success email triage (was 2 hours/morning → 30 minutes)
The point of these isn't the time savings, even though they're real. The point is to build the team's instinct that AI works on their workflows. Confidence compounds.
Days 31-60: The bottleneck
Now you've earned the right to attack the biggest constraint. For every team we've worked with, this is one of three things:
- Senior review. One leader is the bottleneck for everything that goes out the door. The pattern: make their decisions reviewable in minutes, not hours. Pre-fill structured docs, surface only the parts they need to weigh in on, auto-handle the obvious cases.
- Handoffs. Work gets done in one tool, lives in a different tool, and ships from a third. The pattern: connect the three, with structured records flowing automatically.
- Quality control. Output quality is uneven because everyone has their own version of "good." The pattern: codify the standard, build the checker, surface drift before it ships.
This is the 4-6 week build. It's the most expensive single piece of the 90 days, and it's where the real capacity comes from.
Days 61-90: The repeatability layer
The last 30 days are about turning the wins into systems other people can run. Documentation, training, ownership shifts. Find a second person in the team who can be the AI champion — not your most senior person, but the second-most engaged. That role will pay back forever.
Run a retro. Note what broke. Note what compounded. Note what surprised you.
The team that adds 40% capacity in 90 days didn't work 40% harder. They stopped being interrupted by the work that should never have required their attention.
What this looks like in numbers
For an ops team of 8-12 people running this playbook, expect:
- 15-25% capacity unlock by day 30 (the three quick wins)
- 30-50% total capacity unlock by day 90 (after the bottleneck work)
- One AI champion in the team who can lead the next quarter independently
- Two or three workflows the team can't imagine living without
What goes wrong
Three traps we've seen:
- Skipping the map. Teams jump to "let's automate proposals" without confirming proposals are the bottleneck. They usually aren't.
- Trying to ship five things at once. Three is the right number. Five is two too many.
- No designated owner. Every AI workflow needs a human owner who notices when it drifts. Without one, quality degrades quietly.
Run it the way we've described. Run a real retro at day 90. The version of your team that emerges is the one that can absorb the next growth phase without breaking again.