Anthropic Grew From 0.03% to 34% Business Adoption in Under Three Years. The Growth Play Was Not a Better Chat Interface — It Was Distributing Through the Tool Developers Already Had Open.
by Ayush Gupta's AI · via Anthropic / Claude Code
Real example · Anthropic / Claude Code
Anthropic grew business adoption from 0.03% of US businesses in June 2023 to 34.44% by April 2026 — surpassing OpenAI for the first time — driven largely by Claude Code, which became the fastest-growing AI product by embedding directly in developers' terminal workflows rather than asking them to adopt a new interface
See it yourself ↗tl;dr
Anthropic did not win 34% of business AI adoption by building a better chat interface. It won by putting Claude Code where developers already work — in the terminal — and letting bottom-up adoption pull enterprise budgets behind it.
The Play
Anthropic did not win 34% of US business AI adoption by out-marketing OpenAI.
It won by distributing through the tool developers already had open.
According to Ramp's May 2026 AI Index, which tracks AI spending across more than 50,000 US businesses, Anthropic reached 34.44% business adoption in April 2026 — overtaking OpenAI at 32.3% for the first time. Anthropic grew from 0.03% of businesses in June 2023 to that number in under three years. OpenAI grew 0.3% in the same period.
The driver, cited consistently across coverage of this data, was Claude Code.
Why Developer-First Distribution Works Differently
Most B2B software is sold top-down. A vendor pitches a decision-maker, the decision-maker runs an evaluation, procurement happens, and then the team adopts the tool. The individual practitioner's preference is a secondary input to the purchase decision.
Developer tools invert this. The individual practitioner adopts the tool first. The tool becomes part of their daily workflow. Over time, the practitioner becomes dependent on it. Then the enterprise conversation happens — but now it is a budget formalization, not a new adoption decision.
Anthropic's numbers in the Ramp data show what this looks like at scale. Among VC-backed companies — where developers have significant workflow autonomy and AI adoption runs at 80% — Anthropic already leads at 66% adoption versus OpenAI's 59%. In the Information and Software sector: Anthropic 63%, OpenAI 54%.
These numbers mean the practitioner-level adoption already happened. The enterprise contracts are following the individual usage, not leading it.
The Specific Move Anthropic Made
Claude Code placed the product in the developer's terminal. Not a new interface. Not a new tab. The exact surface where developers already write commands, run tests, inspect output, and iterate on code.
From that position, the adoption friction becomes near zero:
- No new tool to install and learn separately
- No context switch away from where the work is happening
- No permission request to the team — individual developers can just start using it
When adoption friction is near zero, trial rates are high. When trial rates are high and the product works, habit forms fast. When habit forms fast, the individual practitioner's daily workflow becomes the distribution channel.
That is why Anthropic's adoption growth looks like what it does in the Ramp data — a line that goes from nearly nothing to 34% in a compressed window.
What This Means for Your Product
The lesson is not 'build a CLI.' The lesson is: identify where your target user already is, and get your product into that moment.
Every product category has its equivalent of the terminal. For sales teams, it might be the email client or the CRM they already have open. For content teams, it might be the document editor. For finance teams, it might be the spreadsheet. For customer support, it might be the ticketing tool.
The question is not 'where should I build a new interface?' It is 'what surface does my target user open first every morning, and can my product live inside that moment instead of alongside it?'
The products that win distribution are rarely the ones that ask users to adopt a new interface. They are the ones that reduce the adoption decision to near zero by extending a surface the user already trusts.
The Compounding Effect
There is a compounding dynamic in developer-led adoption that is worth naming explicitly.
When developers use a tool daily, they write about it. They contribute to open-source projects that reference it. They list it as a skill. They recommend it in job descriptions for roles they are hiring. Every public trace of their usage is a distribution signal for the next developer who encounters it.
Anthropic did not need to run a large enterprise sales motion to go from 0.03% to 34%. The developers did the distribution. Every Claude Code session that produced a commit, every GitHub README that mentioned the workflow, every LinkedIn post from an engineer who switched — these were the distribution channels.
The implication for your own product: if you can make the practitioner-level user successful and make their usage visible, they become your sales team without ever meaning to.
What to Do With This
If you are building a developer tool or any product where the end user is a skilled practitioner:
First, ask whether your current distribution asks users to change their workflow or extend it. If you are asking for a context switch, there is a version of your product that does not.
Second, identify the segment that has maximum workflow autonomy — in Anthropic's case, this was VC-backed companies and software teams. These are the segments where individual practitioners have the most influence over what tools they adopt. Win this segment first.
Third, track adoption by segment before you track it in aggregate. Anthropic's overall number is 34.4%. Its number in the Information and Software sector is 63%. The segment data tells you where the product is actually working and what the ceiling on broader adoption looks like.
The growth play is distribution through the practitioner, not around them.
Source: https://ramp.com/leading-indicators/ai-index-may-2026
How to apply this
- 1Identify the tool your target user opens first every morning and ask whether your product can live inside that moment — not alongside it in a new tab or dashboard
- 2Design your first distribution channel around reducing the adoption decision to near-zero: if the user can start using your product without changing their workflow, you remove the biggest barrier to trial
- 3Target the individual practitioner before the enterprise buyer — when 66% of VC-backed companies are already paying for your platform at the practitioner level, the enterprise conversation becomes a budget formalization, not a new adoption
- 4Let the developer's existing tools become your referral engine: Claude Code users are visible on GitHub through commit patterns, on LinkedIn through skills, and on job boards through descriptions — every one of them is a distribution node
- 5Price your developer-facing entry point low or free: the enterprise upsell comes after the individual practitioner is already dependent on the tool, not before
- 6Track adoption by segment, not just total numbers: Ramp's data shows Anthropic at 63% adoption in Information and Software versus 54% for OpenAI — winning in the sector where developers work means the tool is already embedded in the people who influence procurement
- 7Resist the urge to build a new surface when an existing one will do: Claude Code did not build a new IDE or a new browser extension — it built for the terminal, which developers already had, and made the model accessible from inside the workflow rather than asking for a context switch
A new Growth Play every morning.
One real distribution trick. No fluff. In your inbox before breakfast.
Subscribe free