← kevin.energy Essay · No. 01

Strategic Essay

Can Anthropic escape the commoditization trap?

Customer support is a $95B market where the answer plays out in the next 18–24 months — and why moving up the stack is Anthropic's make-or-break move before the window closes.

Kevin Nguyen · February 2026 · 8 min

Anthropic makes Claude — the AI brain powering customer support, code generation, and research for 300,000+ businesses. But what happens when every AI brain becomes equally smart? Every infrastructure company faces this moment eventually: the technology matures, benchmarks converge, and the model stops being the differentiator. What matters then is who owns the workflow, the data, and the customer relationship. For Anthropic, that moment is not abstract. The four-phase commoditization clock is already running — and the clearest escape route is also the largest proven AI use case in enterprise software.

This essay works backward from the economics to the strategy. It does not argue that staying in infrastructure is wrong in principle — AWS proved it can work. It argues that for Anthropic specifically, at this moment specifically, the application layer is where the leverage lives, and customer support is where the entry point is sharpest.

§ 01 — The fork

The strategic choice every infrastructure company faces

Infrastructure companies face the same question when their technology matures: stay invisible under the hood, or own the product that customers actually use? The historical record is specific. AWS chose infrastructure — and won, because cloud computing kept getting more complex. More complexity means more value in the plumbing. Stripe chose payments infrastructure — and watched vertical SaaS companies capture the customer relationships above it. Twilio chose communications APIs — then watched its stock fall 80% from peak as AI made its APIs interchangeable.

The pattern: infrastructure companies that do not move up the stack get commoditized when the technology matures. For AI, the maturation is happening faster than anyone expected — because every competitor, OpenAI, Google, and Anthropic, is racing toward the same benchmark finish line.

When Claude, GPT, and Gemini all score similarly on benchmarks, the model stops being the differentiator. What matters is who owns the workflow, the data, and the customer relationship.

Anthropic is a brain surgeon — the best in the world. But right now, other companies are hiring the surgeon, putting their own name on the practice, and keeping the patient relationship. Anthropic does the hard work; everyone else captures the value.

§ 02 — The gap

$0.03 versus $10 — where the value actually lands

The economics of the current arrangement are not abstract. Per customer interaction, Anthropic captures roughly $0.03 in API token revenue. Application-layer companies — Sierra, Decagon — capture $10 per resolution on the same interaction. That is a 300x gap, and it compounds with every passing quarter that the application layer deepens its moat.

$0.03 API LAYER 300× $10 APP LAYER
FIG_001 — Revenue per customer interaction, API vs. application layer. © 2026 Anthropic provides the intelligence but captures pennies. Application-layer companies like Sierra and Decagon capture dollars on the same interaction — and their switching costs rise with every ticket processed.

The table is stark on four dimensions that matter most to long-run competitive position:

Revenue per interaction
API layer: $0.03   |   App layer: $10
300× gap
Switching costs
API layer: low — models are easy to swap   |   App layer: high — deep workflow integration
Structural disadvantage at the API level
Pricing power
API layer: declining   |   App layer: increasing
Diverging trajectories
Data ownership
API layer: none   |   App layer: full customer interaction history
The compounding asset Anthropic does not currently own
FIG_002 — Competitive position, API layer vs. application layer. © 2026 Every row favors the application layer — and the data ownership row is the one that compounds. An AI agent that has processed two million of a company's tickets is extraordinarily hard to replace.

§ 03 — The market

Why customer support is the right entry point

There are dozens of enterprise categories Anthropic could pursue. Customer support is the right one for compounding reasons. It is a $50 billion market heading to $95 billion by 2031. These are not projections based on hype — enterprise buyers are already writing seven-figure checks for AI customer support at proven scale:

Anthropic does not need to educate the market. It needs to redirect spending that is already flowing — toward a segment with a specific underserved pocket: healthcare, financial services, insurance, and government customer support is roughly 40% of the total market, and the most underserved by AI-native startups like Sierra and Decagon, whose customer bases skew toward internet-native companies (Notion, Duolingo, Eventbrite).

And before January 2026, Anthropic had no credible path to the application layer without a multi-year product build. The Cowork plugin architecture changes this: a CS plugin is a set of markdown skill files and MCP connectors. Companies already paying for Claude Enterprise can activate customer support workflows for zero incremental cost. The procurement cycle shrinks from months to weeks.

§ 04 — The threat

The four phases of commoditization

Commoditization is not sudden. It is a gradual erosion that compounds quarter over quarter. Application companies are actively building "model swap" layers — architectures designed to make the foundation model interchangeable. Decagon already uses multiple models. Sierra's Agent OS can swap between providers. Zendesk and Salesforce support multiple AI backends. Their customers never know or care which model powers the response.

Phase 01 · Now

Model Abstraction
Apps route queries to whichever AI is cheapest. Anthropic's brand becomes invisible to the end customer. ████░░░░░░░░

Phase 02 · 6–18 mo

Data Moats
Apps fine-tune proprietary models on customer data. General Claude loses its performance edge. ████████░░░░

Phase 03 · 12–24 mo

Pricing Squeeze
With model-swap capability, apps negotiate prices down. Margins compress structurally. ████████████░░░░

Phase 04 · 24–36 mo

Ecosystem Lock-Out
Buyers evaluate the app, not the model. Anthropic competes on price alone. ████████████████████
FIG_003 — The four phases of commoditization. © 2026 Each phase makes the next one harder to prevent. By Phase 3, the leverage has shifted permanently to the application layer. The clock started in Phase 1 — it is running now.

By Phase 3 — 12 to 24 months out — the leverage has shifted permanently. The window to act is not measured in years. It is measured in quarters.

§ 05 — Advantages

Three asymmetric advantages no competitor can replicate

Anthropic has three structural advantages. The strategy should be built on these — not on imitating what Sierra or Decagon have already built.

The Trust Brand

Anthropic's safety reputation uniquely positions Claude for HIPAA, SOX, and PCI-DSS — regulated industries that represent roughly 40% of the customer support market. PwC, Accenture, and Epic are already partnering with Anthropic for compliance use cases. Sierra and Decagon have no equivalent credential, and building one takes years, not months.

Bundled Distribution

300,000+ businesses already use Claude. Adding customer support is usage expansion, not a new sale — near-zero acquisition cost. This is the same wedge that Slack, Zoom, and Figma used to displace incumbents. The buyer already trusts the product; the conversation is about adding a workflow, not a vendor.

Intelligence That Compounds Globally

One model upgrade benefits every customer simultaneously. Sierra's bespoke integrations require custom work per account. Margin trajectory follows: 50% in 2026 moving to 77% by 2028. Integration-first architectures cannot match that trajectory because the labor cost scales with every new account.

01
The Trust Brand
Safety credential opens regulated industries — ~40% of the CS market. PwC, Accenture, Epic already partnering.
Opens the door that Sierra and Decagon cannot unlock
02
Bundled Distribution
300,000+ businesses on Claude Enterprise. CS is an expansion motion, not a new sale.
Near-zero acquisition cost — same wedge as Slack and Figma
03
Intelligence Compounds Globally
One upgrade lifts every customer at once. Margin trajectory: 50% (2026) → 77% (2028).
Integration-first cannot match this at scale
FIG_004 — Anthropic's asymmetric advantages. © 2026 The trust brand opens the door. Bundled distribution eliminates the sales cycle. Intelligence compounding wins the long game — because the margin trajectory improves with scale rather than degrading with it.

§ 06 — The playbook

Three horizons — sequenced to avoid partner conflict

The strategy has to navigate a partner tension. Anthropic cannot alienate Sierra and Decagon — who drive significant API revenue — while competing with them. The answer is market segmentation. The first horizon targets mid-market self-serve, deliberately avoiding Sierra and Decagon's enterprise segment. Subsequent horizons expand from that base.

Horizon 01 · 90 days

CS Plugin
Ship ticket triage, response generation, and brand voice customization via the existing Cowork plugin architecture. Build MCP connectors for Zendesk, Salesforce, and Intercom.

Horizon 02 · 6–18 months

Regulated Wedge
HIPAA/SOX/PCI compliance templates. Voice-native CS. "Powered by Claude" partner certification. Proactive support: AI initiates before issues escalate.

Horizon 03 · 18–36 months

Intelligence Layer
Unify CS data across product, sales, and marketing. Per-resolution pricing (300× API revenue). Agent-to-agent protocol via MCP.
FIG_005 — Three-horizon sequenced execution plan. © 2026 Horizon 1 requires no new product build and targets mid-market self-serve — deliberately avoiding Sierra and Decagon's enterprise segment. The procurement cycle is weeks, not months. There is no partner conflict because the buyer segment does not overlap.

The math of Horizon 1 is what makes the sequence compelling. Cowork plus plugin architecture already exists. The target buyer already has a Claude Enterprise account. The data flywheel — every support interaction generating signal about what customers complain about, what products fail, what language works — begins accumulating from day one and compounds in three directions: fine-tuned model improvement, cross-functional intelligence feeding product and sales, and switching costs that grow with every processed ticket.

§ 07 — The thesis

So, can Anthropic escape the trap?

The numbers at the API layer are extraordinary: $14 billion ARR, 10× annual growth, $380 billion valuation. But 80% of revenue comes from API consumption — which means 80% depends on other companies' decisions about which model to use. That concentration is not a moat. It is an exposure.

The historical analogies are direct. For Intel, the trap was mobile. For IBM, it was cloud. For Twilio, it was AI. The companies that recognized the shift and moved up the stack survived. The ones that did not became commodity inputs in someone else's value chain — still essential, but no longer in control of their own pricing, positioning, or future.

The question is not whether to act. It is whether to act now — while the window is open and while Anthropic still holds three advantages that compound rather than decay.

Customer support is not the only possible answer to the commoditization question. But it is the one where the market is already proven (Sierra to $100M in 21 months), where Anthropic's trust brand creates a wedge no competitor has (regulated industries, 40% of the market), and where the entry cost is lowest (plugin architecture that already exists). The 18-to-24-month window is not a metaphor. Phase 2 — data moats being built by application companies using Anthropic's own API — is already underway.

The escape from the commoditization trap runs through the application layer. Customer support is where the door is open right now.

Kevin Nguyen is a product manager exploring consumer, fintech, and AI. Sources: market figures from Sierra, Decagon, and Zendesk public reporting; commoditization phase analysis original. Connect on LinkedIn →

╌╌  END  ╌╌