Raviv Kerem - GTM Lead, Thread AI

August 26, 2025

Enterprise AI with Zero ROI? Not for Thread Customers.

MIT's recent report, "The GenAI Divide", claims that 95% of organizations are getting zero return on GenAI Investments.

We’re not surprised. Operationalizing AI is hard. It requires specific, purpose-built infrastructure and organizational buy-in. As we've written about before, realizing the promise of AI requires better tools that democratize AI infrastructure. Many companies we meet have been burned by promising pilots that can't scale or brittle tools constrained by their underlying infrastructure choices. The barriers the MIT Project NANDA noted are playing out across enterprises every day: a lack of durable infrastructure, knowledge gaps driven by siloed initiatives, and a fundamental disconnect between generic models and the dynamic workflows of a real business.

But for our customers, it’s a different story. They are already embedding AI into core workflows and achieving substantial results. A global media agency has automated components of its RFP response process, reducing response time by 70% and increasing its annual submission volume by 250%. Another large enterprise is reducing hundreds of monthly review hours for its accounts-payable staff. An autonomy company is leveraging the platform to power critical responses across languages when every second matters.

These outcomes are a direct result of treating the challenge of operationalizing AI as, at its core, an infrastructure and collaboration problem.

From Stalled Pilots to Scaled Production

The promise of AI is not another siloed application. It's the ability to compose intelligent, automated processes that connect disparate systems, data sources, and models. Today, most organizations lack the foundational layer of infrastructure to build and manage these workflows in a reliable, secure, and observable way.

Without this layer, even the most impressive demo will fail when faced with the complexity of an enterprise environment. The 95% failure rate cited by NANDA is a direct consequence of brittle integrations, challenging data governance, and siloed, static workflows.

At Thread AI, we engineered our composable AI orchestration platform, Lemma, to be that missing infrastructure layer. We combine a durable, secure, and observable backend with a powerful low-code UI and a robust SDK, enabling builders of vast technical levels to collaborate on developing and deploying mission-critical workflows and agents on a single platform.

Lemma's unique combination of flexible interaction patterns, multimodal data handling, and central observability enables an enterprise-grade experience for powering mission-critical workflows. Its ability to resume stopped or paused processes as well as its built in guardrails and complex authentication schemes, are key elements that make it truly enterprise-ready, giving our customers the confidence to automate their most critical processes and leverage AI to transform their operations.

Our customers are crossing the GenAI divide by building on this foundation. See a few of our case studies here:

Case Study 1: Automated Proposal Management (RFP) at a Global Media Agency

For a leading global agency with thousands of employees and projects, the proposal management process is resource-intensive, often requiring days or even weeks of coordinated effort from multiple teams.

Using Lemma, the agency built a sophisticated, AI-powered workflow to streamline this entire process. The solution leverages a Retrieval-Augmented Generation (RAG) framework composed of two key Lemma Workers:

A Data Hydration Worker that ingests and processes all of the agency’s relevant knowledge to build a clean, vetted corpus of information stored in a vector database.

And a Retrieval Worker that when a new RFP (Request for Proposal) arrives, this worker fields the questions and generates a search. Cited responses allow a human reviewer to instantly verify the source data. This powerful capability is plugged directly into the tools their team already uses, like Google Sheets, with no disruption to their established processes.

The results are tangible: a 70% improvement in proposal response time and a 250% increase in RFP submission capacity. Agency employees are now free to focus on high-value, strategic client work.

Case Study 2: Automating Accounts Payable for a Global Enterprise

Manually reviewing thousands of invoices is a classic back-office bottleneck - a tedious process prone to human error. One of our enterprise customers is tackling this head-on.

They’ve used Lemma to deploy an intelligent invoice processing Worker that automates the entire accounts-payable cycle.

The Worker:

Ingests invoices from multiple sources.

Uses AI-powered states to perform data extraction on key fields like supplier, invoice number, and line items.

Validates the extracted information by reliably connecting to internal systems of record.

Leverages Lemma’s native Handoff State to seamlessly route exceptions or high-value invoices to an AP clerk for human review and approval in a simple, intuitive UI.

This workflow is on track to save their finance team hundreds of review hours every month, improving accuracy and freeing up staff for more strategic financial oversight.

Case Study 3: Automating Pre-Meeting Intelligence for a Global Consulting Firm

For one of the leading global consulting firms, ensuring front-office employees are deeply informed before any client meeting is critical to success. The traditional process of compiling a Public Information Book (PIB) was a manual, time-consuming effort requiring analysts to hunt down, aggregate, and synthesize information from a wide array of disparate sources.

Using Lemma, the firm automated this complex diligence process by building an automated Just-in-Time Insights Brief that mirrors the workflow of creating a Public Information Book. This workflow transforms pre-meeting preparation from a multi-hour research project into an on-demand, automated capability.

The Worker:

Orchestrates diligence across multiple diverse and distributed sources of information, simultaneously pulling data from public, private, and proprietary sources.

Ingests diverse, multi-format data, including:

Public Filings: 10-Ks and 10-Qs from the EDGAR database.

Earnings Calls: Transcripts from public news sources.

Market Commentary: Proprietary analyst reports.

Live News and Market Intelligence: Real-time news feeds and stock price movements.

Extracts and synthesizes the most critical information from these sources into a single, structured, and easy-to-digest report, that overlays sentiment analysis and other key for signals to allow front-office employees to come prepared with the most up-to-date information.

The result is a dramatic reduction in manual labor, cutting down research that formerly took hours to just minutes. More importantly, it ensures every employee enters a meeting fully equipped with the latest, most relevant intelligence, enabling them to lead more insightful and impactful client discussions.

Crossing the Divide

These organizations are just a sample of customers who aren’t just adopting AI; they are operationalizing and productionizing it. They are succeeding because they are building on a platform designed for the complex realities of the enterprise, doing so in a collaborative manner that compounds value across the organization.

The GenAI Divide isn't permanent, but crossing it requires a shift in focus—from the model to the infrastructure. It requires a composable, durable, and collaborative platform that empowers your teams to build workflows that deliver real, measurable value. Your AI investments should be a strategic lever that compounds, not just iterates.

Our platform addresses the top most common concerns pulled directly from this report:

Our customers trust us with hard problems.
While the results speak for themselves, we are determined to meet our customers where they are and build that trust from day one. We understand that many organizations, especially established enterprises, have legacy systems. We don’t shy away from these hard systems and problems. In fact, we believe cutting-edge technologies can learn a lot from the robustness and security of these proven architectures. Our goal is to bring the latest and greatest AI capabilities to these environments hand-in-hand with our customers.

Lemma allows for complex, cross-team Workflows to be composed.
Collaboration is key when building critical workflows and decision-making in enterprises is often federated across people and business units. Each of these roles contributes to a larger workflow, and Lemma is designed to support this kind of cross-functional collaboration so that technology decisions and business outcomes are in lock-step.

It is designed for minimal disruption.
Lemma acts as a connective tissue, integrating with your existing systems—from modern SaaS applications like Salesforce to legacy databases by integrating at the protocol layer. Our platform is agnostic to your model or cloud provider, so you can avoid vendor lock-in and leverage the best tools for the job, without ripping and replacing what already works.

The platform supports clear data boundaries and adherence to entitlements.
Lemma offers enterprise-grade security and governance, supporting complex authentication models for both human and non-human identities. Lemma Workers and resources can import existing user groups from the organizations IdP (Identity Provider), maintaining the same level of permissions and security posture set by the enterprise. Lemma Workers are also built with distinct identifiers for observability and maintain a separation of the data and compute layers, allowing each run to maintain its own context.

Lemma allows for complex, cross-team Workflows to be composed.
Collaboration is key when building critical workflows and decision-making in enterprises is often federated across people and business units. Each of these roles contributes to a larger workflow, and Lemma is designed to support this kind of cross-functional collaboration so that technology decisions and business outcomes are in lock-step.

Lemma is built for continuous improvement.
Our platform's "human-in-the-loop" guardrails allow you to integrate human judgment into automated workflows. This means you can have real-time oversight and make corrections, ensuring that your AI-powered processes learn and adapt over time. This data can be stored depending on retention policies as a built-in reinforcement learning layer.

Flexibility is at the core of Lemma's design.
The platform supports dynamic, "agentic" workflows that can adapt to changing business needs. Whether you're dealing with structured or unstructured data, or need to handle streaming, batch, or interactive processes, Lemma can support you. Our low-code UI builder and SDK options allow your teams to quickly modify and deploy new workflows as your processes evolve.

By providing a flexible, secure, and collaborative platform, we're not just helping our customers avoid becoming another statistic in a report—we're empowering them to fundamentally transform their operations with AI.

If you’re ready to move to the right side of the divide, we’d love to show you how Lemma can help.