Introducing multi-branch logic: if/else workflows that handle every edge case

The single most limiting thing about linear automation is that the real world isn't linear. Leads have different scores. Invoices have different types. Support tickets have different severities. A workflow that treats all inputs identically will always require human intervention for the cases that don't fit the mold. Multi-branch logic is our answer to that problem — and it changes what's possible with Flowpath in a fundamental way.

How it works

The Branch node is a new canvas element that you drop between any two steps in your workflow. When a run reaches a Branch node, it evaluates your condition and follows the appropriate path. The condition can reference any data field from any previous step — enrichment data, AI step outputs, form fields, API responses, anything that's been produced earlier in the workflow.

Operators include: equals, does not equal, contains, does not contain, is empty, is not empty, greater than, less than, and is between. You can combine multiple conditions within a single branch using AND/OR logic — so you might say "company_size is between 50 and 500 AND funding_stage equals Series A or Series B."

Each condition creates a named path on the canvas — a branch that continues independently from that point. You can add as many subsequent steps as you need on each path, including additional Branch nodes for nested logic.

Sequential vs parallel execution

By default, branches execute sequentially — the workflow finishes one path before evaluating the next. But for cases where you want multiple things to happen at the same time, you can enable parallel execution on any Branch node. In parallel mode, all branch paths that match the condition execute simultaneously, and the workflow continues to the next shared step only after all parallel branches have completed.

This is particularly useful for notification workflows. Instead of notifying Slack and then creating a HubSpot record (sequential, ~0.8s total), you can do both at once (~0.4s total). For high-volume workflows processing thousands of runs per day, that latency saving adds up.

Score-based thresholds

We've built score-based thresholds as a first-class feature rather than a manual condition configuration. When your previous step is an AI scoring step, the Branch node automatically detects the numeric output and offers a threshold interface — you set your cutoff points and the paths are labeled automatically. This is designed specifically for the lead scoring use case but works for any numeric output — invoice amounts, sentiment scores, engagement rates, error counts.

Visibility and debugging

Every Branch node in the run logs panel shows a breakdown of how many runs went through each path over your selected time period, expressed as both a count and a percentage. If you set up a lead scoring workflow and discover that 94% of leads are going to the "not qualified" path, that's a signal your ICP definition or your scoring prompt needs adjustment — and you'll see it immediately without digging through individual run logs.

What this enables

The workflows that were previously impossible or required workarounds are now straightforward. A single invoice processing agent that handles standard invoices, amended invoices and credit notes — three paths, one canvas. A support triage agent that escalates critical tickets immediately, acknowledges standard tickets within an hour and batches low-priority tickets for daily review. A content pipeline that routes long-form articles to one reviewer, short-form posts to another and anything flagged as sensitive to legal review. These aren't hypothetical — they're workflows our customers built in the first 48 hours after we shipped branching.

Multi-branch logic is available on Growth and Enterprise plans today. Starter plan users can view workflows that use branching but cannot add Branch nodes to new workflows.