The Missing Infrastructure for AI Agents: Documents | Reducto’s Adit Abraham
October 21, 2026
AI Summary
5 min readThe Hidden Complexity of Making AI Agents Read Documents
In 2023, the day Adit Abraham and his co-founder got into Y Combinator, they told their partner they were pivoting. The problem they landed on was deceptively simple: nobody thinks about complexity when they upload a document to ChatGPT—they just expect it to work. But the tools available to solve document processing were suddenly much more powerful than before, thanks to vision language models. Abraham, now CEO of Reducto, describes documents as "abstractions and artifacts of human work." White-collar work across healthcare, legal, finance, and insurance lives inside PDFs, spreadsheets, and scans. If agents are going to do that work, they need to read those documents with the same fidelity a human would—because "it's not like an oops to say, oh, I messed up a checkbox. It's like, oh, that's a patient's outcome."
Why Traditional OCR Isn't Enough
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of a16z Show
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (02:02) **From Memory to Documents: The Origin Story** - Adit explains why a team focused on memory pivoted to document processing when they realized agents would need to read and write the artifacts of human work.
- 2 (03:57) **Why Traditional OCR Isn't Enough** - Adit breaks down the three epochs of document processing and explains how vision language models (VLMs) change the game.
- 3 (06:22) **Building a Specialized Model (R1)** - How Reducto post-trained its own foundation model to get better accuracy, speed, and cost than off-the-shelf models.
- 4 (08:50) **Accuracy Thresholds and the Human Loop** - Why each magnitude of accuracy unlocks new use cases and how agents can now beat humans at structured extraction.
- 5 (10:46) **Cost as a Scaling Lever** - At one cent per page, what becomes possible when processing is no longer a bottleneck.
- 6 (13:48) **The Harness Beyond Parsing: Deep Extract and Workflow** - Reducto is moving from pure parsing to building tools for evaluation, review, and workflow construction.
- 7 (15:09) **Why the Labs Won't Solve This** - Adit explains why frontier model labs won't go deep into document infrastructure, leaving room for a focused company like Reducto.
+ Full timestamped outline available in the app
Show Notes
AI agents can reason over enormous amounts of information, but much of the world’s actual work is still trapped in PDFs, spreadsheets, scans, and other formats that models struggle to reliably understand. Reducto co-founder and CEO Adit Abraham argues that solving this document layer is essential if agents are going to take on consequential work in healthcare, legal, finance, and insurance.
Adit joins a16z General Partner Jennifer Li to explain why traditional OCR breaks down on messy real-world documents, how Reducto combines vision-language models with specialized infrastructure to reach high accuracy without sacrificing cost or latency, and why frontier models can silently drop information on long extraction tasks. Reducto now processes billions of pages each month, including its seven billionth page the morning of the recording.
They also discuss Reducto’s R1 model, agents that can review and correct their own document extraction, and the company’s push beyond simply reading files toward letting agents reliably work with and edit them. Underneath it all is a higher-stakes problem: when a misread word can affect a healthcare claim or financial decision, getting the input right matters just as much as making the model smarter.
Resources:
Learn more about Reducto: https://reducto.ai
Follow Adit Abraham on X: https://x.com/aditabrm
Follow Jennifer Li on X: https://x.com/JenniferHli
Stay Updated:
Find a16z on YouTube: YouTube
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Show on Spotify
Listen to the a16z Show on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.