Early in an AI company's life, founders often imagine building both — a platform API others build on, and an end-user application directly. It's tempting and rarely correct. The strategic decision between platform and product AI shapes everything: team composition, go-to-market, pricing, capital needs, exit profile. This post is the strategic fork and what each path demands.
What platform AI means
Sells APIs, SDKs, infrastructure, tools to developers. Buyers are technical — engineers, data teams, product engineers at other companies.
Examples: OpenAI, Anthropic, Google (Gemini), Pinecone, Weights & Biases, LangChain, Mistral, Cohere. Different shapes, same basic pattern.
Pricing: usage-based (tokens, API calls, compute). Customer pays for what they use.
Go-to-market: developer-first. Documentation, SDKs, community, content marketing. Sales for large deployments but self-serve for volume.
What product AI means
Sells applications to end users. Buyers are business leaders or operators — heads of sales, marketing, finance, operations.
Examples: Harvey (legal), Glean (enterprise search), Abridge (medical transcription), Hippocratic AI, Perplexity (consumer), ElevenLabs (voice).
Pricing: seats, tiers, outcome-based. Customer pays for value delivered, not compute consumed.
Go-to-market: business-buyer focused. Sales teams, partnerships, category leadership, customer success. Heavier lift than pure developer product.
What each path demands
Platform demands deep engineering culture. Reliability, developer experience, API design, documentation, scaling. Engineering leadership is the core.
Product demands deep domain expertise. Understanding the user's job, workflow, existing tools. Domain experts as co-founders or early hires.
Platform demands capital efficiency at scale. Margin comes from volume over high-fixed-cost infrastructure. Path to breakeven clear; path to hyperscale capital-intensive.
Product demands go-to-market investment. Sales, marketing, customer success proportional to revenue. Scales differently than platform.
Strategic implications
Team composition. Platform: mostly engineers. Product: engineers + product + sales + CS.
Investor fit. Platform VCs vs product VCs look for different signals. Knowing your category helps pitches.
Exit profile. Platform companies often acquired for technology or customer base. Product companies often acquired for revenue or category leadership.
Network effects. Platform has developer ecosystem effects. Product has customer-data network effects. Different moats.
Why hybrids are hard
Different cultures. Developer products and business products need different instincts in product, engineering, marketing. Hard to do both well.
Different GTM. Developer sales and business sales motions are different. Trying to run both splits resources and expertise.
Platform risk from product ambitions. If you build a product on your own platform, customers on your platform may see you as competitive. Stunts platform growth.
Examples of successful hybrids are rare. OpenAI has ChatGPT on top of API; Google has Gemini API and consumer products. These are exceptional and took years to balance.
When to pick which
Pick platform if: founders are engineering-first; capital is abundant; category is developer-scalable; long time horizon OK.
Pick product if: founders have deep domain expertise; faster revenue needed; enterprise buyer is accessible; target customer is a business function.
Pick hybrid only if: truly unique distribution advantages; patient capital; team can cover both. Exceptional circumstances only.
Evolution paths
Product companies sometimes evolve to platforms. Successful product might open up APIs for ecosystem. Salesforce did this; Stripe did this for payments but started as a platform.
Platforms sometimes launch products. OpenAI launching ChatGPT; AWS launching services on top of primitives. Usually from established platform position.
Timing matters. Evolution too early dilutes; evolution at maturity can accelerate.
Implications for AI specifically
Platform providers face commoditization. Foundation model pricing has dropped 10x+ since 2022. Platform providers competing hard on price and capability.
Product providers face platform risk. Your key capabilities may be what foundation providers offer as base features later. Build moats beyond model capability.
Vertical specialization helps product companies. Deep in one vertical where you have data and workflow expertise that foundation providers don't.
See vertical vs horizontal post for further on vertical specialization as a moat.
Closing
Pick with clarity early. Many founders try to keep both options open; the result is usually suboptimal in both. Commit, execute, revisit later if evolution path opens.