How Sembly AI built an integration engine that scales to 40+ apps without scaling headcount

2025-05-21 Horia Clement Case Studies

Goals

Ship dozens of integrations fast without growing engineering overhead
Support wide-ranging customer workflows across CRMs, docs, and task tools
Build an integration architecture that compounds value instead of adding technical debt

Highlights

40+ integrations live using unified integration logic
One authentication layer powers all apps via Membrane’s infrastructure
Reusable integration packages drive automation from Sembly to CRMs, productivity tools, and more
AI-ready architecture positioned for MCP-native workflows and bidirectional data access
New GTM opportunities unlocked through integration-based co-marketing and marketplace visibility

Company Overview

Founded: 2019
Industry: AI / Productivity SaaS
Location: Global, HQ in New York
Website: https://www.sembly.ai/

About Sembly AI

Sembly AI is a smart meeting assistant that turns spoken conversation into structured outcomes. It records, transcribes, and analyzes meetings – then transforms that data into summaries, tasks, sentiment scores, and deliverables like sales proposals and project plans.

At the heart of Sembly is Semblian 2.0, a proprietary AI engine that augments worker intelligence by generating high-quality documents from meeting context. Drawing on conversations across multiple discussions, Semblian streamlines the creation of tailored deliverables and next steps, saving hours of manual effort.

Sembly is used by teams across sales, product, operations, and HR – including in highly regulated environments requiring secure, HIPAA-compliant infrastructure. For Sembly, integrations aren’t a feature – they’re foundational to the product’s value proposition. Without deep connectivity to CRMs, task managers, and documentation tools, the AI-generated insights stay trapped in Sembly instead of flowing into the tools where work actually happens.

The Challenge: Why traditional iPaaS platforms failed

Sembly is a horizontal product used across roles, departments, and industries. That meant they needed deep integration coverage from day one, and a platform that would enable:

Support for 30+ apps across CRMs (Salesforce, HubSpot), task tools (Trello, Asana, Monday), and documentation platforms (Notion, Google Docs, Confluence)
Ability to build once and deploy across all apps without rewriting logic per integration
Modern, developer-first experience with fast iteration cycles
Scalable pricing that wouldn’t explode as they added customers or usage volume

Sembly explored traditional iPaaS platforms like Workato, but quickly discovered several limitations:

Proprietary lock-in: Integration logic lived inside the iPaaS platform, not in Sembly’s codebase. No code ownership, no portability.
Black-box infrastructure: Limited visibility into what was happening under the hood. Hard to debug, hard to customize, hard to trust in production.
Long lead times just to prototype: Couldn’t validate integrations without thousands of dollars in upfront commitment
Prohibitively expensive pricing models that scaled unpredictably with API calls and execution volume.

“Every app today needs to behave like an enterprise app due to strict data security and privacy requirements, but legacy iPaaS tools were built for yesterday’s enterprise” said Artem Koren, CPO at Sembly AI.

Sembly AI needed infrastructure that would allow them to build their own integration logic once, and adapt it to every customer’s unique requirements at scale. That’s when they found Membrane.

Why Membrane: Infrastructure for AI-powered integrations

Membrane offered something others couldn’t: infrastructure for generating and operating integrations with AI-powered tooling and full code ownership.

Instead of rigid, pre-built endpoints or proprietary platforms, Membrane provides infrastructure that lets teams build integration logic, field mappings, and workflows that work consistently across every external app – all while owning the code.

For Sembly, Membrane checked all the boxes:

Developer-first experience: Code-first packages, IDE-native, no vendor lock-in. Sembly owned the integration logic.
Fast iteration: Build, test, and deploy integrations in minutes instead of weeks.
Predictable pricing: A scalable cost model that didn’t explode with usage or API volume.
Agent-compatible infrastructure: Built to work seamlessly with AI coding agents and MCP (Model Context Protocol), positioning Sembly for AI-native workflows.
Build once, deploy everywhere: Define integration logic once and reuse it across 40+ apps with minimal code changes per app.

“Membrane gave us a modern infrastructure layer that actually works for modern tools,” says Artem. “It’s a rinse-and-repeat engine now. New app? Same logic, same flows – just different mappings. Membrane let us move fast, build fast, and ship fast.”

And critically, Membrane positioned Sembly’s integration architecture for the future: bidirectional data access, AI agents pulling context from CRMs and task tools, and MCP-native workflows that enable richer, context-aware AI experiences.

Implementation Journey: From pilot to production

Sembly’s integration journey began with in-house testing. Their first three integrations – Notion, HubSpot, and Trello – were chosen to reflect a cross-section of customer needs across CRM, task management, and documentation.

This allowed them to build a solid foundation and validate real-world behaviors before expanding. Once the architecture was in place, they ramped quickly – shipping dozens of integrations by reusing the same logic.

Universal authentication

Before Membrane, authentication was a pain point. Each new app meant writing new OAuth logic, storing and refreshing tokens, and handling revocations and user permission edge cases.

Membrane’s configurable infrastructure enabled Sembly AI to create authentication logic that would power all their integrations. This unlocked:

A one-click “Connect App” UI for all integrations, built using Membrane’s API
Auth flows fully handled by Membrane: Secure token storage, automatic refresh logic, and real-time webhook updates to reflect auth status in Sembly’s UI
One implementation, 40+ apps: Instead of 40 different OAuth flows, Sembly shipped one – and reused it everywhere

Universal flows

This is where Membrane’s infrastructure delivers its most significant advantage: the ability to define integration logic once and reuse it across every app.

Sembly’s engineers built integration packages using Membrane’s infrastructure. These packages define how data moves and transforms post-meeting – and they work across all connected apps without rewriting logic per integration.

How it works:

Ingest raw audio and transcript from Microsoft Teams, Google Meet, or Zoom
Run it through a proprietary AI engine, which generates structured summaries and action items
Route this data to connected tools: Notion, Slack, Trello, Salesforce, etc.

Another example of an implementation:

Trigger action: New meeting processed → AI summary generated
Universal flow: Push the same data to Notion, Salesforce, and Google Docs at once

Membrane’s infrastructure automatically translates Sembly’s integration logic into each platform’s unique schema, API quirks, and webhook structure – with no need for custom endpoint coding, or rewriting business logic per app.

This architectural approach unlocked non-linear scale. Adding a new integration no longer meant starting from scratch – it meant configuring Membrane’s infrastructure for a new app while reusing all the core logic Sembly had already built.

The Result: Integration architecture that fuels GTM

With Membrane, Sembly didn’t just build 40+ integrations faster – they built a scalable integration engine powered by modern AI-native infrastructure. Now, Sembly can:

Skip the auth boilerplate: One authentication layer across 40+ apps
Own their integration logic: Code-first packages they control, not vendor lock-in
Build once, deploy everywhere: customizable infrastructure for custom endpoints, flows, and field mappings

Membrane has also unlocked GTM growth through integration-based discoverability and co-sell opportunities:

Customers see integrations as a trust signal and maturity marker
Integrated apps unlock co-marketing and co-selling opportunities
Marketplace listings create organic visibility and inbound traffic

“Integrations give us credibility and distribution. They’ve become a core part of our GTM motion,” said Andrey Fediv, Growth Lead at Sembly AI. “Thanks to Membrane we are a serious contender and it helped us climb the ranks, including the Gartner Magic Quadrant report.”

What’s Next: Bidirectional integrations and MCP-native workflows

Sembly’s next phase is bidirectional integration. Instead of just pushing data out, they want to pull data in – enabling richer, context-aware AI experiences.

Here’s what’s coming:

Mark tasks as “done” in third-party apps directly from meeting wrap-ups
Use CRM data to inform follow-up summaries or AI-generated email copy
Pull deal history or ticket backlog to enrich meeting context and recommendations

With Membrane’s infrastructure, Sembly can support MCP-based workflows with minimal effort. The same integration packages that power Sembly’s current push-based workflows can be extended to pull context from external apps – enabling AI agents to work with live CRM data, project status, and customer context in real time.

Sembly isn’t just keeping up with AI trends – they’re building the integration architecture that will power the next generation of AI-native products.