March Release Update

This March, we shipped three major enhancements, plus a handful of quality‑of‑life improvements, all designed to make FlowRunner more powerful, more flexible, and more collaborative. Here’s what’s new:

Create Transcription Action

We’ve added a new Create Transcription action block, powered by AI, which returns a word‑for‑word transcript of any audio file in its original language.‍

‍Imagine your marketing team records weekly podcast episodes or customer interviews. With Create Transcription you can automatically generate a full text transcript as soon as the recording is uploaded. From there you can:

• Index the transcript for searchability in your knowledge base
• Auto‑generate show notes or blog posts
• Feed the text to other automations (e.g., sentiment analysis, keyword extraction)

AI Transform operation

The Transform Data block already supports over 30 built‑in transformations (filter, map, join, pivot, etc.). Now it’s been infinitely expanded with an AI Transform operation. You can supply any data payload, CSV rows, JSON objects, unstructured text, and prompt AI to reshape or enrich it however you like. There is no limit to the number of examples and the kind of transformations possible with this operation. Consider the following to give you an idea for the possibilities:

Data Normalization & Cleaning

• Use case: You receive customer addresses in wildly inconsistent formats _{(“123 Main St.”, “Main Street #123”, “123 Main”)}.
• Prompt to AI Transform:
  “Normalize address into this JSON schema: {street_number, street_name, unit, city, state, postal_code}”

• Result: A clean array of structured address objects you can feed directly into geocoding or mailing‑label automations.

Natural‑Language to Structured Query

• Use case: Non‑technical stakeholders type questions like “Show me all orders over $500 from last quarter” into a chatbox.
• Prompt to AI Transform:
  “Convert this request into a SQL query against our orders table with columns (order_id, amount, date, …).”‍

• Result: A ready‑to‑run SQL string, e.g.

SELECT * FROM orders  
WHERE amount > 500    
AND date BETWEEN '2024-01-01' AND '2024-03-31';

Multilingual Translation & Summarization

• Use case: Customer support tickets come in different languages and varying lengths.
• Prompt to AI Transform:
  “Translate support ticket into English, then summarize the main issue in one sentence.”
  ‍
• Result: A list of {original_language, translated_text, summary} objects you can route to the right team.

Product Description Generation

• Use case: You have a CSV of new SKUs with raw attributes (name, features, specs).
• Prompt to AI Transform:
  “Generate a 2‑sentence marketing blurb for each product, highlighting its top three features.”
  ‍
• Result: A table of polished descriptions ready to publish on your website.

Entity Extraction & Categorization

• Use case: You’re ingesting free‑form “notes” fields from a CRM.
• Prompt to AI Transform:
  “Extract these entities from the provided note: customer_name, product, issue_type, priority, and output as JSON.”
  ‍
• Result: Structured JSON you can automatically tag, escalate, or integrate into reporting dashboards.

Each of these examples shows how AI Transform can adapt to virtually any data challenge - turning unstructured or semi‑structured inputs into precisely formatted, actionable outputs, all within a single Transform Data block.

Granular Team Permissions

Collaboration just got easier, even on our free plan. Teams can now assign permissions for FlowRunner features at the individual level. You can grant or revoke access to:

• Edit Flow - allows a user to modify a flow’s steps and logic
• Access Flow Permissions - lets a user view or change who else can see or edit a flow
• Access Version Administration - enables a user to roll back or promote flow versions

With this change, managers and app owners can safely give teammates specific responsibilities without exposing every aspect of the workspace.

Other notable improvements

• Added drop‑down selection for array elements in the Expression Editor
• Enabled object rendering in action inputs and improved data‑type recognition in transformers
• Active tab stays open when switching between flow versions
• FlowManager now remembers the last‑selected flow and version on login
• Items like “Response” from Find Records are now correctly recognized as arrays