TableFlow V2: AI Teammates for Data Tasks
Today, we're excited to introduce TableFlow V2. We've reimagined how teams handle data by building AI teammates that work alongside you. These agents extract data from messy sources, verify it against internal records, make intelligent decisions, and route the data where it needs to go.
Operations and data teams spend hundreds of hours a month buried in PDFs, spreadsheets, and faxes. They are stuck copy-pasting, double-checking, and cross-referencing messy files against internal systems. It's tedious, error-prone, and wastes the time of some of the smartest people on the team. That is the problem we have been obsessed with solving.
Our answer is AI teammates that can be quickly created, trained, and deployed to handle messy and confusing data tasks. Each one combines pre-made components and is trained for jobs like customer onboarding or shipping reconciliation. We usually start with one simple question: “How would you explain this process to a new teammate?” We capture that workflow and provide the underlying instructions as context, and the agent takes it from there.
Starting with Files
Our journey started with CSV imports. We built the most popular open-source CSV import tool, used by thousands of developers.
But CSVs were just the beginning. The real world runs on messy Excel files with multiple tabs, inconsistent headers, and tables that look like art projects. So we built AI tooling that could pick the right tab, find the table, interpret the columns, and surface the data you actually need.
Then came PDFs, faxes, and images. They turned out to be one of the best use cases for LLMs. Unlike OCR, which requires brittle templates, the latest AI models can read them like a person and understand them with surprising accuracy.
From Extraction to Teammates
At first, we thought getting the data out of these files was the hard part. But even after the data was pulled out, teams were still stuck verifying it, reconciling it with other records, and routing it to the right person or system. As we helped teams tackle these challenges, we built tooling to automate and monitor each step of the process.
We realized these additional tools were just as critical, and together they became the foundation for TableFlow V2. What began as specific solutions for customer pain points can now be quickly added to your AI teammates to handle the full process: verifying extraction results, reconciling records, and delivering clean data into your ERP, CRM, WMS, or whatever system your team uses.
Problems We Solve
We have seen TableFlow thrive in ops and data teams across supply chain, finance, manufacturing, real estate, and more. The common thread is teams drowning in repetitive, manual data work.
Some of the problems we have tackled:
- • Reconciling packing lists against purchase orders line by line
- • Decoding inconsistent, complicated sizing charts
- • Ingesting faxed sales reports (yes, faxes!)
- • Automating catalog pricing updates across tens of thousands of SKUs
- • Entering sales orders from handwritten documents and emails
These are the kinds of tasks that used to eat up hours. Now they take minutes, enabling teams to move faster and deliver better service without adding headcount.
This is TableFlow V2: AI teammates for data tasks.
They do not replace creativity or judgment. They take on the repetitive data tasks no one enjoys. You train them like a new hire, and once they are up to speed, they just work.
Our goal is simple: give every team dealing with messy data an extra set of hands without the extra headcount.
We're incredibly grateful to all the teams who've worked with us to build and refine these capabilities. Your feedback and partnership have been invaluable in shaping what TableFlow has become.
If you're ready to onboard AI teammates that actually understand your data processes, we'd love to show you what's possible.
— Eric & Mitch