Ramp turned its own card and bill-pay data into a public scoreboard this year. The data covers more than 85,000 American businesses1. Three vendors now run away with their categories: Claude Code, Granola and Wispr Flow.
Claude Code holds 87 percent of the coding-agent market on Ramp's board1. Granola holds 66 percent of meeting notetakers2. Wispr Flow holds 93 percent of AI dictation apps3.
Each figure counts the share of businesses that bought anything in that category over the past year, narrowed to the vendor they chose most. Ramp measures every category the same way.
Three Leaders, Three Wide Margins
Wispr Flow's Lead Over Its Nearest Rival Reaches 82 Points
Adoption rate, the share of category buyers choosing each vendor on Ramp's card and bill-pay data, as of Oct. 31, 2026
The gray dot is the runner-up's adoption rate in each category; the magenta dot is the leader's; the rod is the gap between them.
Sources: Ramp [1]; Ramp [2]; Ramp [3]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Runner-up | Category leader |
|---|---|---|
| Coding Agents | 41% | 87% |
| Meeting Notetakers | 41% | 66% |
| AI Dictation | 11% | 93% |
Claude Code's nearest rival is OpenAI's Codex, at 41 percent of coding-agent buyers. Cursor holds 35 percent1. Claude Code also pulls the most switchers away from a rival. Its 38 percent switch rate is the highest Ramp measured in the category.
Granola's closest competitor is Fireflies.ai, at 41 percent. Fathom trails at 1 percent2. Granola adds 2.3 points of adoption a month, faster than any other notetaker on Ramp's board.
Wispr Flow's gap is the widest of the three. Superwhisper, the next-closest dictation app, holds 11 percent. Aqua Voice holds 4 percent3.
Company size barely moves the number. Small firms adopt it at a 93 percent rate. Mid-sized firms land at 96 percent, and the largest firms sit at 94 percent.
Ramp's own data team flags a limit worth weighing here. Its customers skew toward high-growth, tech-forward businesses, a narrower slice than the typical American firm7. A reader at a slower-growing company should treat these shares as a map of what tech-forward peers buy. The map offers a hint for a different kind of firm, short of a guarantee.
Startups Win Four of Five Categories
Look past these three vendors, and a pattern holds across the rest of Ramp's board. The giant AI labs rarely win these fights. Independent startups do.
Sierra, the customer-service startup Bret Taylor founded after leaving Salesforce, leads AI support bots with 34 percent4. Attention leads AI-first sales agents with 31 percent5. Count Granola and Wispr Flow alongside them, and startups lead four of the five applied AI categories Ramp measured.
Startups Lead 4 of 5 Applied AI Categories on Ramp's Board
Which kind of company sells the top vendor in each of Ramp's five applied AI tool categories, Oct. 31, 2026
Each cube is one category. Magenta cubes are categories an independent startup leads; the gray cube is the one category an AI lab's own tool leads, Anthropic's Claude Code in coding agents.
Counts five applied AI categories only; Ramp's Foundational LLMs category, which labs lead by definition, sits outside this count.
Sources: Ramp [1]; Ramp [2]; Ramp [3]; Ramp [4]; Ramp [5]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| Led by an independent startup | 4 |
| Led by an AI lab's own tool | 1 |
| Whole | 5 |
Anthropic's Claude Code wins the fifth category, coding. Coding is also the one spot built directly on the lab's own model.
A reader could object that Anthropic's base model leads its own board too. It holds 80 percent adoption against OpenAI's 78 percent6. Counting both would put the labs ahead twice, in coding and in the model underneath it.
The point still stands: every tool on this board, Claude Code included, already runs on somebody's base model. A business picks those five tools once it already has a model to run them on.
What Each Tool Costs, and Its Catch
A team buying Claude Code pays through a Claude Pro, Max or Ultra plan. For groups of two to 150, a $20-per-seat Team plan, billed yearly, is the other option12. Anthropic's enterprise tier adds spend limits and audit controls, priced per customer.
Granola's free Basic plan keeps 30 days of meeting history. Its Business plan costs $14 a seat a month. Enterprise runs $35, with single sign-on starting at 50 seats14.
Granola blocks OpenAI and Anthropic from training on a customer's notes. It trains its own model on scrubbed data, unless a user opts out.
Wispr Flow's free plan caps a user's weekly meeting transcription and keeps its shared word list for personal use only13. Its Pro plan runs $12 a seat a month paid yearly, or $15 billed monthly. Growth costs $18 a seat with basic admin controls. Enterprise carries a custom price for each customer.
The Case for Trusting the Receipts
Tanay Kothari founded Wispr Flow in 2023. He told Computerworld what changes once a worker tries voice input instead of a keyboard. "People report that their lives look different after using Wispr Flow," Kothari said. "On average, they reduce their daily typing time from five hours to three"9.
Chris Pedregal co-founded Granola and runs it as chief executive. He turns down requests from company leaders who want to read a worker's own meeting notes. Pedregal told Platformer why he holds that line, even when it costs a sale. "Companies hate us for this, they're like, hey, there's all this content, why am I paying you money?" he said11.
The Catch the Top Seller Still Carries
A plaintiff filed a proposed class action against Granola in a federal court in California. It says the company records every voice in a meeting once a single person agrees to it10. The suit argues that a guest's words still feed Granola's transcripts and model training by default. This claim holds even for a guest who skipped signing up.
Enza Iannopollo studies AI and privacy risk as a vice president and principal analyst at Forrester. She told Computerworld why a notetaker worries her more than an ordinary recorder. "AI notetaking is more dangerous than any other type of traditional recording apps and tools," Iannopollo said8.
She wants every person in a meeting told plainly when a tool is listening. Each person should also hear what happens to the recording after.
The lawsuit stands apart from Granola's sales lead. A buyer of the category's biggest seller still owns the same consent questions as a buyer of its smallest.
What to do
Recheck price and seat minimums before buying. Ramp's board moves monthly, and a quote from October can change by January.
Ask any notetaker vendor, in writing, how it treats a meeting guest who kept the app closed. Get the training-data opt-out confirmed in that same message.
Before adding seats of Claude Code, Granola or Wispr Flow, talk to one more team. Find one already using the runner-up, and ask what it would miss by switching.
By the numbers
- 87 percent: the share of coding-agent buyers on Ramp who chose Claude Code, against 41 percent for OpenAI's Codex1.
- Wispr Flow: 93 percent adoption in AI dictation, well past Superwhisper's 11 percent3.
- Granola: 66 percent adoption in AI meeting notetakers, with Fireflies.ai next at 41 percent2.
- 85,000-plus: American businesses sit inside the Ramp card and bill-pay data behind every figure here1.
- Switch rate: 38 percent of Claude Code's buyers came from a rival coding agent, the highest Ramp measured in that category1.
- $35: Granola's Enterprise price per seat a month, the plan that adds single sign-on and its deepest security controls14.
- Sierra and Attention: 34 percent and 31 percent adoption, the startups leading AI support bots and AI-first sales agents45.
What to watch
Watch Ramp's board next month. Codex and Cursor are both still gaining on Claude Code's lead, even if slowly1. Watch the Chamberlain lawsuit's next filing for whether a court lets Granola's consent defaults stand10. A ruling against Granola would reach every notetaker that still records before every voice agrees to it.
Sources
- Ramp, "Coding Agents Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/code-ai
- Ramp, "AI Meeting Notetakers Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/ai-meeting-notetakers
- Ramp, "AI Dictation Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/ai-dictation
- Ramp, "AI Support Bots Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/ai-support-bots
- Ramp, "AI-First GTM (Agentic Systems) Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/ai-first-gtm-agentic-systems
- Ramp, "Foundational LLMs Ramp Rate: A Data-Backed Look," Ramp, Oct. 31, 2026, https://ramp.com/vendors/categories/foundational-llms
- Ramp, "How Ramp data works: an explainer," Ramp, Oct. 15, 2026, https://ramp.com/data/how-ramp-data-works
- Matthew Finnegan, "Granola lawsuit raises concerns over AI note-taking app privacy," Computerworld, Aug. 6, 2026, https://www.computerworld.com/article/4206255/granola-lawsuit-raises-concerns-over-ai-note-taking-app-privacy.html
- Matthew Finnegan, "Wispr CEO: What a post-keyboard office might look like," Computerworld, Dec. 31, 2025, https://www.computerworld.com/article/4107331/wispr-ceo-interview-post-keyboard-office.html
- Ryan Merket, "Granola sued over bot-free meeting capture and AI training defaults," Runtimewire, Aug. 1, 2026, https://runtimewire.com/article/granola-sued-bot-free-meeting-capture-ai-training
- Casey Newton, "This AI notetaker won't sell surveillance to your boss," Platformer, Jul. 30, 2026, https://www.platformer.news/granola-chris-pedregal-interview/
- Anthropic, "Claude Code," Claude (Anthropic), Oct. 20, 2026, https://claude.com/product/claude-code
- Wispr Flow, "Wispr Flow Pricing," Wispr Flow, Oct. 20, 2026, https://wisprflow.ai/pricing
- Granola, "Granola Pricing," Granola, Oct. 20, 2026, https://www.granola.ai/pricing
