Post-SaaS Abiogenesis: How Businesses Grow Their Own Software

A few weeks ago I asked Claude to explain to me what workflow automation platforms are, and what kinds of problems businesses solve using them.

A workflow automation platform is a surface of a kind of post-SaaS abiogenesis — individual nodes, such as Gmail or Airtable, are glued together by primitive nodes, such as Human Input or Data Transformation, then combined into reusable segments (templates) and eventually into complete workflows that solve specific business problems.

The very existence of such platforms represents a fascinating phenomenon: an attempt by businesses to adapt software to their own processes, rather than adapting their business processes to existing CRM/ERP solutions. It is a space where the market itself says exactly what it needs.

I found this intriguing — and it also presented an opportunity to run an experiment: “write a piece software exclusively using an LLM”, without ever looking at the code, not even for a code review. The result is an SQL RAG combined with a chatbot, in which the LLM provider is used solely for intent classification and answer formatting.

Before jumping into the insights, a few words about methodology.

I pulled data from 9 automation platforms (Activepieces, FlowFuse, IFTTT, Make, n8n, Pipedream, Relay.app, Windmill, Zapier) and analyzed how often each integration appears in templates. Basic building blocks (flow control, schedules, helpers) were excluded. One caveat: an integration appearing in templates doesn’t necessarily mean people use it heavily. To partly account for this, anything appearing on fewer than 3 platforms was dropped.

You can chat with Clanker via the chatbot interface.

Total, absolute dominance of Google services

Google Calendar, Gmail, Google Drive and Google Sheets are in the top 7 integrations across all platforms. Often, a workflow starts and ends in a Google service.

If I were planning to enter this market, I would take care to make my APIs easy to integrate with Google services, without requiring intermediate data transformation nodes. In other words, keep onboarding simple.

Another use of this insight is intent SEO and marketing channels: queries like “How do I turn my Google Sheet into X?” should lead to your sales page.

IFTTT is still the consumer remote control

20 moats, all consumer or device-flavored. Ten years on, IFTTT’s moat is the same one it had at launch: the phone in your pocket and the things in your living room.

Zapier owns the marketing/funnel stack

20 moats, almost all SMB revenue tooling: LeadConnector (624 templates), Klaviyo, Close, Microsoft Dynamics 365, Microsoft Excel and Office 365, Kajabi, Clio, the entire web-form long tail (Squarespace Forms, Unbounce, Leadpages, WPForms), and a paid-ads cluster (Facebook Conversions, TikTok Lead Generation, LinkedIn Ads). Zapier is where SMBs wire their funnel.

n8n is for developers and AI primitives

20 moats split cleanly into two camps: databases and infra (Postgres, Supabase, Redis, MongoDB, NocoDB, S3, GraphQL) and AI agent building blocks (Information Extractor, Text Classifier, Sentiment Analysis, Reranker Cohere, Ollama Model, plus tool-shaped wrappers around Wikipedia and SerpApi). The shape says n8n is what you reach for when the brief is “Zapier-but-with-Postgres” or “Zapier-but-with-an-LLM-loop.”

Make is the long-tail business connector

Operational SaaS the others didn’t bother with: Pipedrive CRM, QuickBooks, Zoom Administration, Teamleader, Bitrix24, plus niche picks like Go4Clients, WhatsScale, DataForSEO, Koncile OCR, BrowserAct. Make also owns two pieces of native plumbing — the RSS module and a built-in Data store — table-stakes infrastructure it ships natively, while peers route users to third-party Airtable.

Possible opportunities

Unfortunately, I haven’t come up with a good business idea from the plain data so far. A promising direction is probably the principle of compressing complexity. Besides the platform itself, every single node represents a separate failure point, and sometimes the connections between them do too.

Complex setup, lack of observability, error-prone flows, many different integrations per workflow, pay-per-step or per-volume pricing — all this together solves the user’s business flow while at the same time creating another huge pain point. End users want “something simple, reliable, all in one place.”

Niche SaaS products that provide all the benefits of custom workflows for a small, narrow set of customers, but without the pains of workflow automation, seem like the right direction to me.

I welcome anyone to talk with my automation robot — he is friendly and absolutely free. https://crux.do