Selling Intelligence Was the Warm-Up For AI – Owning Vertical Markets May Be The Play

Selling intelligence is no longer the most profitable strategy in artificial intelligence. The real opportunity lies in providing a model to a company, integrating it into their core operations, gaining insights from within, and eventually launching a competing product.

In an earlier article for StartupSuperb, the notion was suggested that the black box approach was losing importance in the boardroom, and that enterprise AI, developed according to customer specifications, represents the avenue worth pursuing. Recent months have made this argument more evident. Leading AI labs are now venturing directly into their clients’ markets, and they are doing so transparently.

The Pattern Has Hardened Into A Playbook

By reviewing the actions taken over the past year, a clear pattern emerges. The protocol involves selling models to the world’s top businesses, assisting in embedding these solutions into their critical and high-margin projects, studying the internal workings of the enterprise, and then converting those insights into a product, ultimately emerging as a competitor to the customer.

Each individual step appears to be a partnership, yet when combined, it reveals a strategic intent.

Execution remains key. Just because a model company understands how a client operates does not guarantee it can outperform that client. However, the labs maintain this is the natural trajectory of technological advancement, and there is a degree of truth to that. It also represents a reliable method to shift from a supplier-client relationship to becoming a market player.

Figma Illustrates the Trend

Figma serves as a prime example. The company collaborated with Anthropic on AI design tools, a partnership that initially seemed to validate its status as a newly public entity. However, shortly after, Mike Krieger, Anthropic’s chief product officer, stepped down from Figma’s board, and just three days later, Anthropic revealed Claude Design, a tool targeting Figma directly.

Following this news, Figma’s stock dropped by around 7%, and activist fund Findell Capital has urged the board to reassess its relationship with Anthropic.

At a private event, Figma’s Dylan Field echoed the words that OpenAI’s board used to justify its actions against Sam Altman, stating that the lab had not been consistently transparent. Thus, a partner swiftly transformed into a competitor within a single news cycle.

Harvey’s Practical Case

Harvey offers a more tangible example, as the evidence is present on Anthropic’s own website, which continues to showcase the legal startup as a customer success story in contract analysis and litigation. In May, Anthropic introduced Claude for Legal, beginning to market those workflows directly to law firms.

With an $11 billion valuation, Harvey is now contending with the same model that initially supported it as a competitor.

Repeating Patterns in Different Sectors

Similar patterns occur wherever valuable workflows exist. Intercom developed its support agents based on OpenAI’s models, and now OpenAI sells Presence, its own chat and voice support agent. Abridge and Ambience relied on OpenAI’s API for clinical documentation, and OpenAI has launched ChatGPT for Healthcare for hospitals, featuring a complimentary tier for individual clinicians.

Benchling integrated Claude into biotech R&D across over 1,300 companies, and Anthropic currently markets its research workspace, Claude Science. The same principles are emerging in financial services, where Claude interfaces with market-data providers, executing credit memos and customer verification directly within banks. In every instance, the supplier observed the customer validating the market, then stepped into it.

Microsoft’s Strategic Moves

This compelling incentive is evident in the industry’s earliest partnerships. Microsoft owns approximately 27% of OpenAI and acquired GitHub in 2018 for $7.5 billion. OpenAI is now developing a job platform for professional hiring and reportedly a code repository, moves that could extend into LinkedIn and GitHub’s territory, both owned by Microsoft.

This does not suggest that any participant has been excluded; rather, it illustrates the extent of the drive toward customer markets, even drawing against the interests of an early strategic investor.

Pharma’s Competitive Edge

Pharmaceutical discovery represents a scenario that warrant caution from every board. Companies like Novo Nordisk and Bristol Myers Squibb channel revenue and workflow data into Claude, treating Anthropic as merely a supplier. In May, BMS introduced Claude to over 30,000 employees as a shared intelligence platform for research, clinical development, and manufacturing.

On June 30, Anthropic debuted Claude Science and simultaneously announced its intentions to manage its own preclinical drug discovery programmes.

Customers, through their unexamined reliance on the model, are essentially financing what they perceive as a vendor while inadvertently allowing a competitor to emerge from the shadows, enhancing capabilities that could one day challenge their own pipelines.

The Financial Landscape Fuels Risks

Following the money reveals a clearer picture. Anthropic’s revenue run rate reached approximately $65 billion by the end of July, a dramatic rise from roughly $1 billion in December 2024, with a market valuation of $965 billion and an upcoming IPO filing. OpenAI has similarly seen its revenue double to around $40 billion.

Currently, eight of the Fortune 10 are Claude clients, indicating how significantly a single supplier now influences the economy. Selling tokens has become a low-margin commodity business, whereas positioning the finished product in a market with substantial capital returns is where the true value lies.

A firm carrying a near-trillion-dollar valuation cannot realistically vow to leave the most valuable segments to its clients indefinitely. This doesn’t imply any bad intentions; the incentive is clear, which makes such transitions more probable.

AI Narrative Develops Complexities

Here is where public perceptions diverge from operational realities, leading founders to reconcile the two. The labs promote narratives of safety, abundance, and responsible management, while in the same year, they have entered sectors such as design, law, support, healthcare, and drug discovery.

Investor Gavin Baker shared on the All-In podcast in August that he has heard from reliable sources that Dario Amodei mentioned Anthropic might someday become the only private entity, with only governments alongside it.

Anthropic outright denied this account, labelling it false, and Amodei himself refuted it. Mark Zuckerberg later published an essay advocating for open models and a distribution of power rather than its concentration.

This debate is significant, not because it has reached a consensus, but due to the unresolved nature of the discourse. Currently, no one within or outside these companies can provide clarity on how it will unfold. This narrative supports the anticipated valuation of a large IPO, but the underlying incentive and the risk it poses for those building on these platforms are undeniable.

Addressing Exposures Ahead

Alex Karp from Palantir has pinpointed the inherent risk, asserting that businesses are paying for tokens that yield no value while compromising their competitive edge. Michael Burry sharply remarked that Anthropic is effectively taking business from Palantir. The narratives may evolve, but the foundational opportunities will remain unchanged.

For India, these exposures are magnified. The country’s software companies and its 1,900-plus global capability centres are among the primary consumers of advanced APIs, deeply integrating client workflows through these models, consequently passing client advantages upstream with every incorporation. The insights drawn from the preceding examples underscore that being a substantial and steadfast customer does not offer protection; it can merely serve as reconnaissance.

The Urgency for Vertical AI Ownership

The opportunity mirrors the risk discussed earlier, which has only heightened in urgency. The focus should shift towards vertical AI, based on domain expertise the labs cannot perceive from the outside. This includes private and on-prem deployment of open models, safeguarding proprietary data within the organisation.

Utilising open-weight models that are isolated from the internet and tailored for specific enterprises is the direction forward. Intercom has already started transitioning its agent to an in-house open-weights model instead of relying on a competing supplier, establishing a clear trajectory.

India’s initiatives towards sovereign models and its vertical software developers should view this moment as the formation of a new market preference. Every Indian founder and CTO is now confronted with the critical question ahead of their next integration: precisely what alpha is being transmitted upstream, and can it be safeguarded before it emerges as someone else’s offering? While it remains uncertain if these labs will be the final survivors, firms that make the proactive choice to secure their alpha will position themselves advantageously.

The post Selling Intelligence Was the Warm-Up For AI. Owning Vertical Markets May Be The Play appeared first on StartupSuperb Media.