Figma Launches Tools to Keep AI Design Agents On-Brand

Figma Gives AI Agents a Design Rulebook

Figma, the design tool favored by startups, content teams, and product teams alike, announced a set of new features aimed at solving a familiar problem: as AI agents start designing and writing code inside the platform, how do you keep everything consistent with your brand?

The centerpiece, called Figma Specs, is a Markdown-based editor that lets teams define their design language in a simple, clear way — colors, typography, component rules, and more. This document can serve as a single source of truth for both human team members and AI agents while working on new designs. Teams can also embed canvas elements, FigJam diagrams, and code snippets directly inside it, keeping all the context centralized in one place.

At the same time, Figma moved its code layers feature — which lets you add real, functioning code directly to canvas designs — into public beta, enabling much faster interactive prototyping. Users can ask an agent to turn a design into code, or tweak existing code to refine a prototype. The feature had been in private beta since June and is now open to everyone.

Figma is also expanding support for roughly 20 additional AI tools via the MCP protocol, including Open Code, Dia, Grok Bot, and Indian startup Sarvam — allowing external AI agents to plug directly into Figma projects while following the same design rules.

Why it matters for small and medium businesses: As more design and development teams lean on AI agents to speed up work, the risk of brand inconsistency — wrong colors, mismatched fonts, incompatible components — grows. Tools like Figma Specs give smaller businesses a simple, low-cost way to maintain control over their brand identity, even when part of the work is done by AI.

Who's Actually Using Which AI Agents? A New Survey Reveals the Gaps

Separately, a new VentureBeat Intelligence survey (September 2026, 107 companies with 100+ employees) offers an interesting snapshot of the enterprise AI agent platform market: 43% of companies using an agent platform use OpenAI's, 22% use Anthropic's, and 21% use Google's.

However, when it comes to platforms companies are considering adopting in the near future, Anthropic nearly ties with OpenAI (33% vs. 30%) — meaning the competition is far from settled. Another interesting data point: according to Ramp, more businesses currently pay for Anthropic (43.8%) than for OpenAI (39.8%) in practice — though Ramp's sample skews toward smaller, more tech-forward companies.

Why it matters: For a small or medium business considering adopting an AI agent for the first time, this data shows there's no clear "winner" yet in the market — it's worth evaluating a few options before committing, especially since switching providers usually requires several days of engineering work and can temporarily hurt performance.

In Brief

Google embedded its agentic coding platform, Antigravity, directly into the Gemini app on Mac (requires an AI Ultra subscription) — a direct follow-up to yesterday's Gemini 4 Argon launch.