Veritone wants your archive to earn its keep

Hockey Canada had a library with a peculiar filing system: part of it lived inside people’s heads. The footage and photographs were spread across systems and hard drives. Knowing where to look could depend on knowing whom to ask. For an organization with six decades of sporting history, that is a rather fragile arrangement. A puck is supposed to be hard to catch. A photograph should be easier.

The library in four people’s heads

Veritone’s account of the Hockey Canada project describes more than 80 terabytes of material. The problem was access. Communications, marketing and partnership teams needed specific moments, while the archive’s informal custodians supplied the map. Jennifer Robins, Hockey Canada’s director of creative services, described the old arrangement as “four people having the library in their heads.”

Digital Media Hub replaced that dependence with a shared platform. Veritone says automated tagging, transcription and visual recognition made the collection searchable across teams. Its customer account reports finding moments in seconds rather than hours. Treat that as a reported result from this deployment, rather than a stopwatch guarantee for every archive. Still, the underlying change is easy to grasp: knowledge about the files became something colleagues could query.

This is a good entrance to Veritone because it begins with an ordinary irritation. A file exists. Someone needs it. Between those two facts lies an afternoon. The company sells ways to shorten that afternoon, then attaches other jobs: distributing a clip, checking its rights, preparing evidence for release, or converting media into data that another AI system can use.

A switchboard for artificial intelligence

Chad and Ryan Steelberg founded Veritone in 2014. Their earlier ventures included digital advertising and broadcasting businesses. That background helps explain why recorded media runs through the company’s story: Veritone was dealing with audio and video before generative AI became a household conversation.

Its central product, aiWARE, coordinates multiple cognitive and generative AI engines. A workflow can ingest media, apply processing, index the results and make them available to applications. Different jobs call for different capabilities: speech recognition for a transcript, visual analysis for an image, text analysis for what was said. Veritone’s proposition is that customers should be able to combine those capabilities without rebuilding the whole system around each model.

Developers can work through APIs; Automate Studio offers a low-code workflow designer. Digital Media Hub puts media management and discovery in front of people who need footage. Public-sector applications put related processing inside evidence workflows. The expertise being sold includes integration and deployment, as well as the software interface.

There are alternatives. Dalet Flex handles media asset management; Axon Evidence handles digital evidence. Organizations can also assemble their own pipelines. These overlap with different pieces of Veritone’s portfolio. Veritone’s distinguishing argument is the combination of model coordination, industry applications and services. Whether that combination earns its fee depends on the job and the systems already in place.

Same files, different obligations

Consider two video recordings. One contains a championship highlight. Another contains an encounter recorded by a police camera. Both can benefit from transcription and search. Their owners face very different questions about what happens next.

For a sports archive, discovery can lead to a sale. On September 29, 2026, Veritone announced a multi-year agreement with ATP Media covering North American licensing of ATP Tour archive content. The arrangement includes archiving, cloud hosting, AI processing and licensing. Broadcasters, brands and production companies gain another route to footage for documentaries, campaigns and other creative work.

The NCAA relationship supplies another example. Veritone’s Digital Media Hub account describes championship material on deteriorating tape, followed by digitization and indexing so people can locate athletes, sports and moments. The commercial opportunity requires rights management alongside retrieval. Finding a beautiful shot does not grant permission to put it in an advertisement. Search brings the asset into view; permission determines its possible career.

An archive can be full of value and still be useless on deadline.

THE PRACTICAL PROBLEM

For public agencies, the next step may be redaction. Veritone Redact automates parts of removing sensitive information from recorded material. The company’s Escondido Police Department account says an hour of video previously took five to ten hours to process, with complex cases taking much longer. It reports that one person now delivers output previously requiring two. Pasadena’s published customer account describes reducing a two-week redaction task to a few days.

These are company-published case studies, so they belong to their particular workloads. They make the labor problem visible without proving that every agency will achieve the same saving. A buyer should examine the complete process: detection, correction, review and release. Sensitive material still demands an accountable decision about what may be disclosed. The Intelligent Digital Evidence Management System, or iDEMS, addresses the broader management and analysis of evidence.

The price of becoming an AI company

Veritone’s business has grown in several directions. In 2021 it announced the PandoLogic acquisition at up to $150 million, including contingent payments; the initial consideration included $50 million in cash and $35 million in shares. Broadbean followed in 2023, with subsequent reporting recording $53.3 million in purchase consideration. Together, these businesses extended Veritone into programmatic recruitment advertising and job distribution.

Expansion also created a portfolio that required explanation. In October 2024, Veritone sold its advertising agency, Veritone One, to an Insignia Capital affiliate. The announced consideration was up to $104 million, including a potential $18 million earn-out. That headline was conditional. The 2025 annual filing says the revenue targets for the earn-out were not met. Management presented the sale as a way to concentrate on enterprise AI and improve liquidity.

The financial test continues. In the second quarter of 2026, revenue was $24.3 million and the net loss was $22.2 million. Annual recurring revenue was $62 million. Recurring revenue gives a view of contracted business; it does not settle the question of profitability. The quarter’s revenue growth and loss deserve to be read together.

A June restructuring filing anticipated a workforce reduction of at least 25% relative to March 31, 2026. By August, Veritone reported $11.3 million in annualized cost reductions from the first phase. On October 1 it announced a registered direct offering expected to raise roughly $15 million gross, with an expected October 2 closing subject to conditions. The stated uses included debt repayment or restructuring, working capital and general corporate purposes.

Those decisions show the cost of pursuing the strategy. They also temper the company’s cheerful assertion, “AI makes people better.” That is an ambition. Customer outcomes and company economics require their own measurements.

Start with the search you cannot finish

What should a prospective customer borrow from this story? Start with a task that is already expensive. Locate an old interview. Prepare a recording for release. Get a licensed sports clip to a producer. Use representative material and define what a successful result looks like before the demonstration begins. Count the time spent reviewing errors, rather than stopping the clock when the software produces an answer.

The commercial model also matters. Veritone combines subscriptions, usage-based processing and storage, professional services, and managed content licensing. Data Refinery advertises per-token pricing for turning unstructured media into AI-ready data. Digital Media Hub contracts depend on negotiated terms. A practical comparison therefore includes migration, integration, processing volume and ongoing review, alongside the subscription bill.

The case is strongest when an organization has a substantial collection, repeated demand and a clear use for the results. Small archives with occasional searches may offer too little labor saving. Poor recordings make analysis harder. Unclear ownership can prevent a licensing deal even when the search works perfectly. These are purchasing conditions, not small print.

Veritone’s September 2026 showcase added Video Intelligence and semantic-search developments to the media proposition. The useful test remains wonderfully unglamorous: can the person with a deadline find the right thing, establish what they may do with it, and finish the task? A company can own decades of history. Making that history available is another business entirely.