Valora Technologies

Somewhere right now, a corporate file share is quietly metastasizing. Twenty years of contracts, half-finished memos, scanned faxes, and folders ominously named "FINAL_v7." Nobody knows what is in there. Nobody wants to look. Valora Technologies looks.

From a building in Westford, Massachusetts, a team of roughly seventeen people runs software that reads other companies' documents - all of them, cover to cover - and tells those companies what they actually own. Not the file names. Not a sample. The full text of every file. It is the kind of work most organizations would rather pretend is unnecessary, right up until a regulator, a lawsuit, or an AI project forces the question.

Above: the unglamorous truth of the modern enterprise, rendered as a pull quote so it feels important.

Dark data, and the cost of not knowing

Here is the inconvenient arithmetic. Every company generates documents far faster than it can ever organize them. The result is "dark data" - information you possess but cannot see. Most of it is ROT: redundant, obsolete, or trivial. Some of it is radioactive: personal data, privileged communications, regulated records hiding in a folder no one has opened since a previous decade.

For years, the standard response was to hire people to read and tag files by hand. This works beautifully, provided you have unlimited interns and no deadlines. For everyone else, manual review collapses somewhere around the first terabyte. The data keeps growing. The humans do not.

Valora's bet was that classification is a machine problem wearing a human costume. If software could read a document the way a trained records manager would - understanding what it is, who it concerns, whether it is sensitive, whether it can be safely destroyed - then governance could finally keep pace with the data. That bet has a name: AutoClassification.

An MIT marketer and a chief architect walk into a data problem

Valora was founded in 1999 by Sandra Serkes and Aaron Goodisman. Serkes, the CEO, arrived with a background that reads like a tour of hard-to-automate fields: speech recognition, computer telephony, document processing, and analytics, plus degrees from MIT and Harvard Business School. Goodisman, the CTO, became the chief architect of the platforms that would eventually carry names like PowerHouse and BlackCat.

Their timing was either prescient or stubborn, depending on how you keep score. In 1999, "machine learning for unstructured data" was not a pitch deck phrase; it was a research footnote. They built the company through the eDiscovery boom - the legal world's scramble to find relevant documents during litigation - which taught Valora to be thorough in a way that few software companies ever need to be. When you classify documents for a lawsuit, "mostly right" is not a category.

That eDiscovery heritage is the quiet throughline. It is why a small Massachusetts firm could later be trusted by enormous global companies: it had spent years being held to courtroom standards of accuracy.

PowerHouse reads. BlackCat shows. Connectors reach.

The Valora platform is three pieces that pretend to be one. PowerHouse is the engine. It connects to a repository through its API, scans the contents, performs a full-text analysis of each file, then applies customized classification tags and automated disposition rules. It does not care whether your data lives in the cloud or on a server in a closet, whether it is structured or unstructured. It reads it in place.

BlackCat is the part humans actually touch - a metadata interface of charts, reports, and collaborative workflows that lets teams approve disposition manually, in bulk, or fully automatically. Connectors are the custom plumbing that links PowerHouse to wherever the data happens to be hiding.

PowerHouse comes in three tiers, and the difference is raw speed. Starter handles about 6,250 files an hour. Foundation roughly triples that. Enterprise reaches around 50,000 files an hour. The chart below is the entire sales pitch in one picture.

Who actually trusts a 17-person company with their secrets

Big ones, as it turns out. Valora was engaged by a consortium representing five of the largest oil and energy companies in the world to AutoClassify and manage their contracts - the sort of documents where a misfiled clause can be expensive in ways that make headlines. A separate multinational oil and gas company brought Valora in for a stack of information governance work, including the deeply unglamorous and deeply necessary task of answering GDPR data subject access requests.

The use cases fan out from there: data discovery and classification, ROT and file clean-up, records management, data privacy and GRC, legal hold and eDiscovery, and increasingly, getting messy enterprise data into shape for AI. Each one is a variation on the same task - know what you have, keep what matters, defensibly delete the rest.

Turning data from a liability back into an asset

Strip away the acronyms and Valora's mission is almost old-fashioned: help large organizations understand, manage, repair, govern, protect, and report on their data. The company frames its approach as "innovative technology, expert support, and tailored solutions," which is the polite way of saying the software is sophisticated but a human still picks up the phone.

There is a real principle underneath. Data you cannot see is data you cannot protect. The sensitive document you do not know you have is the one that leaks. The record you should have deleted years ago is the one that surfaces in discovery. Valora's entire mission is to make the invisible legible - and then to act on it responsibly.

The AI era runs on data nobody cleaned

Every organization now wants to feed its data to a model. Few of them have read that data. They are about to discover what Valora has known since 1999: that an enterprise's documents are a beautiful mess, riddled with duplicates, sensitive material, and outright garbage. Train on that, and the model learns the mess. AutoClassification - identifying, tagging, and curating data before it ever reaches a model - turns out to be the unglamorous prerequisite for the glamorous AI everyone is chasing.

So return to that file share - the one quietly metastasizing at the top of this page. Twenty years of contracts and "FINAL_v7." After Valora runs through it, the picture changes. The duplicates are gone. The personal data is flagged and protected. The records past their retention date are defensibly disposed of. What remains is something a company can actually use, govern, and trust. The folder nobody wanted to open becomes the asset nobody knew they had.

That is the whole story. A small company in Massachusetts spent 25 years reading the documents nobody else would - so that everyone else finally knows what they own.