Smarter Workflow Suggestions

For two years, we've watched our users build the same kinds of workflows over and over. A new customer signs up — same onboarding sequence. A high-priority ticket comes in — same routing logic. A renewal date approaches — same outreach pattern.

Today we're launching something that's been six months in the making: AI agents that learn from your team's workflows and proactively suggest improvements.

What it actually does

When you connect a data source to Nimbus, our AI agent analyses your team's historical decisions — the tickets routed to engineering, the leads escalated to sales, the support cases marked as urgent. Within 72 hours, it starts surfacing suggestions like:

• "247 of your last 300 enterprise leads were routed to Sarah. Want to make her the default for company size > 500?"

• "Your team rejected 89% of refund requests over $50 last quarter. Should we add a manual approval step?"

• "Tickets tagged 'billing' resolved in 4.2 hours on average. Tickets tagged 'integration' took 18 hours. Consider splitting these queues?"

These aren't generic AI hallucinations. They're patterns extracted from your actual data, with confidence scores attached to every suggestion.

How confidence scoring works

Every suggestion comes with a percentage — that's not marketing fluff. It's a real measure of statistical confidence based on three signals:

1. Sample size — how many historical decisions back this pattern

2. Consistency — how often the pattern held without exceptions

3. Recency — how recent the supporting data is

A 94% confidence suggestion means the pattern held in 94% of similar cases. Anything below 70% gets surfaced as "might be worth exploring" rather than a direct recommendation.

We turned on AI suggestions on a Tuesday. By Friday, we'd implemented three of them. By the following week, our average ticket resolution time had dropped by 31%.
— Marcus Chen, Head of Operations at Flowpilot

What you control

We're aware AI features can feel like losing control. So three things stay 100% in your hands:

• What gets suggested. Tag any workflow as "no AI suggestions" and it's excluded entirely from analysis.

• What gets implemented. Every suggestion requires human approval. Nothing auto-applies, ever.

• What threshold matters. Set the confidence floor per workflow — 70%, 85%, 95%, whatever fits the stakes of that automation.

Available today on Plus and Pro

AI Workflow Suggestions ships today for all Plus and Pro customers. Free tier users can try it for 14 days during our beta period — grab the toggle here.

If you've been waiting for AI features that actually understand your team's context, this is it. Set it up in your workspace settings — should take about 90 seconds.

— Daniel Kim