Vellum

Name: Vellum
Brand: Vellum
Rating: 4.8 (20 reviews)

AI Native

Agents for everything. Except self control.

Est 2023|Valuation $90M|Raised $24.5M|33 Ref Customers

LinkedIn 7K|Website|Web traffic 245K|FTEs 37|HQ New York, New York, US

MarketsAI Assistant Agent Builder Platforms LLM Observability

Published Oct 14, 2026

01Common complaints

- If you are evaluating “Vellum” based on reviews that mention: - “ebook formatting,” “Kindle/EPUB,” “Mac publishing,” “one-time $199.99” - Those are about a different product (Vellum by 180g) and are not relevant to vellum.ai; you should reset your evaluation dataset. ([trustradius.com](https://www.trustradius.com/products/vellum/reviews?utm_source=openai)) - If you need open-source + self-host by default - Langfuse and Arize Phoenix may be better fits for teams prioritizing code access and infra control. ([langfuse.com](https://langfuse.com/pricing?utm_source=openai)) - If your stack is all-in on LangChain and you want lowest-friction integration - LangSmith can reduce integration effort and provide ecosystem alignment. ([langchain.com](https://www.langchain.com/pricing?utm_source=openai)) - If you primarily need security testing / red teaming / CI eval gates - Promptfoo (community/self-host) may be a better fit for “testing-first” workflows. ([promptfoo.dev](https://www.promptfoo.dev/pricing/?utm_source=openai)) - If your procurement requires public, downloadable compliance artifacts - Vellum’s SOC 2 report is “available by request,” which can slow down procurement compared to vendors that publish trust portals. - Current limitations acknowledged by the vendor - Vellum notes attribution gaps (“Unknown Task”) and that they are working on making background spend more visible and easier to control. ([vellum.ai](https://www.vellum.ai/docs/pricing?utm_source=openai)) - Security/compliance gaps - ISO certifications and detailed privacy compliance frameworks were not publicly documented in sources reviewed. - Technical/UX limitations cited in reviews - Complex/advanced flows can be difficult to implement (G2 reviewer). ([g2.com](https://www.g2.com/products/vellum/reviews?utm_source=openai)) - Market/brand limitation (material) - The name “Vellum” is heavily associated online with the book-formatting app, creating ongoing review and procurement confusion (TrustRadius “Vellum” listing is about the publishing tool, not vellum.ai). ([trustradius.com](https://www.trustradius.com/products/vellum/reviews?utm_source=openai))

02Why they buy

- Potential unique selling points supported by evidence - Workflow builder + eval automation + side-by-side comparisons are praised in at least one detailed G2 review as accelerating iteration and enabling non-technical contributions. ([g2.com](https://www.g2.com/products/vellum/reviews?utm_source=openai)) - SOC 2 Type II compliance is publicly claimed (report via request). ([vellum.ai](https://vellum.ai/blog/vellum-product-update-february-2024?utm_source=openai)) - unique selling points that are mostly marketing until verified - “Your assistant, your data” and “credentials stored separate from AI” are presented as product principles; buyers should request architecture diagrams and threat models. ([vellum.ai](https://www.vellum.ai/?utm_source=openai)) - Areas where competitors clearly have advantages - Langfuse: explicit data region options and an OSS model that reduces vendor lock-in risk. ([langfuse.com](https://langfuse.com/pricing?utm_source=openai)) - LangSmith: deep LangChain ecosystem synergy. - Promptfoo: strong positioning around LLM security/testing and self-hosting. ([promptfoo.dev](https://www.promptfoo.dev/pricing/?utm_source=openai))

03Field reports

- Data quality warning (high importance) - Multiple review platforms use the listing name “Vellum,” but some entries clearly refer to Vellum (180g), the Mac book-formatting software, not vellum.ai. - Example: TrustRadius “Vellum” page describes a publishing tool, Mac-only, with a published one-time fee—this is not vellum.ai. (TrustRadius product reviews page) - Some Reddit “Vellum” threads are also about book formatting and are not relevant to vellum.ai. - G2 (vellum.ai platform context) - Positive quote (reviewer: Head of Engineering): - “Every aspect of iterating on AI products is easier” and highlights side-by-side prompt/model comparison, evaluations to catch regressions, and workflow building enabling non-technical contributors. (G2 product reviews page) - Negative/limitation quote (same review): - “Some advanced flows are difficult to implement” though the reviewer frames it as workable with practice and support. (G2 product reviews page) - Capterra (vellum.ai platform context) - Capterra positions Vellum AI as a platform for building/evaluating/deploying AI workflows; overall rating appears very high with a small review footprint as of the most recent crawl. - Caveat: the detailed text of negative reviews was not accessible from the excerpted crawl here, so this research cannot quote 1–2 star reviews from Capterra directly. (Capterra product reviews page) - Reddit / community discussions (vellum.ai personal assistant context) - Recent threads comparing Vellum to other personal assistant tools (Hermes/OpenClaw) claim positives like “quick install,” “explicit permissions,” and local-first feel. These are anecdotal and may reflect an open-source/local component. (Reddit community threads) - Caveat: the specific features, reliability, and security posture in these threads are not independently validated. - Gartner Peer Insights - Attempted: a direct Vellum.ai product listing/reviews were not found in accessible results; Gartner Peer Insights access often requires interactive browsing/login and may be blocked. - In this research, only the Gartner Peer Insights portal homepage was accessible, not a Vellum.ai listing. (Gartner Peer Insights portal) - Why customers chose this product over alternatives (signals) - G2 reviewer chose it for: - Side-by-side prompt/model comparison - Version tracking - Automated evaluations to catch regressions - Visual workflow building enabling non-technical collaboration (G2 product reviews page) - Why some customers left for competitors (churn reasons) - Insufficient direct evidence from negative reviews specific to vellum.ai in accessible sources. - Common churn drivers in this category (inferred from general LLMOps buyer behavior, not Vellum-specific): pricing scaling with usage, missing self-hosting, weak RBAC/SSO for enterprises, limited observability depth, or eval workflows not fitting CI/CD.

Vendor-published metrics and claims

- Vellum’s official blog post introducing “Personal Intelligence” claims internal productivity improvements: engineers across the team showed a notable increase in PRs per month during an internal multi-month period. (Official vendor blog post) - Critical view: - This is internal/self-reported, likely confounded by other factors (focus, staffing, roadmap pressure), and lacks methodology details (baseline period, PR definition, exclusions). Treat as directional only. (Official vendor blog post)

Third-party/customer-success style compilations

A customer-success–style PDF report (FeaturedCustomers–style) mentions Vellum being helpful for prototypes and training senior engineers, but it is not a rigorous case study and may be curated marketing collateral. (Third-party customer-success PDF)
Critical view:
- Without attribution, company context, and methodology, treat as weak evidence. (Third-party customer-success PDF)

What’s missing

No detailed, independently verifiable case studies (with named companies, baseline metrics, and measured outcomes) were found in accessible primary sources during this research.

04Best fit

Ideal Customer Profile

- Company sizes - Platform ICP (inferred): - Mid-market to enterprise product teams (organization-scale customers) shipping LLM features into production with a need for testing/regression control. - Smaller startups building rapidly with mixed technical/non-technical contributors. - Personal assistant ICP: - Individuals and small teams wanting an always-on assistant without running their own infrastructure. - Industries and verticals - Not enough direct named-customer evidence in primary sources in this crawl. - Roles and personas - Platform: - ML engineers / applied AI engineers - Product managers and prompt engineers - QA / reliability engineers for evaluation gates - Security/compliance for SOC2 evidence, access controls - Personal assistant: - Founders, executives, operators, engineers seeking tool-using assistants - Qualification criteria: what makes a GREAT fit - You need: - systematic prompt/model iteration with evals - collaboration between engineers and non-engineers - structured workflows/agents rather than isolated prompts - Deal disqualifiers: what makes a POOR fit - You require: - fully open-source stack with no vendor-managed control plane - clearly documented ISO certifications and deep enterprise identity lifecycle (SSO/SCIM/audit logs) without NDAs - a product category with low tolerance for brand ambiguity during procurement

05The pitch

What the product does (core value proposition in objective terms)

- Vellum presents two adjacent product "stories" under the same brand/domain: - An enterprise LLM development platform (prompts/workflows/evals/deployments) documented in the vendor's developer documentation. It positions itself as an end-to-end platform to build, test, and deploy AI-powered applications, supporting both technical and non-technical contributors (vendor docs). - A "personal AI assistant" product (always-on assistant with memory, multi-channel access) documented on the vendor website. It emphasizes persistent memory, tool use, and access via web/mobile/desktop and chat/voice channels (vendor product pages). - For buyers: this dual positioning can be a strength (broader reach) but also creates evaluation ambiguity (different buyers, pricing logics, and security expectations). Some third-party review sites also confuse Vellum with an unrelated product that shares the same name, so diligence is required when reading "Vellum reviews" (third-party review sites).

Target market and market positioning

Platform (B2B): teams building LLM apps who need prompt iteration, workflow orchestration, evaluation, monitoring, and deployments—often categorized under LLMOps / prompt engineering lifecycle.
Personal assistant: individual prosumers and small teams wanting an "always-on" assistant with memory and integrations, without self-hosting infrastructure (vendor product pages).

Get a comprehensive analysis of Vellum including market position, competitive landscape, adoption trends, and peer benchmarks.

Download full report →