August’s Competitive Edge: Legal AI Built on Reducto

“Reducto helped us unlock the last mile of tough legal documents that gave us the competitive edge we needed to provide the accuracy and rigor our customers look for.” - Thomas Bueler-Faudree, Co-founder of August

August is emerging as one of the AI-native platforms to watch in the legal space, especially among the midsize firms it serves. While most legal AI tools offer disconnected features—”draft this clause, summarize that document”—August has oriented itself around something more demanding: complete legal workflows that produce attorney-ready work products. 

The platform works the way lawyers actually work: a client email arrives in Outlook, research happens on the web app, a memo gets drafted in Word, and the response goes back through Outlook, all with full context preservation across every step.

That vision requires a level of document understanding that goes well beyond standard text extraction. August’s customers upload everything from clean email threads to multilingual productions, scanned exhibits, and the kinds of irregular PDFs that circulate in litigation.

What sets August apart is how its output reads. Lawyers consistently report being impressed that August's work product matches their firm's style and reasoning, not generic AI output that needs heavy editing, but drafts they can use with minimal revision. That quality depends on having genuinely accurate document understanding as the foundation.Its platform drafts privilege logs, analyzes productions, assembles playbooks, and works natively inside Word and Outlook, tools where lawyers already spend their time.

For the platform to automate real legal work, it must read all of it—consistently, accurately, and with full traceability back to the underlying documents. When the patchwork of custom OCR, table parsers, and multilingual hacks began slowing the team and compromising precision, August chose the straightforward fix: rebuild the ingestion foundation with Reducto.

Replacing the document stack with Reducto

Before Reducto, August ran a homegrown ingestion stack—a mix of extraction, OCR, and improvised table handling that held together until edge cases began to surface. Multilingual productions from firms in India and Asia weren’t parsed reliably, tables flattened into unusable text, scans and images frequently failed, and without bounding boxes the team couldn’t point lawyers to supporting evidence from the source documents.

“As our usage expanded, the system became high maintenance,” said Dominic Lee, founding engineer at August. “The results weren’t where we needed them to be.” At the same time, August was shipping more agentic workflows in which accuracy and traceability are non-negotiable. 

When Dominic joined, he took over the upload pipeline and led a full rebuild on top of Reducto. The switch was surprisingly smooth: “It was basically a rip-and-replace of our whole upload system,” Dominic said. “And it was pretty painless to implement.”

Documents are now parsed and chunked with Reducto, and stored in Weaviate and DynamoDB for downstream workflows. As volumes grew, August simply moved from polling to webhooks—no re-architecture required. Overnight, the quality of their inputs improved. Reducto delivered:

• High-fidelity OCR and multilingual parsing
• Robust, structured table extraction that preserved rows, columns, and headers
• Bounding boxes for every chunk

Those primitives quickly became foundational, powering August's citation system and grounding the agentic workflows that legal firms depend on. August's technical team has contributed to Reducto's documentation more than a dozen times and maintains daily communication with the Reducto team, a level of integration that gives August access to capabilities and optimizations that other platforms simply don't have.

Automating Discovery and Privilege Review

One of August’s most popular use cases is building workflows for discovery—the phase of a case where teams must sift through enormous productions of emails, message files, spreadsheets, scans, and PDFs to understand what matters and why. Even in routine disputes, a firm might receive up to hundreds of thousands of documents, each of which needs to be categorized, summarized, and woven into the broader factual record.

Privilege review sits at the center of that work and is one of the most sensitive tasks in litigation. Lawyers must determine whether a document is protected by attorney-client privilege, accountant-client privilege, or work-product doctrine, and explain the basis for that determination. Accidentally producing a privileged document can jeopardize a case—or waive privilege entirely—so firms must dedicate time and resources into checking who sent what, why, and in what context.

These workflows are perfect candidates for AI assistance, but only if the underlying document understanding is airtight. Discovery productions contain every imaginable format: clean Outlook messages, multilingual threads, financial models in Excel, tables embedded in PDFs, and the messy 10–15% of scanned or image-based documents that legacy systems often fail to parse. And in privilege review, every conclusion needs to be supported by verifiable evidence in the source material.

“Citations are essential,” Dominic said. “A lawyer won’t trust an answer without being able to verify the evidence.”

This is why rebuilding its ingestion pipeline on Reducto became foundational for August. With Reducto’s multilingual parsing, structured tables, and bounding-box citations, August’s agents can read a full production far faster. They cluster communications, flag likely privileged content, draft rationales, and generate the logs and work product lawyers rely on, all grounded in precise, document-level citations.

Looking ahead

August is building further deeper integrations with Microsoft Word and Outlook, not as standalone features but as a unified environment where AI agents work alongside lawyers throughout their entire workflow. When a lawyer receives a client email in Outlook, starts research on the web app, drafts a response in Word, and sends it back through Outlook, August maintains full context across all of those steps. That level of continuity, which requires both technical sophistication and careful product design, is what enables August to handle associate-level work rather than just providing disconnected tools.

The team's agent-first architecture is designed for this future. While some platforms rely on rigid workflow builders where users manually configure each step, August's agents can plan, coordinate, and execute complex tasks with minimal instruction. As models continue to improve, this approach creates an exponential advantage: tasks that were impossible six months ago are now within reach, and August's platform is already structured to take advantage of them.

While most productions consist of emails and straightforward text files, every matter contains a stubborn 10–15% of documents that legacy platforms like Relativity, iManage, and NetDocs struggle to parse. For August, Reducto handles that last mile of difficult document understanding—the foundation that lets August's agents produce the quality of work that modern firms expect.