Reducto + Ribbit

Our Token Thesis

Reducto can become foundational infrastructure for the Token Revolution

Tokens = Missing Link

Today's digital systems struggle to capture context-rich, unstructured data like conversations, PDFs, or spreadsheets that human advisors intuitively understand. Tokenization bridges this gap. Tokenizers convert messy inputs into structured, actionable data, enabling LLMs to grasp critical context they'd otherwise overlook. Though digitizing documents has proven challenging over the past decade, Reducto's technology positions it uniquely as a universal data API, potentially eliminating millions of hours spent weekly on manual data management and unlocking entirely new workflows.

From Tool to Teacher

AI agents will increasingly depend on high-quality tokens to complete complex tasks. At first, humans will prompt agents. But soon, agents will rely on humans to generate tokens (feedback, exceptions, edge cases) that improve reasoning and accuracy. This shift creates a new labor model: experts become token suppliers, training agents through real-world decisions and earning rewards when their logic is reused. In this future, tokens aren't just data, they're a new kind of work product with compounding value. Reducto will be at the forefront of this.

Expert Tokens Are the Future

The most valuable tokens won't come from the open web but from experts doing real work in high-context domains like finance. Workflow tools will turn this activity into structured training data, making professionals the primary source of learning for AI. As agents improve, they shift from passive tools to active collaborators—surfacing insights, prompting decisions, and automating actions. The flywheel is simple: more expert input leads to smarter agents, which drives broader use and more input. Reducto will be the critical enabler that allows companies to gain an advantage by capturing this domain-specific context faster.

Read Ribbit Token Thesis

Some Companies We'd Love to Introduce

What Your Customers Are Saying

Customers are big advocates of Reducto

Best at Complex Docs

Customers consistently highlight Reducto's unmatched performance handling complex document layouts—including tables, checkboxes, and handwritten inputs—surpassing AWS, Azure, Google, and niche startups. Its OCR paired with semantic comprehension dramatically improved accuracy, significantly reducing manual intervention and outperforming previous solutions that typically reached only 30–50%.

High Speed to Trust

Despite initial hesitations due to its early stage, customers rapidly adopted Reducto across mission-critical operations, from sales enablement and procurement intelligence to analytics. Exceptional founder-led support, agile product iteration, and scalability solidified trust. This deep integration positions Reducto as increasingly indispensable and difficult to replace, fueling continuous expansion into new workflows.

Areas of Feedback

Customers emphasized that Reducto could improve by adding automated schema setup, better confidence scoring, and clearer intermediate outputs. They also highlighted parsing improvements for multi-page tables, checkboxes, Excel sheets, and mixed media. One Customer specifically noted willingness to pay for auto-tagging, reusable schemas, and greater control over backend LLM selection.

Detailed Call Notes

Complete transcripts and summaries from recent customer conversations

Customer #1

Customer #2

Customer #3

Customer #4

We Are Excited By What We've Heard

A few things we've heard on podcasts and interviews that have us excited

Default Ingestion Layer

We love the vision of Reducto becoming the ingestion team for AI applications, offloading the messy work of parsing PDFs, spreadsheets, and scans into clean, structured data. As enterprises adopt agents and LLMs, there is a lot of value in becoming the unified ingestion pipeline, staying at the forefront of what high-quality, AI-ready inputs look like across formats, use cases, and increasingly complex model demands.

Not Over-Verticalizing

We like Reducto's decision to stay horizontal, rather than going too deep for verticals too early. By building a flexible system that handles PDFs, spreadsheets, and scans across industries, Reducto avoids the constraints of vertical-specific tooling. This broad scope allows its models to improve faster through diverse inputs, and strengthens their position as the foundational layer for AI products working with unstructured data.

Unlocking Possibilities

We're excited by Reducto's ability to unlock use cases companies didn't think were possible like processing scanned documents with no metadata or handling hundreds of pages in seconds. Customers are now building features they previously couldn't support, thanks to Reducto's accuracy and speed. Reducto's ability to power entirely new product capabilities across AI applications creates a powerful advantage.