The Product-Market Fit Moment: Why Anthropic and OpenAI Have Finally Crossed the Chasm

On the surface, this week's news cycle reads like a disjointed collection of headlines: the Vatican invited Anthropic to the Pope's AI encyclical presentation [2]; Cisco announced a sweeping enterprise engineering partnership with OpenAI's Codex [3]; and a prominent tech analyst declared that both Anthropic and OpenAI have finally found product-market fit [1]. These aren't isolated events. They are signal flares from a single, tectonic shift in the AI industry—the moment when frontier AI labs stopped being research curiosities and became indispensable infrastructure.

The thesis, articulated by Simon Willison in a widely circulated editorial, is deceptively simple: "I think Anthropic and OpenAI have found product-market fit" [1]. But beneath that understated headline lies a brutal, clarifying analysis of an industry that has spent three years burning billions of dollars on compute, chasing benchmarks, and struggling to articulate a coherent value proposition beyond "chatbot." The evidence—drawn from enterprise adoption, cultural legitimacy, and developer tooling—suggests the chasm has been crossed. The question now is not whether these companies have product-market fit, but what kind of fit they've achieved—and who gets left behind.

The Vatican Signal: When AI Gets a Blessing

Start with the most culturally dissonant data point. The Vatican, under Pope Leo, invited Anthropic to the presentation of the Pope's first encyclical on AI [2]. This is not a photo op. The Catholic Church, an institution that has navigated questions of moral authority and technological ethics for two millennia, chose to elevate Anthropic as a dialogue partner in the most authoritative doctrinal document a Pope can issue [2]. The Wired report frames this as "an unprecedented alliance between the Church and Silicon Valley" [2].

This is product-market fit of a specific kind: legitimacy market fit. Anthropic has positioned itself as the safety-first AI company, founded by former OpenAI employees who left over concerns about AGI development pace and direction. The company's public benefit corporation structure and its emphasis on "constitutional AI" aren't just marketing—they are the precise attributes that make the Vatican comfortable putting Anthropic in the same room as a papal encyclical [2]. No other AI company could have pulled this off. OpenAI, with its for-profit PBC structure and sometimes chaotic public persona, would have been a harder sell for the Holy See. Google DeepMind, despite its DeepMind Ethics & Society unit, lacks the narrative clarity of Anthropic's origin story.

But mainstream coverage misses a key point: the Vatican invitation is not just about ethics. It's about trust infrastructure. For enterprise buyers—banks, healthcare systems, government agencies—the question of AI adoption has always been secondary to the question of AI trustworthiness. If the Vatican, an institution that has seen empires rise and fall, is willing to engage with Anthropic at the highest doctrinal level, that sends a signal to risk-averse procurement officers everywhere. The Church's blessing is, in a very real sense, a certification of Anthropic's brand safety. That is product-market fit for a market that doesn't just want intelligence—it wants permission.

The Cisco Deal: Enterprise Engineering Gets a Copilot

If the Vatican represents the soft power of legitimacy, the Cisco-OpenAI partnership represents the hard power of operational integration. On May 27, OpenAI announced that Cisco is "redefining enterprise engineering with Codex," using the AI system to "scale AI-native development, accelerate AI Defense work, and automate defect remediation" [3]. This is not a pilot program or a proof-of-concept. Cisco, one of the largest enterprise networking and cybersecurity companies globally, is embedding Codex into its core engineering workflows [3].

The details matter. Codex, which translates natural language to code, has been available since 2021. But the Cisco announcement suggests a product maturation. Cisco uses Codex not as a toy for generating snippets, but as a tool for "AI-native development"—a phrase implying Codex is a first-class citizen in the software development lifecycle [3]. The "AI Defense work" reference is particularly telling: Cisco uses AI to build AI security products. This recursive loop validates the entire premise of AI-assisted engineering.

This is product-market fit for the developer tools market. OpenAI's API, which provides access to GPT models and Codex, has been a workhorse for startups and individual developers. But the Cisco deal signals that Codex has crossed into the enterprise tier. When a company with Cisco's engineering rigor and security requirements trusts Codex with defect remediation, the technology has moved beyond "interesting" to "indispensable."

The timing is not accidental. The MIT Technology Review report, published the same week, reveals a massive gap between enterprise ambition and execution: "Although 85% of organizations say they want to be agentic within the next three years, 76% say their current operations and infrastructure can't support that change" [4]. This is the chasm that OpenAI and Anthropic are bridging. The 76% figure represents a market of desperate buyers. They want AI agents and automation, but they lack the "people, processes, and workflows" to get there [4]. Companies like Cisco are showing that Codex can be the bridge.

The Agentic Gap: Why 76% of Enterprises Are Stuck

The MIT Technology Review piece, titled "Rethinking organizational design in the age of agentic AI," provides the macroeconomic context that makes the Anthropic and OpenAI product-market fit claims credible [4]. The data is stark: 85% of organizations want to be agentic within three years, but 76% say their current operations can't support it [4]. That's a 9-point gap between desire and readiness. In any other technology cycle, that gap would be a death sentence. In AI, it's a business opportunity.

The article identifies "a lack of readiness across people, processes, and workflows" as the primary bottleneck [4]. This is not a compute problem or a model quality problem. It's an organizational design problem. Enterprises have spent two years buying API keys and running pilot programs. They have not spent enough time restructuring their teams to accommodate AI-native workflows. The MIT Tech Review calls this "the sticky tape problem"—the challenge of integrating AI into existing organizational fabric [4].

This is where Anthropic and OpenAI diverge in their approach to product-market fit. OpenAI, through Codex and the API, sells a tool. The Cisco deal is a classic enterprise software play: embed the tool into the workflow, measure productivity gains, and expand. Anthropic, through its Vatican-level legitimacy and safety-first branding, sells a relationship. The product is not just Claude—it's the trust that comes with Claude.

Both approaches work, but they work for different segments of the 76% stuck enterprises. OpenAI targets companies that want to move fast and break things (with AI). Anthropic targets companies that want to move carefully and avoid lawsuits. The market is large enough for both.

The Open Source Shadow: What the Numbers Tell Us

No analysis of product-market fit is complete without examining the competitive landscape. Verified data from HuggingFace, the leading open-source model repository, reveals a fascinating picture. The gpt-oss-20b model has been downloaded 8,216,581 times, while the larger gpt-oss-120b model has 5,020,781 downloads. The whisper-large-v3-turbo speech recognition model has 7,948,622 downloads.

These numbers are massive. They suggest that open-source AI is not dead—it's thriving. But nuance matters: downloads do not equal deployment. The 8 million downloads of gpt-oss-20b could represent a million developers trying it once, or ten thousand developers using it daily. The data does not specify.

What the open-source numbers reveal is the ceiling on proprietary product-market fit. OpenAI and Anthropic have found product-market fit for the enterprise and legitimacy markets. But for the hobbyist, researcher, and cost-sensitive markets, open-source models remain the default. The 8 million downloads of gpt-oss-20b represent a constituency not yet converted to paid API usage. If OpenAI and Anthropic want to maintain their product-market fit as the market matures, they must either (a) make their proprietary models dramatically better than open-source alternatives, or (b) find ways to monetize the open-source ecosystem.

The OpenAI Downtime Monitor, a free tool tracking API uptime and latencies for various LLM providers, is a telling artifact. It's a freemium tool—meaning someone built a business around the fact that OpenAI's API goes down. That is both a vote of confidence (people care enough about uptime to monitor it) and a warning (reliability is not yet a solved problem).

The Hidden Risk: Product-Market Fit as a Trap

Here is what the celebratory coverage misses. Product-market fit is not a destination. It is a dynamic equilibrium that any number of forces can disrupt: regulatory action, open-source commoditization, a catastrophic model failure, or a competitor's breakthrough.

The Vatican invitation is a double-edged sword. If Anthropic's Claude ever produces a genuinely harmful output that contradicts Catholic doctrine, the Church's endorsement would amplify the reputational damage. The Cisco deal is similarly fragile. If Codex introduces a security vulnerability into Cisco's AI Defense products, the backlash would be severe.

The MIT Technology Review's 76% figure is also a warning. If enterprises cannot fix their organizational design problems, the agentic AI revolution will stall. The 85% who want to be agentic will become frustrated. They will blame the AI vendors, not themselves. Product-market fit can evaporate when the market realizes the product alone cannot solve the problem.

There is also the question of pricing. OpenAI's API pricing is listed as "Unknown." That opacity is a feature, not a bug—it allows OpenAI to charge different prices to different customers. But it also creates uncertainty. If enterprises cannot predict their AI costs, they will hesitate to scale. Anthropic's pricing is similarly opaque. The product-market fit of 2026 may look very different in 2027 if customers revolt against pricing models.

The Editorial Take: We Are Witnessing the Platformization of AI

What the Vatican invitation, the Cisco deal, and the MIT Tech Review data collectively reveal is that AI is undergoing a platformization process. In the same way cloud computing moved from "experimental" to "default infrastructure" between 2010 and 2015, AI is moving from "chatbot curiosity" to "enterprise necessity" between 2023 and 2026.

Anthropic and OpenAI have found product-market fit not because their models are perfect—they are not—but because they have solved the distribution and trust problems that plagued earlier AI companies. Anthropic has distribution through cultural legitimacy [2]. OpenAI has distribution through enterprise partnerships [3]. Both have trust through demonstrated reliability and safety positioning.

The 76% of organizations not ready for agentic AI [4] represent the next wave of growth. They will buy not just models, but solutions—consulting, integration, training, and organizational redesign. The companies that win will be those that can sell the full stack, not just the API key.

The open-source ecosystem, with its 8 million downloads, will continue as a parallel universe. It will serve researchers, tinkerers, and the cost-sensitive. But for the enterprise, the Vatican, and the Fortune 500, the product-market fit of Anthropic and OpenAI is now undeniable.

The question for the rest of the industry is simple: What is your product-market fit? The window for finding it is closing. The Vatican has already chosen its partner. Cisco has already chosen its tool. The 76% are watching, and they are ready to buy.

---

References

[1] Editorial_board — Original article — https://simonwillison.net/2026/May/27/product-market-fit/

[2] Wired — Why the Vatican Invited Anthropic to the Pope’s AI Encyclical Presentation — https://www.wired.com/story/anthropic-christopher-olah-pope-ai-encyclical/

[3] OpenAI Blog — Cisco and OpenAI redefine enterprise engineering with Codex — https://openai.com/index/cisco

[4] MIT Tech Review — Rethinking organizational design in the age of agentic AI — https://www.technologyreview.com/2026/05/26/1137584/rethinking-organizational-design-in-the-age-of-agentic-ai/