Back office ops · Production

Stepstone scales to 700+ n8n workflows and builds an AI chatbot to onboard employees into automation

The problem

As Stepstone's n8n footprint grew to hundreds of workflows, the shared Community Edition instance became structurally unsafe: credentials were shared across users, workflow ownership was muddled, execution monitoring was thin, and a single master account was unsustainable. Simultaneously, enabling a large non-technical employee base to automate safely—without producing a graveyard of abandoned workflows—required structured onboarding that did not exist.

First attempt

The Community Edition became untenable at enterprise scale: credential sharing, no proper role-based access, thin monitoring, and a single master account that could not support team-level isolation—including a requirement to keep GDPR-regulated employee data separate from main-instance admins.

Workflow diagram · grounded in source
1
User describes automation need
trigger
“A user opens it, says which team they're on, describes what they want to automate in plain language”
2
Feasibility questioning
ai_action
“the bot walks them through the feasibility step by step. It asks the kind of questions a senior engineer would ask”
3
Idea scoring
ai_action
“It scores how realistic the idea is”
4
Diagram and steps generated
output
“If the score is high, it draws a diagram of how the workflow should be built and lists the steps”
5
Claude scaffolds the workflow
ai_action
“The next iteration uses Claude as the underlying model and the n8n MCP to scaffold the workflow itself. Not perfect, by design. "The resulting workflow is intentionally not perfect, as this requires the user to engage with and learn the …”
6
User refines and learns
feedback_loop
“this requires the user to engage with and learn the underlying principles of automation, providing them with a valuable foundation to build upon”
Reported outcome

Since moving to Enterprise, Stepstone now runs over 700 active workflows in production—more than three times the volume from a year earlier—with the IT security team operating its own instance.
An AI chatbot guides new users through automation feasibility assessment. Operationally, a daily batch of contract mappings now runs in about 20 seconds, and a large product cleanup that would have consumed several working days was completed rapidly.

Reported metrics
Active production workflows700+
Workflow year-over-year growthmore than three times the volume from a year earlier
Job-related data documents parsed monthlytwo to three million
Daily contract mapping batch processing time (after)around 20 seconds
Show all 8 reported metrics
active production workflows700+
workflow year-over-year growthmore than three times the volume from a year earlier
job-related data documents parsed monthlytwo to three million
daily contract mapping batch processing time (after)around 20 seconds
manual time per contract mapping (before)two to three minutes
daily contract mappings requiring manual processing (before)60 to 70
time saved on product cleanup projectequivalent of two full working days for five full-time employees
contract entries updated in product cleanupabout 4,500
Reported stack
n8nClaudeOktaConfluenceJira
Source
https://n8n.io/case-studies/stepstone/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Since moving to Enterprise, Stepstone now runs over 700 active workflows in production—more than three times the volume from a year earlier—with the IT security team operating its own instance.

What tools did this team use?

n8n, Claude, Okta, Confluence, Jira.

What results were reported?

Active production workflows: 700+; Workflow year-over-year growth: more than three times the volume from a year earlier; Job-related data documents parsed monthly: two to three million; Daily contract mapping batch processing time (after): around 20 seconds (source-reported, not independently verified).

What failed first in this deployment?

The Community Edition became untenable at enterprise scale: credential sharing, no proper role-based access, thin monitoring, and a single master account that could not support team-level isolation—including a require…

How is this back office ops AI workflow structured?

User describes automation need → Feasibility questioning → Idea scoring → Diagram and steps generated → Claude scaffolds the workflow → User refines and learns.