Writer: GenAI with ROI

We are delighted to co-lead Writer's Series C alongside Radical Ventures & ICONIQ Growth, and join co-founders May Habib, Waseem Alshikh, and the rest of the Writer team on their journey in building a generational enterprise GenAI business. Amidst an enterprise GenAI landscape crowded with point solutions, Writer stands out as a true platform that has cracked the code on delivering tangible business value to enterprises through AI-powered workflow automation. The company's explosive growth is powered by marquee enterprise customers with diverse use cases across multiple lines of business. Writer is the panacea for enterprises seeking to move GenAI applications from POCs & experimentation into production, transforming the future of work.

Crossing the Chasm for Enterprise AI Deployment

Through our conversations with enterprise executives, we've observed a striking disconnect: while enterprises have eagerly embraced GenAI experimentation, with many establishing dedicated AI budgets, only ~5-10% of enterprise AI applications have successfully moved into production. This implementation gap stems from several interrelated challenges that emerge as organizations attempt to move beyond experimentation.

‍

Foundation model players like OpenAI’s primary goal is to crack “AGI”—they are not providing white-glove services to enterprises. As such, enterprises are forced to invest in infrastructure to build on top of FMs, in order to: cleanse datasets, fine-tune models, build vector stores, utilize RAG/agent frameworks, and build guardrails for evaluation, security, and observability. These are complex tasks that require strong AI/ML teams, large budgets, sprawling infrastructure, and collaboration with business users to understand their needs. The reality is, many organizations have tried and failed to build DIY solutions, typically stymied by #1) the ability to provide tangible value to business users, #2) the technical complexity of managing this infrastructure, #3) implementing guardrails effectively, and #4) managing hallucination risks. This challenge is particularly acute in regulated industries like healthcare and financial services.

‍

Even when technical hurdles are overcome, enterprises struggle with change management, training, and driving consistent adoption across different business units. As business unit use cases & needs evolve, the underlying AI products need to evolve in lockstep. Internal AI/ML teams build applications, but they are not trained to check user engagement, adoption, and satisfaction over time. As a result, many applications have questionable ROI or fail to move into production, creating a clear market opportunity for platforms that can bridge this gap.

‍

The Comprehensive, Purpose-Built Platform for All Enterprise Workflows

‍

Writer is building a GenAI workflow automation platform that addresses this enormous market opportunity. Writer has taken a differentiated approach to enterprise AI implementation by focusing on automating mission-critical workflows, rather than providing a generic “copilot.”  The Writer platform enables enterprises to deploy custom AI applications across business functions while maintaining rigorous controls over compliance and governance.

‍

Our general thesis is that FMs are going to win the battle for “general purpose” tasks—but two additional massive opportunities exist: #1) highly specialized vertical applications utilizing FMs to solve domain-specific problems (e.g., our investments in Hippocratic AI and EvenUp), and #2) generative AI-native applications that are deeply customized and integrated into enterprise workflows—and this is where Writer plays.

‍

Writer wins for two reasons: technology and methodology. Let’s start with technology.

Writer's Technological Edge

Writer has built the industry’s first full-stack platform, which includes a family of Writer-built LLMs, a knowledge graph with integrations for RAG, guardrails, and an AI Studio designed for both devs (low code) and business users (no code).

‍

Writer's rationale for training its own LLMs is simple: there are limits to what large, generalized FMs can do. As such, Writer has trained Palmyra models for domain-specific tasks (across healthcare, financial services, customer support, creatives, etc.) in a highly cost-efficient manner. Writer utilizes techniques such as reduced training costs with early stopping, novel inference approaches that improves performance of long-context prompts, synthetic data generation, and a unique design to their transformer architecture to accelerate inference.

‍

In addition, Writer is focused on delivering high accuracy. Hallucination is a big impediment for bringing GenAI apps into production. Traditional vector-based retrieval approaches face limitations, because they treat all enterprise data as distinct, unrelated vectors. Nearest neighbor searches necessitate a trade-off between speed/computational intensity and accuracy. As speed increases, accuracy drops materially, especially when dealing with concentrated pockets of similar concepts. These approaches result in hallucination rates of 20%+—a showstopper for mission-critical enterprise workflows. What if you could embed semantic relationships between concepts within each vector (i.e., a “map”)? The analogy here is the ability to search a database with a pre-built index—the speed is much faster, yet accurate.

‍

Writer has done exactly this with its Knowledge Graph, an innovative approach to knowledge retrieval that is centered around a graph-based RAG architecture. Writer has trained a specialized model that converts enterprise information into a graph structure, where each node represents a distinct entity and edges capture the semantic relationships between them. When processing enterprise documentation, this system can clearly differentiate between closely related but distinct concepts, maintaining the precise relationships and hierarchies that exist in the source material, rather than treating similar items as potentially interchangeable vectors that might blur together in vector space.

‍

This architectural choice, combined with Writer's proprietary, domain-specific Palmyra models, has drastically reduced hallucination rates on Writer's platform and improved retrieval accuracy. The system's effectiveness is further enhanced by its ability to maintain knowledge freshness: Writer's graph structure allows for incremental updates as new information becomes available. Rather than being limited by context windows, it can traverse relationships in the graph to answer complex queries that require synthesizing information from multiple sources. This is particularly valuable in regulated industries where accuracy and auditability are paramount.

GTM: the Enterprise-Focused Engine that AI Research Labs Lack

Writer has instrumented a world-class go-to-market machine that provides enterprises exactly what they want: turnkey GenAI solutions that address their pain points here and now with quick time-to-value. Rather than providing yet another generic "AI copilot," Writer helps enterprises identify specific high-value workflows that customers want to rebuild using GenAI, and quickly bring those visions into reality on Writer's platform. To provide a concrete example, Writer might work with a team that needs to package materials into a 40-page client brief that is sent out monthly. The data might be sourced from 10 different applications—some SaaS, some on-prem, some custom. Writer will work with this specific team to integrate with the right applications, extract the appropriate data, build a knowledge graph, understand the tone/style of communication, design the appropriate guardrails, and automate the process of creating the 40-page brief, saving the team 80-90% of the time required.

‍

This is a single workflow, in one function, in a single enterprise—and Writer is doing this at scale, across LOBs, across enterprises.

‍

How?

‍

Writer drives an intensive discovery process where they work hand-in-hand with customers to map use cases across business units, prioritizing them based on potential value creation, and then quickly building production-grade apps and workflows. The impressive feat here is that Writer has maintained best-in-class software gross margins despite its white-glove approach.

‍

Already, several large enterprise customers have seen viral growth of Writer's usage across lines of business, ranging from supply chain to investor relations to legal/compliance and more. Customers on Writer have been able to deploy custom apps into production within weeks, with consumer-grade end user adoption. We heard from customer after customer that Writer was by far their highest-ROI generative AI investment. It's no wonder that even the most sophisticated F500 customers have started to turn to Writer to realize tangible ROI from their GenAI budgets.

Building the Future of Enterprise AI

Writer has already demonstrated immense product-market fit, and the team is executing on an incredibly ambitious vision to become the comprehensive platform for AI-powered workflows in the enterprise. One of the central pieces of Writer's future roadmap is its AI Studio, which democratizes the creation of enterprise AI applications. While Writer currently provides a white-glove approach to help enterprises deploy custom applications, AI Studio enables enterprises to self-serve through a low-code/no-code environment, accelerating the pace of AI adoption across organizations.

‍

Writer's approach reverses the traditional enterprise software paradigm. Rather than building rigid workflows that humans must adapt to, Writer is creating an AI-first platform that can flexibly adapt to how enterprises actually work. The company has also brought on veterans from legendary enterprise SaaS businesses like Andy Shorkey from Mulesoft and Roger Kopfmann from Coupa who can help execute this vision and have lived the scaling journey to $500M+ ARR.

‍

We were astounded by what we heard from customers on the exponential proliferation of Writer's usage across their organizations, and the immense potential they see for capabilities like Knowledge Graph and AI studio to transform every aspect of their workflows from the ground up. Building a true enterprise AI platform that can transform how organizations work is an immensely ambitious undertaking, but May Habib is precisely the leader to make it happen. Her rare combination of deep technical understanding and commercial acumen, coupled with her relentless focus on customer outcomes, has already enabled Writer to break through where countless others have failed – and we believe this is just the beginning of what Writer will accomplish under her leadership.