Artificial Intelligence / Machine Learning was ranked the seventh most commercially important trend to broadcast and media technology end-users in Devoncroft’s 2018 BBS Broadcast Industry Global Trend Index.  The Big Broadcast Survey (“BBS”) is based on surveys and interviews of technology purchasers and end-users in the global media sector.

Artificial Intelligence / Machine Learning is a newer topic in the global media technology sector.  For instance, neither the NAB Show nor IBC Show have included an exhibitor designation attributable to providing Artificial Intelligence (“AI”) solutions.  The marketing collateral of only 27 exhibitors at the 2018 NAB Show cited AI capabilities, though I expect this will rise substantially at the upcoming IBC Show.

Veritone is an emerging provider of AI solutions in the media sector.  The Company’s recent financial announcements provide a glimpse into the technology supplier perspective on AI usage in the media sector, anticipated adoption rates, and supplier economics.

Veritone’s management has the unenviable task of having to report on the inherently volatile results of a young technology company within the harsh gaze of the public markets – and its quarterly filing requirements.  In its financial reporting, Veritone breaks out the results of its AI platform business.

The AI platform business started offering commercial solutions in April 2015 in the media and entertainment sector.  Though Veritone has since expanded to other verticals, its net revenues from the AI platform business are still primarily from the media sector.  This point was confirmed repeatedly during Veritone’s second quarter 2018 earnings call last week.

The original use case for Veritone’s AI platform in media was monitoring the placement of advertisements and the effectiveness of media campaigns.  A recent (August) investor presentation by Veritone cites the growth in uses cases to activities such as ‘Index and Extend Archived Content’ and ‘Content Moderation and Compliance.’  Several prominent media customers are noted in Veritone’s collateral, including CNBC, ESPN, and Fox Sports.

The Veritone AI platform is a software solution called aiWare deployed in the AWS public cloud.  It then brings together the tasks of ingesting data (and content), orchestrating the processing and analysis of the data, storing the data, and presenting the analytics of the analysis.  The analysis uses third-party cognitive engines (API calls).  In other words, Veritone is providing an orchestration system using the “Artificial Intelligence” libraries of large technology companies like AWS, Google, and IBM (my explanation, not Veritone’s).

Veritone’s investor presentation lists over 200 cognitive engines (third-party API integrations) across 16 cognitive engine categories.  Those categories range from some familiar use cases of translation, transcription, and object recognition to categories not typically associated with cognition such as transcoding.

The business model is a monthly fee (starting as low as $500) for the software orchestration layer, a 30% license fee associated with third-party revenue, and then a metered usage of the “amount of data processed and specific cognitive engines executed.” The quoted text is excerpted from Veritone’s investor presentation in which management also indicates the 30% margin fee on third-party revenue is an ‘expectation.’

How the economics of AI in the media sector evolve is a fascinating data point to track.  My read of the 30% license fee is a 30% margin on top of the on-demand services of organizations like AWS and Google where there is almost total transparency of pricing.  This will then mean media customers will need to countenance a material resale margin on these services or those service providers need to provide for preferential pricing or some combination of the two (equating to 30%).

During Veritone’s Q2 earnings call, management indicated substantially all of the media revenues associated with the AI platform are license fees as opposed to metered usage. There is sound logic in the metered usage approach for pricing AI solutions.  It is generally believed AI predictions will improve with greater volumes of data ingested.  Greater volumes then means greater value and by extension more revenue to the supplier.  Supplier revenue should (or at least ought to) scale with value to the customer.

Commenting on the state of adoption during Veritone’s Q2 2018 earnings call CEO Chad Steelberg said, “The AI industry as we know is still in its early days. The largest and perhaps the earliest vertical to deploy AI for business value is the Media & Entertainment industry.” Continued Steelberg, “If I were to plot this vertical on the product adoption curve, it has crossed the chasm with their first and second evolution in deployments.”

“Crossed the chasm” of course refers to the iconic 1991 book by Geoffey Moore, and within the book to the difficult task of moving market adoption from ‘Early Adopters’ to ‘Early Majority.’ It is interesting to note that transition takes place at the 15% threshold of penetration of technology in a market.  Based on the 2018 BBS data, AI adoption rates are far lower in the media sector.  I don’t believe Mr. Steelberg’s comments was intended to indicate 15% penetration of AI media, but rather a sentiment of greater adoption.

Current revenues of Veritone’s AI platform business reflect low levels of adoption and fast growth – though again off a low base.  Trailing twelve month net revenue for the AI platform business was $3.0 million at the end of the second quarter of 2018.  Veritone doesn’t break out individual market vertical revenue, but did indicate media revenues (which are a majority, but not all of AI platform revenues) were up 250% year-over-year and 25% sequentially in the second quarter.

Veritone’s AI platform processed 2,729,000 hours of data or 311 years’ worth of data.  Total AI platform revenue in the quarter was $860,000.  Since recurring revenue was $214,000 at the end of the second quarter, the balance of net revenue, $646,000, would then approximate metered usage (minimal in media) and third-party fee revenue.   Reducing these figures to a unit fee, equates to a pay scale of $0.23 per processing hour (again this is an estimate).  In the thought experiment where the AI processing replaces humans, the savings would seem considerable.

Management indicated gross margins for AI platform net revenues were on the order of 65% – 70% because of the costs incurred for the underlying cloud platform and third-party cognitive engine.  This then is an almost directly analogous situation to more traditional workflow orchestration and asset management solutions in the media sector.  Those installations tend to have a similar gross margin profile when responsible for integrating the IT portions (storage and servers) of the total solution.

© Devoncroft Partners 2009-2018.  All Rights Reserved.