• by Joe Mandese  @mp_joemandese, January 30, 2024

If 2023 was “the Year of AI” on Madison Avenue, it began for Omnicom in 2018. That’s the year it introduced version 1.0 of Omni, its AI-powered operating system.

And while most big agency holding companies accelerated their AI development in the past year-- especially the deployment of generative AI assistants or related applications -- Omnicom rolled out version 3.0 -- building on the engineering and deployment of 1.0 and 2.0, but leveraging its deep relationships with big generative AI partners to make its interface more intuitive, easier to use, and as Omnicom Digital CEO Jonathan Nelson describes it, “more human-readable.”

The truth is, Omni has been part of a progression for Omnicom to engineer the ad industry’s best-in-class stack to automate -- as well as augment -- the tasks that are better suited for machines to give humans in its workforce “super powers” to do more with less effort.

Omni 1.0 itself was the aggregation and integration of a dozen disparate single-point technology solutions that Omnicom had developed before it, and simply organized it into a single place.

While it had rudimentary elements of AI-inspired ideation to help planners and creatives jump-start their thinking, it was nothing like what Omnicom has developed with 3.0, working with all of the big LLM (large language model) partners, some on a first-mover basis.

Not surprisingly, the primary output of Omni 3.0 is a chatbot, Omni Assist, which gives people the ability to tap into Omni by asking simple human-oriented prompts.

But the back-end is powered by a sophisticated learning and development protocol helping Omnicom’s vast workforce of planners, media and creative people learn as they interact with it.

To date, Omnicom has already onboarded about 50,000 of its 80,000-person global workforce on the platform.

And not surprisingly, media people have been some of the earliest and most active users, but creatives -- and even the most senior managers within the giant holding company -- are also embracing it.

Speaking to Wall Street analysts during a 2023 earnings call, CEO John Wren gave the example of Omnicom’s “multitude of databases” covering its 318-building global real-estate portfolio, estimating it would take a person at least a week “if they worked really hard” to compile a report from that data and give it to him.

"If I type it into OpenAI, I can get the answer by myself in five minutes," Wren said.

Not surprisingly, data is at the heart of Omni -- especially proprietary data that is not replicable in the open marketplace.

“We’re up to 14.4 petabytes,” boasts Omnicom’s Nelson, adding: “Just to put that into perspective, that’s like 550 Libraries of Congress worth of data.”

One of the sources of unique, proprietary data informing Omni has come from a parallel AI-powered technology developed by one of Omnicom’s operating units -- culture consultancy Sparks & Honey’s Q – and another big reason for naming Omnicom MediaPost’s holding company of the year was the way it has integrated Q into Omni. And vice versa.

While Omnicom developers were re-engineering Omni based on the release of generative AI applications such powering Omni Assist, the Sparks & Honey team was doing the same with Q -- creating more frictionless, intuitive, and human-readable ways for its team, its clients, and the rest of Omnicom’s organization to interface with it.

In fact, Q’s design was reskinned this year to blend directly into Omni’s, and users working on either platform might not even know where one begins and the other one lets off.

And that’s by design, because the goal of both platforms is to give Omnicom’s people access to the information they need to generate their own ideas and executions. Omni and Q just give them a running start -- a quick, rough draft, if you will to build on, iterate and perfect.

Both systems are capable of generating briefs, rudimentary creative outputs, and even the precursors to media plans and allocation models.

Omni, in fact, can actually help people build their own models. One of 3.0’s features is a “taxonomy builder” that enables users to generate their own, on-the-fly taxonomy drawing from a virtually unlimited number of possibilities.

“You can use it to drive a thousand-node taxonomy or a million-node taxonomy,” Nelson notes.

Meanwhile, the integration of Omni and Q represents a unique proprietary advantage for Omnicom, because it literally infuses culture into an organization that might otherwise have a more limited -- albeit focused -- view of the world.

Q’s database is already massive and growing by leaps each year, scraping the most meaningful cultural signals identified by Sparks & Honey and organizing them into a proprietary taxonomy known as “the elements of culture” developed by Sparks & Honey. (Think periodic table, but for culture instead of physical elements.)

Meanwhile, the Sparks & Honey team has begun rolling out three new generative AI-enabled applications giving Omnicom’s team easier and more intuitive access to all its data.

Not surprisingly, the first one is a chatbot dubbed Q Assist, in which users can ask simple language questions to generate outputs.

While that might seem like table stakes coming out of 2023, it is part of a suite of apps making the databases even more accessible.

A second new Q app, “Q Sparks,” surfaces cultural signals that can spark human thinking, including Big Ideas.

“In the past, people had to write the booleans [computer-readable expressions] and it would give you all kinds of data, but you had to figure out what to do with it,” explains Sparks & Honey founder and CEO Terry Young. “This is making the leap for you. And if you don’t like the first leap, just hit regenerate and it will give you new ones.”

A third new Q application, dubbed “Q Intelligence,” mines Sparks & Honey’s massive database of its weekday culture briefings (see one forecasting generative AI below) and organizes them based on themes, cultural signals, and product categories based on relevance.

Sparks & Honey’s culture briefings -- already unique, arguably public-service features within the advertising industry -- are open to anyone to view on the agency’s social-media pages, and feature human Sparks & Honey analysts, guest speakers, and a notable third party, Q itself, discussing and identifying trends and making predictions for every imaginable cultural trend.

At launch, Sparks & Honey rolled out Sparks Intelligence with a variety of “vertical” marketing category themes such as “health” and “automotive,” enabling users to quickly dive into a culturally relevant context to help people jump-start their thinking process.

As important as Omnicom’s AI-powered tech development may have been during 2023, it was actually the acquisition of some data that cast MediaPost’s deciding vote to name it Holding Company of the Year: Its $900 million acquisition of ecommerce platform Flywheel Digital from Ascential plc.

While Flywheel was a good get as a complement to Omnicom’s own burgeoning digital commerce and retail-media solutions, it really was about the data the company has accrued over the past decade, giving Omnicom a unique proprietary view of purchase behavior across all the major platforms.

"We are the biggest source of information in marketplaces," Flywheel CEO Duncan Painter told Beet.tv earlier this year.

Painter, who was CEO of Ascential when it sold Flywheel to Omnicom and joined it as CEO of Flywheel, added: “If everyone thinks that someone like Amazon has a lot of data, what they have to remember is that for us to do our job for clients, we have to take it for every major marketplace in the world. So we collect and get the data for all the trading, media, inventory, supply chain across-the-board -- customer satisfaction -- of the top 80 marketplaces in the world.”