Writer (AI company)
Last reviewed
Jun 5, 2026
Sources
35 citations
Review status
Source-backed
Revision
v4 ยท 4,487 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 5, 2026
Sources
35 citations
Review status
Source-backed
Revision
v4 ยท 4,487 words
Add missing citations, update stale details, or suggest a clearer explanation.
Writer is a generative artificial intelligence company headquartered in San Francisco, California, that develops a full-stack enterprise AI platform built on its proprietary family of large language models called Palmyra. Founded in August 2020 by CEO May Habib and CTO Waseem AlShikh, Writer provides organizations with tools for building, deploying, and managing AI applications and agents grounded in company-specific data. The platform combines custom LLMs, a graph-based Knowledge Graph for retrieval-augmented generation (RAG), built-in guardrails for governance, and a no-code/low-code development environment called AI Studio. As of 2025, Writer serves over 300 enterprise customers including Accenture, L'Oreal, Uber, Vanguard, Salesforce, and Prudential, and has raised $326 million in total funding at a $1.9 billion valuation.
Writer's origins trace back to Qordoba, a software localization startup co-founded in 2015 by May Habib and Waseem AlShikh. Habib, a Lebanese-Canadian entrepreneur, had previously worked in investment banking and at the Abu Dhabi-based sovereign wealth fund Mubadala Development Company before relocating to San Francisco to pursue entrepreneurship. AlShikh, originally from Syria, discovered his passion for computer science at age 18 and later earned degrees in electronics from Damascus Polytechnic University and Beirut Arab University. The two connected through a Twitter community of computer scientists in Dubai, where AlShikh's posts about algorithms and Java caught Habib's attention.
Qordoba used machine learning to help enterprises dynamically localize content across multiple languages, raising $21 million in capital during its operation. By 2020, the co-founders recognized a broader opportunity in applying AI to enterprise content and communication beyond localization.
In August 2020, Habib and AlShikh pivoted from Qordoba and launched Writer, building on the same core principle of translating text from one "language" into another, but extending the concept to brand voice, messaging consistency, and AI-powered content generation. The company initially positioned itself as an AI writing assistant for enterprises, competing with tools like Grammarly while emphasizing privacy, security, and enterprise-grade features.
Writer secured a $5 million seed round in 2020 from investors including Aspect Ventures and Upfront Ventures.
In November 2021, Writer raised a $21 million Series A led by Insight Partners, with participation from Gradient Ventures and a group of angel investors. The company continued to develop its proprietary language models and enterprise platform.
By September 2023, Writer had grown significantly and announced a $100 million Series B led by ICONIQ Growth, with participation from Insight Partners, WndrCo, Balderton Capital, and Aspect Ventures. Strategic investors included Accenture and Vanguard, both of whom were also customers. This round valued the company at approximately $500 million and brought total funding to $126 million.
In November 2024, Writer closed a $200 million Series C co-led by ICONIQ Growth, Premji Invest, and Radical Ventures. Additional participants included Salesforce Ventures, Adobe Ventures, B Capital, Citi Ventures, IBM Ventures, and Workday Ventures. The round valued Writer at $1.9 billion, establishing the company as a unicorn. Total funding reached $326 million.
In October 2025, Salesforce Ventures disclosed a new investment in Writer at the Dreamforce 2025 conference, deepening a relationship in which Salesforce was already both an investor and customer. Salesforce Ventures managing director Paul Drews framed the deal around Writer's enterprise traction with accounts such as Goldman Sachs and Accenture, while Habib described the value of grounding agents in customer data that lives inside Salesforce systems. Writer cited distribution across the Salesforce ecosystem as a key rationale for the partnership. The terms of the investment were not disclosed.[26][27]
| Round | Date | Amount | Lead Investor(s) | Valuation | Notable Participants |
|---|---|---|---|---|---|
| Seed | August 2020 | $5M | Aspect Ventures, Upfront Ventures | N/A | N/A |
| Series A | November 2021 | $21M | Insight Partners | N/A | Gradient Ventures |
| Series B | September 2023 | $100M | ICONIQ Growth | ~$500M | Insight Partners, WndrCo, Balderton Capital, Accenture, Vanguard |
| Series C | November 2024 | $200M | ICONIQ Growth, Premji Invest, Radical Ventures | $1.9B | Salesforce Ventures, Adobe Ventures, IBM Ventures, Workday Ventures |
Writer experienced rapid revenue expansion alongside its fundraising. The company reported approximately $2 million in annual recurring revenue (ARR) in 2022, growing to $15.7 million in 2023 and reaching $47 million by November 2024, representing 194% year-over-year growth. The company has maintained over 150% net revenue retention, indicating that existing customers consistently expand their usage of the platform over time.
Writer develops and trains its own proprietary family of large language models under the Palmyra brand, rather than relying on third-party models from providers like OpenAI or Anthropic. This approach gives Writer full control over model architecture, training data, and optimization for enterprise workloads. The Palmyra family includes general-purpose, domain-specific, multimodal, and creative models.
In February 2023, Writer publicly launched its first generation of Palmyra models in three sizes:
These initial models were trained on enterprise-focused data and positioned as alternatives to general-purpose LLMs for business applications.
Palmyra X 003 Instruct is a model with a 32,000-token context window designed for precise and detailed responses, with strong multilingual support covering over 30 languages. It was positioned as Writer's instruction-following model for tasks requiring accuracy and nuanced output.
Released in October 2024, Palmyra X 004 marked a significant leap in capability. The model features a 128,000-token context window and was specifically designed to power AI agents and autonomous applications through advanced tool use and function calling.
Palmyra X 004 achieved the top position on the Berkeley Function Calling Leaderboard (BFCL) with a score of 78.76%, outperforming models from OpenAI, Anthropic, Google, and Meta by nearly 20%. It also ranked in the top 10 on Stanford University's Holistic Evaluation of Language Models (HELM) benchmark, scoring 86.1% on HELM Lite and 81.3% on HELM MMLU.
Writer achieved these results with a model containing approximately 150 billion parameters, trained using synthetic data at a reported average cost of roughly $700,000, a fraction of what major AI labs spend on comparable models. As listed in Writer's developer documentation, Palmyra X4 is priced at $2.50 per million input tokens and $10.00 per million output tokens.[28]
Announced on April 28, 2025, Palmyra X5 is Writer's most advanced foundation model. It features a 1-million-token context window (equivalent to roughly 1,500 pages or six books) and introduces adaptive reasoning, which allows the model to dynamically adjust its strategy based on the complexity of the input.
Technically, Palmyra X5 uses a hybrid attention mechanism that blends linear and softmax attention to process multi-million-token inputs without the memory and speed penalties associated with standard transformer attention. The model can absorb an entire million-token prompt in approximately 22 seconds and return individual function-calling turns in roughly 0.3 seconds.
Palmyra X5 was trained with synthetic data for a reported cost of just $1 million in GPU compute. It is priced at $0.60 per million input tokens and $6.00 per million output tokens. The model supports code generation, structured outputs, and over 30 languages. It is available through Writer AI Studio and Amazon Bedrock.
Writer positioned Palmyra X5 as delivering performance near OpenAI's GPT-4.1 at roughly 75% lower cost, citing per-token pricing three to four times cheaper than GPT-4.1. On OpenAI's MRCR 8-needle long-context test the model scored 19.1%, against 20.25% for GPT-4.1 and 17.63% for GPT-4o, and it scored 70.99% on Big-Bench Hard (BBH). Writer described early enterprise use cases including financial reporting, RFP responses, support documentation, and multi-step agentic workflows.[29]
Palmyra Med is a domain-specific model tailored for the healthcare industry. Built on a 70-billion-parameter architecture with a 32,000-token context window, it was trained on specialized medical data. In testing, Palmyra Med averaged 85.9% accuracy across standard medical benchmarks, surpassing Med-PaLM-2 by approximately 2 percentage points. It is designed for use cases such as clinical decision support, medical documentation, and patient communication.
Palmyra Fin is a financial domain model built on 70 billion parameters with a 128,000-token context window. It was specifically developed for the financial sector and demonstrated its capabilities by passing the CFA Level III exam. The model supports tasks such as financial analysis, regulatory compliance, risk assessment, and portfolio research.
Palmyra Vision is Writer's multimodal model, announced in February 2024. It extends the Palmyra family beyond text to include image understanding and visual question answering. The model features an 8,000-token text context window and can extract handwritten text, classify objects, analyze charts and graphs, and answer questions based on visual inputs.
On the VQAv2 benchmark, Palmyra Vision scored 84.4%, outperforming both GPT-4V and Gemini 1.0 Ultra. Enterprise use cases include image-based compliance checks, automated product description generation, chart interpretation, and handwritten text digitization. It is priced at $0.005 per image plus $7.50 per million text output tokens.
Released in December 2024, Palmyra Creative is a model purpose-built for creative thinking and writing tasks. It features a 128,000-token context window and focuses on three core capabilities: divergent thinking to generate a wide range of unique ideas, co-creation that adapts outputs based on user feedback, and the delivery of solutions tailored to specific creative challenges.
The model supports narrative development, poetry, scriptwriting, marketing copy, character creation, and dialogue writing. It is available through Writer AI Studio, the Writer Framework, and NVIDIA NIM inference microservices. Pricing is set at $5.00 per million input tokens and $12.00 per million output tokens.
In 2025, Writer released its Palmyra-mini family as open-source models on Hugging Face, marking the company's first contribution to the open-source AI community. The family includes three variants:
Palmyra-mini achieved a score of 0.818 on the GSM8K benchmark and is notable for being the first Writer LLM capable of running entirely on an iPhone. These models are intended to complement larger Palmyra models for speed-sensitive and cost-sensitive parts of enterprise workflows.
| Model | Context Window | Key Capability | Pricing (per 1M tokens) | Release |
|---|---|---|---|---|
| Palmyra X5 | 1,000,000 tokens | Adaptive reasoning, agentic workflows | $0.60 in / $6.00 out | April 2025 |
| Palmyra X4 (X 004) | 128,000 tokens | Tool calling, function calling, RAG | $2.50 in / $10.00 out | October 2024 |
| Palmyra X 003 Instruct | 32,000 tokens | Instruction following, multilingual | $7.50 in / $22.50 out | 2024 |
| Palmyra Creative | 128,000 tokens | Creative writing, ideation | $5.00 in / $12.00 out | December 2024 |
| Palmyra Vision | 8,000 tokens | Image understanding, VQA | $0.005/image + $7.50 out | February 2024 |
| Palmyra Med | 32,000 tokens | Healthcare, clinical decision support | Enterprise pricing | 2024 |
| Palmyra Fin | 128,000 tokens | Finance, CFA-level analysis | Enterprise pricing | 2024 |
| Palmyra-mini | N/A | Lightweight on-device reasoning | Open source | 2025 |
Writer's developer documentation lists Palmyra X5 as the flagship general-purpose model and the recommended replacement for its older specialized models. The company scheduled five earlier models for deprecation on July 13, 2026: Palmyra X 003 Instruct, Palmyra Vision, Palmyra Med, Palmyra Fin, and Palmyra Creative. A separate Palmyra Fin 32k variant had already been deprecated in March 2025. As of mid-2026 the actively listed Palmyra models centered on Palmyra X5 and Palmyra X4.[28]
Writer distinguishes itself from competitors by offering what it calls a "full-stack" approach to enterprise AI. Rather than providing models alone, the platform integrates custom LLMs, data infrastructure, application development tools, and governance controls into a single system. This architecture is designed to reduce the complexity that enterprises face when assembling AI solutions from multiple vendors.
The platform's core components include:
Writer's Knowledge Graph is a graph-based retrieval-augmented generation system that stores enterprise data as a network of semantically connected concepts, entities, and events. Unlike traditional vector database approaches that rely on distance-based similarity, Writer's Knowledge Graph maps explicit semantic relationships between data points, enabling more accurate retrieval for AI-generated responses.
In benchmarking on the RobustQA dataset, Writer's Knowledge Graph achieved 86.31% accuracy, significantly outperforming seven competing RAG approaches that scored between 32.74% and 75.89%. Writer reports 67% lower cost compared to leading alternative RAG implementations, with savings increasing at scale.
The Knowledge Graph supports multiple data connectors, enabling enterprises to ingest data from internal systems and keep the graph continuously updated. It forms the foundation for grounding AI outputs in accurate, company-specific information and reducing hallucinations.
AI Studio is Writer's development environment for building and deploying enterprise AI applications. It offers both no-code and low-code tools that allow technical developers and business users to collaborate within a shared workspace.
The no-code interface features a visual editor with a library of drag-and-drop blocks, enabling business users to create AI applications that automate workflows, generate content assets, or answer questions based on company data without writing code. For more technical users, AI Studio provides API access, the Writer Framework (a Python-based development toolkit), and integration with external systems.
AI Studio supports all Palmyra models and allows enterprises to build applications ranging from simple chatbots to complex multi-step agent workflows.
In November 2025, Writer launched Writer Agent, a unified interface that combines its Ask Writer conversational assistant with an autonomous Action Agent. The release introduced several new capabilities:
All of these capabilities are included in the core platform pricing without additional charges.
Writer extended the Agent platform with two major releases in early 2026 aimed at letting non-technical business users build and automate agents.
On March 25, 2026, Writer introduced what it described as the first Agent Skills and Playbooks for business teams. Skills are reusable building blocks that encode a team's methodologies, quality standards, and decision frameworks so that, for example, the way a strategist structures a competitive analysis can be captured once and reused across an organization. The release included a Skills creator and an enhanced Playbook builder that let employees describe a workflow in natural language and generate production-ready agent capabilities, enhancements to shared Voice profiles for brand consistency, an expanded set of third-party connectors, and a ready-to-use library of more than 200 enterprise-specific Skills. Doris Jwo, Writer's vice president of product, said the goal was to move AI productivity gains beyond developers and early adopters and out of the hands of IT gatekeepers. Writer cited a Forrester Total Economic Impact study reporting an average 333% return on investment with a six-month payback period and noted that customers had deployed more than 5,000 playbooks.[30][31]
On April 30, 2026, Writer added event-based triggers to Writer Agent, allowing agents to listen for real-world business signals and execute multi-step workflows without a person initiating the process. The triggers operate across systems including Gmail, Gong, Google Calendar, Google Drive, Microsoft SharePoint, and Slack, automatically starting research, generating deliverables, updating systems, and orchestrating handoffs. The release added a connector for Adobe Experience Manager that gives marketing teams read and write access to pages, content fragments, and digital assets, along with governance controls such as connector profiles with team-specific permissions, customizable Writer Agent profiles, AI Studio observability, a Datadog logging plugin for LLM request and response data, and support for bring-your-own encryption keys on AWS, Azure, and Google Cloud. Jwo characterized event-based triggers as a shift in how enterprises deploy AI agents. Trade coverage framed the launch as Writer competing more directly with autonomous-agent offerings from Amazon, Microsoft, and Salesforce.[32][33]
Writer provides integrated guardrails that keep AI agent behavior aligned with system requirements and enterprise policies. Governance features include identity validation, granular permissions, approval flows, and controls over which data sources and applications agents can access. Enterprises can define brand voice profiles, departmental writing standards, and compliance rules that constrain all AI-generated output.
The platform includes full observability and auditability tools, allowing administrators to monitor how AI operates across the organization and trace outputs back to their data sources.
Writer holds multiple security certifications and compliance attestations relevant to enterprise deployment:
| Certification | Description |
|---|---|
| SOC 2 Type II | Annual examination evaluating controls for security, availability, and confidentiality |
| HIPAA Type 1 | Compliance with healthcare data privacy requirements |
| PCI Compliance | Payment card industry data security standards |
| ISO/IEC 27001 | Information security management system certification |
| ISO/IEC 27701 | Privacy information management certification |
| ISO/IEC 42001 | Responsible AI management certification |
The platform supports all major single sign-on (SSO) providers and multi-factor authentication. Data isolation is enforced between teams within the same organization, ensuring that inputs and data from one team are not accessible by another.
As of 2025, Writer serves over 300 enterprise customers across multiple industries. The company reports an average nine-times return on investment for its customers. Notable customers include:
| Customer | Industry | Use Cases |
|---|---|---|
| Accenture | Consulting | Enterprise content, AI deployment |
| L'Oreal | Consumer Goods | Marketing content, brand consistency |
| Uber | Transportation | Content automation |
| Vanguard | Financial Services | Financial research, client communications |
| Salesforce | Enterprise Software | Internal AI applications |
| Prudential | Insurance/Finance | Compliance, documentation |
| Mars | Consumer Goods | Marketing, product content |
| Intuit | Financial Software | Enterprise AI workflows |
| Dropbox | Cloud Storage | AI-powered productivity |
Writer generates revenue through enterprise software subscriptions, with pricing tiers that include Team, Enterprise, and custom plans. Enterprise plans offer full Knowledge Graph capabilities, unrestricted connectors, departmental brand and voice profiles, system interoperability, and granular agent governance.
Writer competes in the enterprise generative AI market against several categories of providers:
Foundation model companies with enterprise offerings: OpenAI (ChatGPT Enterprise, API), Anthropic (Claude for Enterprise), and Google (Gemini for Google Cloud) all offer enterprise-grade API access and platform features. These providers generally have larger models trained on broader datasets, but Writer argues that its enterprise-first approach, custom training, and full-stack platform offer advantages in data security, brand consistency, and total cost of ownership.
Enterprise AI platforms: Companies like Glean, Cohere, and AI21 Labs also target the enterprise market with their own models and platform capabilities. Writer differentiates through its combination of proprietary models, Knowledge Graph technology, and no-code development tools.
Horizontal AI tools: Broader enterprise AI tools from companies like Microsoft (Copilot), Salesforce (Einstein/Agentforce), and ServiceNow compete for the same enterprise AI budgets, though they typically integrate third-party models rather than training their own.
A key element of Writer's competitive positioning is cost efficiency. Writer has emphasized that it can train models at a fraction of the cost incurred by larger AI labs. The average training cost for Palmyra X4 was reported at $700,000, compared to estimates of $4.6 million for a comparably sized OpenAI model. Palmyra X5 was trained for approximately $1 million in GPU compute.
May Habib is the CEO and co-founder of Writer. Born in rural Lebanon, she immigrated to Canada with her family in the 1990s as the eldest of eight children. She graduated with high honors from Harvard University with a degree in Economics and Near Eastern Languages and Civilizations, where she served as Associate Managing Editor of The Harvard Crimson.
After Harvard, Habib worked in investment banking with technology companies in New York before joining Mubadala Development Company, a sovereign wealth fund based in Abu Dhabi. She moved to San Francisco in 2015 to co-found Qordoba, the localization startup that would eventually become Writer.
Waseem AlShikh is the CTO and co-founder of Writer. He grew up in the mountains south of Damascus, Syria, where he discovered computer science at age 18. He moved to Lebanon for university studies, where he developed a text summarization tool to overcome the language barrier of English-language instruction. AlShikh holds degrees in electronics from Beirut Arab University and Damascus Polytechnic University.
He achieved the highest national rank in the Syrian National Computer Science Competition two years in a row. After graduating, he moved to Jordan for his master's degree and then to Dubai for professional work. He connected with Habib through a Twitter community of computer scientists and joined her at Qordoba before co-founding Writer.
Writer has received several industry recognitions:
On April 7, 2026, Writer published its second annual "AI Adoption in the Enterprise" survey, conducted with the independent research firm Workplace Intelligence. The study drew on 2,400 respondents, split evenly between C-suite executives and employees, surveyed between December 17, 2025, and January 25, 2026, across the United States, United Kingdom, Ireland, Benelux, France, and Germany at companies ranging from 100 to more than 10,000 employees. Among the headline findings, 60% of executives said they planned to lay off employees who would not adopt AI, 92% said they were actively cultivating "AI elite" employees, and 79% acknowledged struggling with issues such as lagging return on investment, strategy gaps, and internal power struggles. The survey reported that 97% of executives viewed AI as beneficial while only 29% had seen significant returns from generative AI. May Habib used the release to argue that "layoffs are not a viable AI strategy," contending that the companies seeing the most benefit were those redesigning operations around human-agent collaboration rather than headcount reduction.[34][35]
In March 2025, Writer announced a significant international expansion, opening new offices in Singapore, Dublin, Chicago, and Austin, while expanding existing offices in San Francisco, New York, and London. The company appointed regional leadership to support market entry and customer growth across Europe and the Asia-Pacific Japan region.