Perly Consulting │ Beck Eco

The State of Play

A living index of AI adoption across industries — where established practice meets the bleeding edge
UPDATED DAILY

The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.

The Daily Dispatch

A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.

AI Maturity by Domain

Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail

DOMAIN
BLEEDING EDGEESTABLISHED

Wireframe generation & design-to-code conversion

BLEEDING EDGE

TRAJECTORY

Stalled

AI that generates wireframes and prototypes from descriptions and converts designs into production code. Includes text-to-wireframe tools and Figma-to-code conversion; distinct from design system generation which creates reusable component libraries rather than individual screens.

OVERVIEW

AI-powered wireframe generation and design-to-code conversion compress prototyping from days to hours, but the practice remains firmly bleeding-edge: powerful for rapid ideation and early validation, but production handoff remains constrained by design system integration, accessibility compliance, and code quality governance. The ecosystem is consolidating rapidly — Anthropic entered the space in April 2026 with Claude Design integrating handoff to Claude Code, Figma deepened bidirectional MCP workflows enabling code-design synchronization, and Vercel invested heavily in v0 as a production platform. Designer adoption is mainstream (92% report faster workflows, 91% improved quality), and 92% of developers use AI coding tools monthly. The defining tension persists unchanged: generated code requires 20-30% rework for production, 92% of AI codebases contain vulnerabilities, and core architectural problems (design-system preservation, component reuse, accessibility) remain unsolved at scale. Teams achieving success front-load precision — explicit design tokens, component naming conventions, and metadata discipline — rather than free-form prompting. Designers are adopting enthusiastically; production scaling at enterprise remains architecturally constrained.

CURRENT LANDSCAPE

Vendor ecosystem consolidation accelerated in April 2026 with Anthropic's launch of Claude Design, integrating design ideation and code generation within a unified product family. Figma continues deepening platform lock-in with bidirectional MCP workflows enabling teams to push rendered UIs back to design files, while Vercel's rebuilt v0 platform now blocks 100k+ insecure deployments and has driven 100M+ user interactions with reported $250M investment. Backend standardization around Claude Opus (Lovable, Bolt, v0) continues with 20% error reduction over prior models. Market forecasts project growth from $5.5B (2024) to $18.16B by 2030 (21.9% CAGR), with Locofy Lightning (one-click Figma-to-code) identified as category-defining innovation.

Developer adoption is mainstream: 92% of US developers use AI coding tools monthly with MIT Technology Review recognition of vibe coding as 2026 breakthrough technology. Designer adoption remains strong: 89% report faster workflows, 91% improved quality. However, systematic code quality barriers are well-documented: 92% of AI codebases contain critical vulnerabilities, code duplication is 4x higher than human-written code, and accessibility failures are systemic (AI-generated React defaults to inaccessible div soup without component abstraction). Teams using v0/shadcn or Figma Code Connect paired with design system enforcement achieve 90% pixel-perfect accuracy; teams generating without structured component constraints see visual drift, component debt, and governance erosion within weeks. Agency testing reports successful prototyping speed (14 days to 4 days documented) but identifies persistent production handoff failures: design-to-code tools ignore existing design system components, requiring 90-minute rebuilds using actual components. The "design system is the API" insight has become central: output quality depends entirely on team metadata discipline (explicit annotations, component properties, design tokens). Enterprise deployments remain confined to prototyping; production scaling blocked by accessibility compliance gaps, design system semantics, and code maintainability concerns.

TIER HISTORY

ResearchJan-2023 → Jan-2023
Bleeding EdgeJan-2023 → present

EVIDENCE (97)

— Ecosystem segmentation analysis: Figma Make outputs developer-ready component code vs Base44 outputs deployed full-stack apps; clarifies that design-to-code requires developer while prompt-to-product does not, defining market bifurcation.

— Critical assessment documenting design system drift in Claude Design, security vulnerability in Lovable (Broken Object Level Authorization), stock market reaction (Figma down 55% YTD to $16.69 all-time low), and realistic limitations of AI-generated design handoff.

— Analysis of AI productivity paradox: code generation tools increase perceived speed but degrade code quality (churn +73%, security flaws in 29% of code); METR randomized trial shows 39-point perception gap between actual performance and developer belief.

— Practitioner workflow guide documenting real Claude Code + Figma adoption pattern, key enablers (iterative loops, context windows, MCP connectors), and human-in-the-loop requirements for production handoff.

— Five-way tool comparison with named early-adopter case studies (Brilliant, Datadog claiming 10× productivity improvements), market data (Figma stock -7%, Lovable $400M ARR), and documented failure modes (METR study showing senior developers 19% slower with AI).

— Official AWS case study documenting v0 reaching 4 million users and generating production-ready React/Tailwind/shadcn UI code from natural language prompts with integration to Amazon Bedrock and Vercel CDN.

— Design systems expert documents specific failures of design-to-code tools (Figma Make, Claude Code, Claude Design) and proposes Component.md specification as source-of-truth layer between Figma and code generation.

— Hands-on feature evaluation showing Figma shipped more AI in 6 months than 3 years prior; MCP Server identified as most valuable, First Draft saves 30–60 min, while Make varies wildly and Figma Make not production-ready for complex apps.

HISTORY

  • 2023-H1: First-generation design-to-code tools (Locofy, Uizard, Anima) gain traction among non-professional designers. Peer-reviewed research confirms LLM code generation works well for data tasks but struggles with visual-graphical constraints. Indie developer case study shows 70% time savings from design-to-code conversion. Critical assessments highlight workflow limitations including design-development misalignment.
  • 2023-H2: Wireframe generation research advances with WireGen prototype achieving 77.5% improvement over baselines through LLM-based generation from natural language. Second named case study (Mealcraft, UK) validates production Figma-to-code deployments with 70% time savings. Figma reveals internal constraints: only 3% of developers highly trust AI code accuracy; full automation abandoned in favor of "intelligence amplification" approach due to framework diversity and code usability issues. Practice remains dependent on substantial manual refinement despite time savings.
  • 2024-Q1: Multimodal LLMs (GPT-4V, Gemini Vision) show capability gains: Design2Code benchmark shows 49% human parity and 64% superiority over reference code on 484 webpages. Research addresses layout preservation bottleneck (LaTCoder: 66% improvement with chain-of-thought). Vendor ecosystem expands (Builder.io Visual Copilot, Uizard Autodesigner). However, Figma user survey reveals gap between expectations (89% expect impact) and adoption reality (72% report AI plays minor role, <33% proud of shipped features). Deployment concentrated among indie developers; enterprise adoption remains blocked by design-system integration, accessibility verification, and maintainability concerns.
  • 2024-Q2: Dedicated platform adoption accelerates: Uizard reaches 3.2M users with 80% of new UIs AI-generated. Figma doubles down with native AI (text-to-layout) and Dev Mode enhancements at Config 2024. Independent evaluations cement maturity messaging while documenting continued limitations: bachelor's thesis on five leading tools (Locofy, Anima, etc.) shows responsiveness and advanced feature gaps; industry review of 50+ tools confirms speed/ideation gains but persistent generic output and design system integration challenges. Vendor ecosystem solidifies around Locofy, Anima, Uizard, Builder.io, TeleportHQ. Gap between marketing (50-80% time savings) and reality (significant post-generation refinement required) persists, but confidence growing among SMBs and indie developers. Enterprise integration remains blocked by design system semantics, accessibility verification, and code maintainability.
  • 2024-Q3: Vendor ecosystem expands with Visily launching sketch-to-wireframe and template-driven design generation. Practitioner testing confirms sustained pattern: tools generate prototypes rapidly but require extensive refinement for production use. Accessibility becomes documented blocker—expert analysis shows AI-generated code systematically fails WCAG 2.2 AA compliance. Localization gaps persist, particularly for non-Latin languages. Figma-to-code plugin competition intensifies (DhiWise vs. Locofy). Practice stabilizes as mainstream for SMBs and freelancers seeking design acceleration but remains blocked for enterprise by accessibility verification, design system preservation, and code quality concerns.
  • 2024-Q4: Uizard sustains 931K monthly traffic with 10.9% growth from Autodesigner feature; Figma deepens native AI integration (text-to-layout, Dev Mode, Code Connect). Industry-wide adoption metrics peak: 98% of 400+ U.S. designers report AI-changed workflows with 91% positive ROI. Builder.io and platform vendors document persistent tension between rapid generation and code quality; LLM research identifies seven categories of non-syntactic errors in generated code. Ecosystem consolidates around core vendors with differentiation on language support and component generation; enterprise adoption remains blocked by accessibility compliance gaps, design system preservation, and code maintainability concerns.
  • 2025-Q1: Figma launches Code Connect in beta, formalizing design-to-code as platform feature. Builder.io reports 1M Figma plugin installs with enterprise production deployments. However, adoption-trust paradox deepens: 90% of developers use AI code generation but only 3% maintain high trust. Code quality barriers intensify—duplicated code eightfold higher, security vulnerabilities prevalent, and 66% of developers spend more time debugging than automation saved. Real-world enterprise case emerges (Emaar Properties, UAE using Uizard), but constrained by extensive post-generation refinement. Enterprise scaling remains blocked by accessibility, design system preservation, and code maintainability concerns; practice consolidates as mainstream for SMB and freelance design acceleration.
  • 2025-Q2: Figma Config 2025 announces Figma Sites and Figma Make, deepening platform consolidation of design-to-code workflows. Independent practitioner testing confirms v0.dev, Lovable, Bolt, and Replit accelerate prototyping but require heavy manual support; none approach full automation. Fundamental technical limitations emerge: LiveCodeBench Pro benchmark reveals frontier AI models achieve 0% accuracy on hard coding problems and 53% on medium tasks, while practitioners identify critical context gaps (project history, implicit knowledge, temporal memory) blocking autonomous code generation. Enterprise deployment remains confined to prototyping; production scaling blocked by state management complexity, design system preservation, and developer skepticism.
  • 2025-Q3: Designer-developer trust divergence sharpens: only 32% of designers trust AI design-to-code outputs vs 82% of developers using AI tools generally; 42% of companies abandon AI initiatives. UX team adoption rises to 75% for wireframing tasks, but developer sentiment on generated code quality deteriorates—trust falls to 29% among 49,000+ surveyed developers, with 66% spending more time debugging than saving. MIT analysis documents systemic enterprise pilot failure: 95% of 300 analyzed deployments deliver minimal ROI. Independent tooling evaluations (Locofy, Anima, v0, UXPin) confirm tools accelerate prototyping but fall short of production readiness.
  • 2025-Q4: Figma launches native AI wireframe generator within Figma Make, consolidating design-to-code capability into platform. Figma reports 34% of users shipped generative AI applications (up from 22%), but trust remains low: only 32% of designers trust AI outputs vs 68% of developers. Independent 30-day tool testing confirms value for prototyping (3-5 hours reduced to 30-60 minutes) but all tools require extensive cleanup (20-30% post-generation rework). Designer-developer adoption gap widens to 28 points (31% designers vs 59% developers). Practice consolidates as mainstream for SMB wireframing but production scaling remains blocked by trust deficits, accessibility compliance, and design system preservation challenges.
  • 2026-Jan: Vercel accelerates v0 with AWS database integration (Aurora, DSQL, DynamoDB) enabling full-stack app generation; reports agentic pipeline improvements (dynamic prompts, LLM Suspense, autofixers) achieve double-digit success rate gains. Locofy Figma plugin sustains adoption growth (2,192 users Jan 1 to 4,788+ by March). However, project failure rates dominate industry discourse (88% AI agent failures per HyperSense analysis), with RAND/Gartner confirming 80% never reach production. Independent v0 testing identifies trade-offs: React-only output and credit-based pricing unpredictability limit adoption. "Last mile" problem crystallizes—design system preservation and component reuse remain fundamentally unsolved at scale. Practice consolidates as mainstream for SMB/freelance/startup wireframing but enterprise production scaling remains blocked by design system semantics, component duplication, and hidden post-generation refinement costs.
  • 2026-Feb: Figma releases bidirectional integrations (GitHub Copilot MCP server, Codex-to-Figma) enabling code-design workflows; v0 rebuilds as production platform with GitHub integration and database support. AI backend standardization around Claude Opus with reported 20% error reduction. Designer adoption peaks (89% faster workflows, 91% improved designs); video-to-code emerges (Replay showing 40→4 hour legacy modernization gains). Critical risk surfaced: high-fidelity AI outputs may bypass validation phases. Platform consolidation accelerates but "last mile" design system and component reuse challenges persist unchanged.
  • 2026-Mar: Bidirectional design-code workflows operationalized at scale. OpenAI+Figma MCP integration (Feb 26) enables AI to reference design context directly; GitHub Copilot MCP now syncs rendered UIs back to Figma as editable frames, closing the handoff loop. Figma AI wireframe adoption doubled YoY. Independent comparative testing (Mark Rosal) shows tool diversity: Claude achieves pixel-perfect accuracy with semantic preservation; ChatGPT infers broader UX intent; Cursor prioritizes speed. Case study evidence emerges: Tortuga founder documented eliminating design handoff entirely, opening Figma only twice for single product (vs every screen previously), shifting to code-first Claude workflows. Industry reports document 30-60% frontend dev time reduction across leading tools (Anima, Locofy, Builder.io); 68% of developers now use AI code generation daily. Trust remains divergent: designer confidence high (89% faster, 91% improved quality) but developer skepticism persists (29% high trust in generated code quality). "Last mile" challenges—design system reuse, component duplication, semantic HTML gaps, accessibility—unchanged despite workflow streamlining.
  • 2026-Apr: Research breakthroughs and production deployment maturity signals coexist with a sharper critique of systemic quality failures. ICLR benchmark (Figma2Code) reveals fundamental trade-off in multimodal design-to-code: proprietary models (GPT-5, Gemini 2.5 Pro) achieve visual fidelity but generate rigid, unmaintainable code; open-source models generate cleaner, responsive output due to layout-aware generation. Anthropic launches Claude Design (April 2026), integrating design ideation with handoff to Claude Code in a unified product family. Figma GA of use_figma MCP tool with write access enables design system-driven code generation; named enterprise adoption: Uber's uSpec system automates component spec creation across seven implementation stacks. Technical research: DOne framework (schema-guided generation) achieves 3x faster code generation and 10% better design fidelity. Critical failure modes newly documented: UXPin analysis names Figma, Cursor, Claude Design, Bolt, v0, and Lovable as all exhibiting visual drift and component debt when disconnected from design systems; Smashing Magazine documents that 92% of AI codebases contain critical vulnerabilities, code duplication runs 4x higher than human-written code, and role creep forces designers to master engineering simultaneously—labelled the "Rework Tax" draining engineering resources. UXMagic 2026 tool survey identifies the "Div Soup Problem" (semantic HTML failures) as a systemic barrier across 12 leading tools. Real-world testing: Workspace agency finds design-to-code tools ignore existing 200+ component libraries, requiring 90-minute rebuilds; Builder.io confirms agent accuracy depends entirely on team metadata discipline. v0 blocked 100k+ insecure deployments with 100M+ user interactions signalling mainstream but constrained adoption. Design system quality remains the binding constraint—"Design system is the API that allows AI to build your product safely." Designer-developer sentiment divergence persists: 89% of designers report faster workflows (91% improved quality); only 32% trust AI outputs vs 68% of developers. "Last mile" unsolved at scale: design system preservation, component reuse architecture, semantic HTML, and accessibility compliance remain primary barriers to enterprise production scaling.
  • 2026-May: Market consolidation accelerates with critical credibility signals mixed with documented risks. v0 adoption confirmed at 4M users with AWS case study validating production deployment of React UIs via Vercel CDN and Amazon Bedrock. Figma stock crashes 55% YTD from IPO high ($142.92 to $16.69), signalling market skepticism on design-to-code ROI despite AI feature acceleration (MCP Server noted as most valuable innovation). Lovable crosses $400M ARR but Broken Object Level Authorization vulnerability (May 2026) exposes 1M+ projects' source code, databases, and credentials from user accounts created before November 2025. METR randomized controlled trial documents 39-point perception gap: developers believe AI tools make them 20% faster but actual testing shows 4% slowdown; GitClear analysis reveals code churn jumped from 3.3% baseline to 5.7–7.1% during AI adoption. Practitioner consensus crystallizes around two insights: (1) Design-to-code requires disciplined component metadata (Component.md proposal) and iterative refinement via conversation, not one-shot export; (2) market bifurcates into design-to-code (requires developer expertise) vs prompt-to-product (no developer needed). Figma AI features assessment shows MCP Server as breakthrough (design-to-code handoff via MCP), while Figma Make generates unreliable complex applications and First Draft saves 30-60 minutes on scaffolding. Claude Code + Figma MCP workflow adoption documented but demands full-day expert effort for architecture/styles/business logic. Designer confidence remains high (89% faster workflows, 91% quality improvement); developer skepticism persists despite mainstream tool adoption. Production scaling barriers unchanged: design system preservation, component reuse, accessibility, and maintainability remain primary blockers at enterprise scale.

TOOLS