The implementation gap
The Adobe-Figma merger ended with a \$1 billion termination fee after regulators in the UK and EU blocked the path forward. This decision forces Figma to defend its market position alone. As Figma rolls out AI design-to-code features, organizations face a significant implementation gap. Many companies treat AI as a procurement event rather than a capability. This mistake results in scattered efforts where one team buys a copilot and another pilots a chatbot, yet no durable wins materialize. You probably know that a working demo does not equate to a production capability. Figma Dev Mode provides specs and tokens, but it provides no usable layout code. Anima provides full layout and semantic code output, such as React, HTML, or Vue. Anima also uses a visual diff engine to track changes between design updates, which reduces hallucinations. Organizations struggle with the delivery gap, specifically around access control and identity integration. They also face issues with monitoring and incident handling. Most teams lack a predictable release process and rollback path. Scaling AI solutions also requires a support model that someone actually owns. Without these, AI work turns into a queue that waits on a few people who know how to ship safely. Companies also struggle when they confuse having data with being able to use it. Key fields remain inconsistently populated and definitions live in people’s heads, creating a direct AI bottleneck.
| Feature | Figma Dev Mode | Anima | Framer |
|---|---|---|---|
| Code Output | Specs and tokens | Full layout and semantic code | Production-ready code |
| Workflow | Inspection | Implementation | Publishing |
| AI Maturity | Limited to specs | Component-aware | Mature production tools |
Vulnerabilities in agentic workflows
Model Context Protocol (MCP) servers create new pathways for attackers to impersonate trusted AI agents. A high-severity command injection bug in the Framelink MCP server, known as CVE-2025-53967, allows attackers to execute arbitrary system commands through the get_figma_data tool. This flaw exists because the server constructs shell commands using unvalidated user input. The application’s backend code takes data directly from the API parameter and executes it as part of a shell command. Security teams must navigate these risks as 56% of organizations view generative AI as a growing security concern. There are already more than 15,000 MCP servers in the world, and hundreds of them suffer from misconfigurations that allow for sensitive data leakage. A scan of 2,000 MCP servers revealed that nearly all of them lacked any authentication or access controls. Precision remains another major hurdle for AI-driven design. Figma’s security agents saw only 15% of findings as valid during their first week of deployment. Will these low precision rates eventually force a freeze on all AI work in the enterprise?
Competitive pressure and technical limits
Figma’s revenue grew 40% to \$303.8 million in the fourth quarter last year, but its AI strategy faces pressure from specialized tools. While weekly users of Figma Make increased 70% from the previous quarter, the tool still hits limits compared to established platforms. More than half of the 1,400-plus customers who spend \$100,000 or more annually on Figma’s core design software use Figma Make every week. Framer remains a more mature choice for production websites because it includes a built-in CMS and global CDN. Figma Sites, currently in open beta, lacks advanced CMS features and granular SEO controls. The failure to deliver usable layout code through Figma Dev Mode means engineers must still manually translate design specifications into functional components, which remains a significant bottleneck for teams relying on AI for rapid production. Adobe also faces scrutiny, as the US Federal Trade Commission investigates its subscription cancellation policies. This investigation could leave Adobe exposed to significant monetary costs or penalties.
