Quantifying AI Adoption Patterns Across Gaming Value Chains
This document supplements our AI Investment Thesis with empirical evidence from the Anthropic Economic Index, a dataset tracking AI adoption patterns across millions of economic tasks. While our thesis explains why AI value concentrates in private markets, this evidence report quantifies how AI transforms gaming workflows at the task level.
Key Finding: Gaming-adjacent tasks show a clear split: 530 tasks (48%) are highly automatable (code, testing, documentation), while 318 tasks (29%) remain human-critical (creative validation, strategic decisions). The remaining 23% are hybrid. This asymmetry — with the overall mean at 58.5% automation potential — is why "Games plus AI" beats "AI-only" plays: you need gaming expertise to know which AI outputs matter.
Gaming is a $263B annual market at an AI inflection point. Fund I's 1.8x Net TVPI demonstrates our ability to identify and ride platform transitions: the VR opportunity with Stress Level Zero (25.5x MOIC), the premium-to-mobile shift with thatgamecompany (11.1x), and the cross-platform emergence with Gardens (3.0x).
AI in gaming has passed the novelty phase. The question is no longer "can AI generate assets?" but "how do we ship AI-generated content at scale?" This shift mirrors the 2004-2008 transition when "can games be online?" became "how do we monetize live services?" Fund I was positioned for that transition; Fund III is positioned for this one.
| Metric | 2024 Baseline | Latest (2024 Surveys) | Trajectory |
|---|---|---|---|
| Studios using AI tools | 49% (GDC 2024) | 73% (a16z 2024) | +24pp |
| AI adoption in teams <20 people | N/A | 84% (a16z 2024) | AI-native baseline |
| Studios reporting >20% efficiency gains | Sporadic | 40% (a16z 2024) | Mainstream |
| AI compute cost (GPT-3.5 level) | $60/M tokens (2022) | $0.06/M tokens | 1,000x reduction |
Sources: GDC State of Game Industry 2024; a16z Games Industry Survey 2024; Stanford AI Index 2024
The Anthropic Economic Index tracks how AI capabilities map to real economic tasks. Drawing on our combined 70+ years and $10B+ lifetime revenue building and scaling gaming hits, we identified which occupations and tasks actually constitute gaming work. Starting from O*NET codes for Video Game Designers, Software Developers, Multimedia Artists, and Marketing Managers, we filtered to the 1,105 tasks that represent what gaming studios actually do day-to-day — distinguishing, for instance, between a Marketing Manager running an FMCG campaign vs. one running mobile UA at scale. The resulting set is anchored to operator experience, not just labor-market taxonomy.
Gaming-adjacent tasks analyzed across software development, art/design, marketing, and analytics occupations
The data reveals a clear pattern: AI handles repeatable execution; humans handle judgment and validation. This isn't "AI replaces humans" - it's "AI amplifies humans who know what to build."
| Task Category | Tasks | Mean Automation | Translation to Gaming |
|---|---|---|---|
| Code & Testing | 525 | 61% | Unity scripting, bug fixes, QA automation |
| Art & Animation | 360 | 59% | Asset generation, 2D/3D modeling, VFX |
| Analytics | 45 | 63% | Cohort analysis, ROAS projections, dashboards |
| Marketing & UA | 41 | 50% | Creative variants, copy, campaign analysis |
| Creative Direction | 18 | 40% | Concept art direction, narrative scaffolding* |
| Other | 116 | 49% | Project management, documentation, localization |
*Creative Direction's relatively low automation (40%) reflects its human-judgment-intensive nature. AI handles option generation (concept variations, mood boards), but the selection of which direction to pursue remains human-critical. This is the "floor vs. ceiling" split in action.
Source: Anthropic Economic Index, Release 2025-03-27, gaming occupation filter
The 530 high-automation tasks concentrate in areas where AI lowers the cost floor:
The 318 human-critical tasks concentrate where AI cannot raise the quality ceiling:
AI lowers the cost floor (cheaper to produce) but doesn't raise the quality ceiling (what "great" looks like). This is why domain expertise matters: knowing which AI outputs to ship and which to discard is the value-creating skill. "AI-only" plays lack this filter; "Games plus AI" embeds it.
Our investment thesis explains the human conflict of interest that blocks incumbent AI adoption. The Economic Index data quantifies the task-level bottleneck that makes organizational transformation so difficult.
Legacy organizations optimize for roles, not tasks. A "QA Engineer" encompasses dozens of tasks: some highly automatable (regression testing, bug documentation), others human-critical (exploratory testing, UX validation). To capture AI value, organizations must decompose roles into tasks, automate selectively, and rebuild workflows around the new split. This is organizationally traumatic.
| Traditional Role | Automatable Tasks | Human-Critical Tasks | Decomposition Difficulty |
|---|---|---|---|
| QA Engineer | Bug documentation, regression scripts, test data | Exploratory testing, UX edge cases, "feel" testing | High (threatens identity) |
| Technical Artist | Asset variants, texture baking, rigging scripts | Art direction, performance tradeoffs, style guides | Medium (clear boundaries) |
| UA Manager | Creative variants, bid reporting, cohort SQL | Budget allocation, channel strategy, CPM negotiation | High (data vs. judgment) |
| Game Designer | Documentation, level blockouts, dialogue scaffolds | Core loop design, difficulty tuning, "fun" validation | High (identity = creativity) |
We observe the decomposition problem directly in portfolio companies:
The pattern is consistent: incumbents can use AI at the tool level (ChatGPT for emails, Copilot for code) but cannot restructure workflows to capture the 10x opportunity. The conflict of interest we describe in our thesis manifests as task-level resistance.
Transcend's portfolio of 43 companies provides direct evidence of AI's impact on studio economics. While comprehensive before/after data is still accumulating as AI-native practices mature, early indicators are clear.
Before examining AI-specific evidence, it's worth establishing why Transcend can evaluate these opportunities:
| Investment | Thesis at Entry | Outcome | Relevance to AI Thesis |
|---|---|---|---|
| Stress Level Zero | VR platform inflection | 25.5x MOIC; DPI generating | Platform wave-riding |
| thatgamecompany | Premium-to-mobile transition | 11.1x MOIC; Sky became global phenomenon | Business model transformation |
| Cosmic Lounge | AI-native mobile studio | 1.6x MOIC; purpose-built-for-AI game engine | AI-first development pipeline in action |
Each investment reflects the same pattern: small, focused teams capturing disproportionate value during platform transitions. VR, mobile, cross-platform — each wave created entry points for teams that understood both the technology and the audience. AI is the current transition, and Fund III is positioned to repeat this pattern with AI-native studios.
Several portfolio companies are demonstrating early AI-native advantages. The pattern that emerges: AI compresses burn rate, not just timeline.
| Company | Fund | AI Integration | Measured Impact |
|---|---|---|---|
| WeLevel (SolidRiver) | III | Proprietary CLAI system: fine-tuned LLaMA for procedural worlds, dynamic NPCs with persistent memory; custom Stable Diffusion for concept art | 12-person team achieving output of 50+; 60-70% reduction in 3D modeling time; AI costs capped at EUR 100K/year |
| Velocity IO | III | AI-native advertising infrastructure: LLM-powered targeting, real-time generative creatives, conversational UX integration | Addressing $30B+ GenAI ad market; 95% of GenAI app users currently unmonetized |
| Kinoa | II+III | AI agents for LiveOps and monetization; platform-level automation of game operations | ARR scaled from $100K to $2M; 50-60% MoM revenue growth; UA partner to multiple portfolio companies |
| Cosmic Lounge | II | Purpose-built-for-AI game engine; AI-native asset creation pipeline | 40% reduction in asset creation time; 16% payer conversion (top-tier for mobile F2P) |
| Eggscape Entertainment | II | Claude/Unity AI coding pipeline; AI-assisted development velocity | Team building features via AI-assisted coding; self-sustaining via Google Aura porting fees |
| Fun Country (PokerSkill) | II | Superhuman-strength poker AI; AI-generated lesson and puzzle content at scale | GTO-quality AI feedback; content pipeline generates at scale what previously required expert authoring |
| Live Aware Labs | II | AI-powered player feedback: auto-transcription, sentiment detection, pain point identification from gameplay video | Feedback loop compressed from weeks to minutes; serving AAA studios (ArenaNet/NCSoft, GSC Game World) |
Fund III investments (WeLevel, Velocity, Kinoa) were selected specifically for their AI-native architectures. Fund II companies demonstrate AI adoption emerging organically across the existing portfolio.
The Margin Transformation: AI-native studios restructure the cost profile. Traditional studios require 80-120 people for $50M revenue (20-25% margin). AI-native studios target 15-25 people for equivalent output (40-50% margin). WeLevel's 12-person team, producing AAA-scope content that would traditionally require 50+, illustrates this structural shift in practice.
Sophisticated LPs ask hard questions about AI investing. Here's how we think about the key risks:
| Risk | Concern | Our View |
|---|---|---|
| IP/Legal | Steam disclosure rules, copyright uncertainty, training data provenance | Early-stage = clean-room opportunity. We evaluate tool provenance in diligence. Studios using licensed/self-trained models are defensible. |
| "Zero Marginal Cost" Flood | AI lowers barriers = more trash games = harder discovery | More content = premium on curation. Transcend's combined decades of experience identifying "fun" — and $10B+ in lifetime revenue across the team — becomes more valuable, not less. |
| AI Valuation Premium | Startups claiming "AI" command inflated valuations | We maintain discipline. AI is an enabler, not a valuation driver. Track record shows seed-stage entry (Fund I 1.8x Net TVPI proves discipline works). |
| Platform Policy Risk | Apple/Google could restrict AI content | Games-first positioning = platform-aligned. Platforms want great games regardless of how they're made. Deep platform relationships provide early signal. |
| Model Commoditization | Foundation models become free; no defensibility | "AI-only" plays lack defensibility. "Games plus AI" creates moats through proprietary gameplay data, retention loops, and IP that AI wrappers can't replicate. |
| Margin Compression | If everyone produces cheaper, market floods with "high quality" games; revenue per game drops | Discovery becomes the bottleneck, not production. This is why Transcend's curation expertise compounds: we identify which games can cut through the noise. Platform relationships (Apple features, Google Play promotions) and retention data become the scarce resources. |
The Economic Index data explains why pure AI plays struggle in gaming:
Ask: "If a better AI model releases tomorrow, does this company get better or worse?" AI-only companies get commoditized (their advantage was access). Games plus AI companies get better (same IP, cheaper production). Fund III invests in the latter.
The Economic Index data validates our thesis at the task level: AI transforms gaming workflows asymmetrically, creating opportunity for teams who can filter signal from noise. Here's why Transcend captures this opportunity:
The Anthropic Economic Index provides empirical confirmation of our thesis:
The thesis is "Games plus AI," not "AI plus games." Gaming expertise is the filter that determines which AI outputs ship. Fund III invests in teams with both.
Dataset: Anthropic Economic Index (Hugging Face). Paper: arxiv.org/abs/2503.04761. Release used: 2025-03-27. Filter applied: Gaming-adjacent occupations (Video Game Designers, Software Developers, Multimedia Artists, Marketing Managers, etc.) using O*NET occupation codes. Total tasks in filter: 1,105. Automation potential calculated as sum of feedback_loop, directive, task_iteration, validation, and learning columns; human_required as filtered column. License: CC-BY.
a16z Games Industry Survey 2024 (n=651). GDC State of Game Industry 2024 (n=2,861). Unity State of AI in Games Report 2024. Bain & Company Gaming Report 2025.
Stanford AI Index Report 2024. Epoch AI Hardware Availability and Pricing 2024. BCG AI Radar 2024.
Fund I performance: 1.8x Net TVPI (Q4 2025). Portfolio company count: 43 across Fund I/II/III. Portfolio insights from internal data systems. Incumbent examples drawn from public statements and direct board engagement.
This report supplements "Why AI Value Concentrates in Private Markets: The Structural Case for AI-Native Gaming Investments" (Fund III Investment Thesis, December 2025). That document covers the Value Multiplier Framework, human conflict of interest dynamics, and exit mechanics in detail.