Is Make.com in Trouble?
Prompt-to-workflow generation is live across five platforms. The visual builder is no longer the product
No-code automation platforms built their value on a visual canvas where users arrange modules, triggers, and actions by hand. That interface is now under structural pressure from two directions simultaneously: agentic AI platforms that execute full workflows from a single prompt, and the automation incumbents themselves, who are racing to add conversational builders on top of their own products. Google Stitch is already doing to Figma what a prompt-to-workflow product would do to Make and Zapier. Vendors and enterprise buyers who treat the visual builder as a durable moat are mispricing the speed of this transition.
Today’s Deep Dive covers:
Who Has Already Shipped Prompt-to-Workflow Automation, and What Does It Prove?
What Does the “Google Stitch for Automation” Look Like, and What’s Still Missing?
What Breaks for No-Code Platforms, and What Should Enterprise Buyers Do Now?
1. Who Has Already Shipped Prompt-to-Workflow Automation, and What Does It Prove?
Prompt-to-workflow generation is not a future capability. It is live, in production, and spreading across every tier of the automation market.
The most direct proof point is Gumloop, a Y Combinator company that closed a $50 million Series B led by Benchmark in March 2026. Gumloop’s Gummie agent takes a natural language description of a desired automation and builds the full workflow: selecting nodes, writing prompts, configuring integrations, and presenting the result on a visual canvas the user can inspect and edit. Customers including Shopify, Ramp, Instacart, and Opendoor are using it daily. During Benchmark’s due diligence, one enterprise customer reported giving employees access to Gumloop alongside two competitors; six months later, staff were using Gumloop daily while the competing tools went untouched. Benchmark general partner Everett Randle called enterprise automation “the biggest category in enterprise AI.”
Gumloop is not alone. At least five automation incumbents, including Make, Tray.ai, and GoHighLevel, have shipped their own conversational workflow builders since late 2025. The pattern is consistent: describe what you want automated in plain language, and the platform generates a workflow you can review and modify. The incumbents recognize that the visual canvas, long their core product, is becoming the output of a conversational layer rather than the primary interface.
On the general purpose agentic side, Perplexity Computer validates that full workflow orchestration from a single prompt is architecturally viable at scale. Launched February 25, 2026, Computer orchestrates 19 AI models simultaneously, connects to 400+ applications, and runs workflows autonomously for hours or days. Its annual recurring revenue grew from $80 million in late 2024 to an estimated $200 million by February 2026. OpenAI’s Codex and Anthropic’s Claude Code are proving the same thesis from the developer side. Claude Code already generates n8n workflows via an open source MCP server that documents 1,396 automation nodes. Developers are building production automations from conversational prompts today.
This follows the pattern we tracked in our March 16 digest, where 61% of IT leaders cancelled planned projects to cover AI overruns, and in our March 9 coverage, where 66% of enterprise AI budgets flowed to infrastructure rather than software. The prompt-to-workflow shift is the automation layer catching up to the same structural reallocation.
The capability is proven. The question is no longer whether AI can generate a workflow from a prompt. The question is what the output looks like, who controls it, and what the economics need to be before the visual builder becomes the secondary interface.
