Onboarding in 4 Days

When Mira Patel raised Flowpilot $4M seed round in March, she had a problem most founders dream about: too many customers signing up.

Flowpilot — a video collaboration tool for marketing teams — had 240 new accounts in their first month after launch. With a team of five engineers, two of them were spending 70% of their time on customer onboarding. Not building product. Not fixing bugs. Just walking new customers through setup, manually provisioning workspaces, and chasing missing fields in welcome forms.

We were the bottleneck. New customers were waiting 3 weeks to fully onboard because we couldn't keep up.

— Mira Patel, CEO at Flowpilot

Three months later, Flowpilot onboarding takes 4 days. Their engineers are back to shipping product. New customer satisfaction scores went from 6.4 to 9.1.

Here's exactly what they built.

The starting point

Flowpilot onboarding had six stages that all required human attention:

1. Welcome email + account creation

2. Workspace provisioning

3. Team member invitations

4. Brand kit upload

5. First video tutorial

6. 30-day check-in

Each stage had a different person doing slightly different things. Some used Linear tickets. Some used a Notion checklist. The CEO was personally onboarding enterprise customers via email.

The result: every customer's onboarding looked different, took different amounts of time, and required different people to remember different things.

Mapping the actual workflow

Mira's team started by writing down what onboarding actually looked like — not what they thought it should be, but what was happening today.

They found 38 steps across the 6 stages. About half were automatic in their existing tools. The other half were "someone has to remember to do this."

The 19 manual steps became Flowpilot automation list.

The 12 workflows

Rather than build one giant "onboarding" workflow, they split it into 12 smaller ones — each triggered by a specific event:

• Account created → triggers welcome email, provisions Flowpilot workspace, creates Slack channel

• First team member invited → sends collaboration tips, schedules check-in email for day 3

• Brand kit not uploaded after 48hrs → sends a "need help?" email with a Loom walkthrough

• First video published → triggers celebration email + invites them to the community Discord

Each workflow handles one job. Together they cover the full journey.

Connecting the data

The hardest part wasn't building the workflows. It was getting the data right.

Flowpilot customer data lived in three places: Stripe (billing), Postgres (product data), and HubSpot (sales). Different identifiers, different formats, different latencies.

Nimbus's integration layer let them define one canonical "Customer" record that updated from all three sources. Now when a customer hits any milestone, every workflow has access to the full context.

The results, three months later

The CSAT jump was the surprise. Mira expected speed gains. What she didn't expect was that automated onboarding actually felt MORE personal than the manual version.

Because every customer got the same level of attention. With manual onboarding, the loud customers got fast service and the quiet ones got forgotten. Now the system pays attention to everyone equally.

— Alexander Vance

What Mira would do differently

• Start with the smallest workflow first. Don't try to map all 38 steps before automating any of them.

• Document the workflow before building it. The act of writing down the steps surfaced 5 redundancies.

• Connect data sources first, then build workflows. Flowpilot tried to build in parallel with data integration and it doubled their build time.

Flowpilot grew from 5 to 12 people in the 8 months since automating onboarding. They've onboarded 8,400 customers without adding a single onboarding-specific role.

“Nimbus didn't replace our team. It freed our team to build the next thing.”

— Sarah Chen