WillItSurvive
AI companies and labs · United States

Vellum AI

Vellum AI builds tools for companies to develop and test LLM-based apps.

Little protection left
Assessed 21 July 2026 · Evidence as of 21 July 2026
Not much stands between this business and its pressures.
On the recordSince 21 July 2026 · Checkable from H1 2027 · Every call on the record Checkpoint: Whether Vellum raises a Series B or stays flat two years after its Series A
The call

Vellum built enterprise LLM tooling with some real customers, then pivoted to a consumer AI assistant with no funding or hiring news to back the shift.

Vellum started by helping companies build and test AI systems, but switched to making a personal AI assistant that competes with ChatGPT and Claude. The company has not grown its team in a year even after raising $20 million in summer 2025, which suggests it is struggling. Big tech companies like OpenAI, Google, and Anthropic are adding the same features Vellum offers directly into their own products, so there is little reason for customers to pay Vellum separately.

Why
01
AI pressure: OpenAI (ChatGPT with memory), Anthropic Claude, Google Gemini. Frontier labs now ship native prompt tools, evals, and memory, so Vellum's new consumer assistant is just a thin wrapper on models it doesn't own. Could AI do the core job? Yes, likely.
02
Bigger players: Azure AI Foundry and Databricks. Cloud platforms are folding prompt management and evaluation into their existing suites, squeezing the enterprise tooling Vellum built first. Could a giant copy it? Yes, easily.
03
What protects it. A handful of enterprise customers like Drata and Swisscom with embedded workflows that would cost time and risk to rip out.
What would change this call
Vellum announces a Series B (or bridge/down-round) within the next two quarters. Shows whether investors still back the pivot or the company is running out of runway.
Headcount on Crustdata/LinkedIn moves off 0% YoY growth. Indicates whether the company is scaling the new strategy or quietly contracting.
OpenAI, Anthropic, or Google ship native persistent-memory assistant features broadly. Would directly commoditize Vellum's thin-UI personal assistant product.
Key numbers
  • Raised $20M Series A (Jul 2025)
  • 34 employees, 0% YoY headcount growth (Jun 2026)
  • Estimated $3.5M-4.3M ARR (unverified) (2025)
Named risks
  • Flat headcount 12+ months after Series A with no follow-on round announced. Severity: high
  • Dual-product strategy straddling enterprise LLMOps and consumer assistant without leadership in either. Severity: high
  • Personal assistant sold at cost with no markup, directly exposed to frontier-lab feature absorption. Severity: high
  • No audited revenue disclosed, only conflicting unverified third-party estimates. Severity: medium
Rivals, as read in July 2026
RivalDirectionEvidence
Anthropic (Claude Cowork)losing toDirectly competes with Vellum's personal assistant product as of Jun 2026
OpenAI (ChatGPT with memory)contestingNative persistent memory threatens Vellum's assistant wrapper, 2026
Google (Gemini)gaining onNative platform agent/eval features absorb standalone LLMOps value
LangChain/LangSmithcontestingListed as direct LLMOps peer competing for enterprise dev-tool share, 2026
DifycontestingOpen-source LLMOps alternative competing on cost and flexibility, 2026
Questions people ask
Is Vellum AI profitable in 2026?
No official revenue or profitability figures have been disclosed. Vellum raised a $20M Series A on July 10, 2025, led by Leaders Fund, but has not published audited revenue since, and third-party revenue estimates are unverified aggregator figures, not company statements.
Has Vellum AI raised a Series B?
As of mid-2026, no follow-on round has been disclosed since Vellum's $20M Series A closed on July 10, 2025. That is over 12 months with no announced Series B, bridge round, or new valuation.
Could AI replace Vellum AI's product?
Vellum's new personal-assistant product runs on top of OpenAI, Anthropic, and Google models at cost with no markup, per its own June 7, 2026 blog post, leaving it exposed if those frontier labs ship native memory and agent features directly to consumers.
Who founded Vellum AI and who owns it?
Vellum was founded in 2023 by Akash Sharma (CEO), Sidd Seethepalli, and Noa Flaherty, previously colleagues at Dover (YC S19). It is privately held and venture-backed, with no acquisition or parent company identified as of the July 2025 Series A.
Who is Vellum AI's biggest competitor?
Vellum lists Hermes, OpenClaw, Claude Cowork, Jan.ai, GPT4All, and Lindy as direct rivals for its 2026 personal-assistant product, while its original enterprise platform competes with LangChain/LangSmith and Dify, per Vellum's own site as of June 2026.
What does Vellum AI actually do now?
Vellum launched as an enterprise LLM development and evaluation platform in 2023, but by mid-2026 its public site vellum.ai was marketed primarily as an open-source personal AI assistant across Mac, iOS, web, voice, email, Telegram, and Slack, per its own June 7, 2026 blog post.
Timeline
  • 2026-06-07Vellum blog details personal assistant with local Ollama and frontier model support
  • 2026-05-28Vellum publishes Open Source LLM Leaderboard comparing model benchmarks
  • 2026-05-23Vellum publishes new assistant pricing: free Base, Pro from $50 per month
  • 2026Vellum launches 'Vellum for Agents' agent builder from plain-language tasks
  • 2025-07-11Vellum announces 150+ paying customers including Drata, Swisscom, Redfin
  • 2025-07-10Vellum closes $20M Series A led by Leaders Fund
  • 2023-07-11Vellum closes $5M seed round with Y Combinator, Rebel Fund, angels
  • 2023Vellum founded by Akash Sharma, Sidd Seethepalli, Noa Flaherty, joins YC W23
Revision history
21 July 2026
First assessed. Little protection left. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources