It’s an exciting time to be working in data and technology. The innovation and creativity facilitated by LLMs is astonishing, creating for me, at all the same time, a feeling of excitement and enthusiasm, with a healthy dose of anxiety and caution. The first question seemed to be the hardest. But then something happened. It was intuitive. You didn’t need to be taught how to use it.
Using these new prompts has reawakened a curiosity, a renewed enthusiasm for learning and understanding. There is no dumb question, and always an enthusiastic response. Interestingly, early adopters are from a broader range of people that might normally be caught early in a tech revolution.
I’m not one for hype, but from someone who experienced the first wave of the WWW, this does feel as significant, and possibly more revolutionary. Time will tell.
With this backdrop we are writing our first post for Ozone Labs. Our first post will lay out our thesis for Ozone Labs, why we are doing this, how we are harnessing our own collective curiosity and enthusiasm, and what you can expect.
Ozone is a data and technology company
Ozone is an advertising platform. We trade in deep audience understanding, advertising activation and measurement. Under the covers we are mostly a data and technology company. Over half of Ozoners are in data roles; data science, engineering, analytics and product. And like any other company like us, machine learning models and data science underpins most of our insights and activations.
The disruptive world of the prompt
Everything we do at Ozone operates at the crossroads of content creation, audience engagement in that content, and marketing to people consuming the content. Without doubt three worlds most likely to be upended by LLMs. As we discuss and debate internally and with customers how best to navigate this new world, it’s easy to focus on the risks, but there are also exciting opportunities. What’s certain is a high level of disruption while adjusting to a new normal.
- Audience engagement: One thing that ChatGPT did for many was highlight how degraded the search experience had become. Clicking on links in Google search felt like going into battle, with dubious content that jumped around the page as ads loaded that you tried not to click on as you navigated around them. By comparison, a clear precise answer to a prompted question seemed like the wholly grail. LLMs have their own shortcomings, and need to find their place in the battle for our attention and time, but even at this early stage they will play a role in most of our lives.
- Content creation: The business of publishing in a digital platform world is complex. The cost of creating journalistic content is high, and while publishing in direct channels creates the most commercial value, consumers spend most of their time on social platforms making it difficult for publishers to monetize. And while search isn’t a great value exchange for publishers, at least there is one. Today, the prompt value exchange is far from clear and far from democratic or transparent. But it’s unlikely the LLM commercial model of today is where we will settle, and publishers hold a very strong hand in ensuring they are fairly remunerated for their investment in quality content.
- Marketing: The introduction of LLMs is already transforming marketing automation. Data will drive most marketing decisions and platform integrations will deliver most media activations. The planning shift from traditional media channels to platforms to reach audiences has already happened, Google, Meta and Amazon capture 70% of global ad spend. OpenAI is pitching itself as the next global platform. What data marketers can get back from the platforms will matter: a strong feedback loop will accelerate growth, limited signal = limited control and learning.
Growing together - the importance of open standards and development
The introduction of LLMs and generative AI has created a pace and energy to product development. Many features and innovations that seemed in the distant future, or unlikely, are all of a sudden possible and doable today.
While many Ozoners have been on their own personal journey since we introduced Claude Code to product and engineering, and more recently Claude Desktop and Gemini to the company , one learning that has resonated is how working together has accelerated the learning and progress. Hackathons that we used to hold a few times a year, we’ve increased to quarterly, and more recently involve not just product, engineering and data folks, but business stakeholders, closing the gap between technical prototyping and solving live business challenges. Collaboration and working together is a core Ozone value. It’s in our DNA as a company founded as a joint venture between publishers.
At such an early stage in this new technology cycle, this is the perfect time for open discussion to debate standards and risks, industry governance, and developing open frameworks and tools to build on the great work of others. The release of AdCP, XX, XX is a great support for some of the new initiatives technology innovation is spawning, facilitating working groups and lots of Linkedin posts.
That said, it feels like the wider media and adtech industry has lost the art of collaboration. Beyond the standards setting and discussion within JICS, how companies are experimenting with these new technologies and standards, what they are learning, what’s working and what’s not working, seems harder to come by. The result is that we all end up working on the same stuff, every company in the same gear at the same time, with no easy way of building on the shoulders of previously great work and moving the whole market forward.
It’s for this reason we are excited to be launching Ozone Labs.
In the spirit of collaboration, through Ozone Labs we intend to share many of our ideas and concepts and prototypes that are not fully formed but that we think are interesting and that others may want to know about, adapt and build on for their own discovery.
Ozone Labs is our channel for sharing our experiments into what the future of publishing, audience engagement and marketing could be, in a world of LLMs and agents.
Ozone Labs is our investment in creating a culture of testing, learning and collaborating with our customers and peers.
Three interconnected Labs - Research, Labs Live and Projects
Work within Ozone Labs will focus in three areas:
- Research: Benchmarking publisher content usage by LLMs
- Labs Live: Partner hackathons and deep dives
- Projects: A collection of prototypes, concepts we are working on
Research: Benchmarking publisher content usage by LLMs
Initially, our research efforts will be focusing on how publishers' content is sourced and used in an LLM prompt ecosystem. The value exchange between publishers, their readers and advertisers was clear in an analogue world. The shift to digital channels and addition of Google and social platforms radically changed the value exchange resulting in antitrust and monopolistic behaviour. The introduction of LLMs is forcing a re-write to how published content is accessed through digital platforms.
The output of this work will be a series of benchmarks and white papers. Building on Ozone’s LLM experimentation platform Athena, we are exploring value creation, configuration and different commercial models of the LLM ecosystem from a publisher perspective, that will produce:
- Benchmarks: A series of benchmarks that compare LLMs in different handling of news content with different prompt profiles
- Studies and white papers: Investigating topics such as accuracy, authority, and grounding
There will be some deep-dive research programmes we will undertake with some publishers that are unique to their business, on a case-by-case basis.
Labs Live
Bringing to life the value of rapid collaboration and prototyping with our partners. We team up with brands, agencies, publishers, and engineers to co-create hackathons. Same table, same sprint, shared outcomes. The best ideas happen when different worlds collide. The principles of each event are:
- Pitch ideas: Anyone can throw an idea on the table — a hunch, a frustration, a “what if.”
- Form teams: Self-organise around the ideas that excite you. Engineers, commercial, data — all welcome
- Build fast: Focused sprint, usually 1–2 days. Working prototypes, not presentations
- Demo and decide: Present to the group. The best ideas move into the Lab for further development
Each hackathon will be specific to each partner and designed to push the boundaries of what’s possible. And to have fun and to learn a lot along the way.
Projects
Not all great ideas make it to products, but a lot can be learned along the way. Projects are a collection of things we are working on that might be a great idea, or could become a great product.
Some projects will be the output of Labs Live events that continue to have a life of their own. Some will be open source repositories we think are useful that others can fork and use.
Some will be our tests and evaluations of the many innovations that are in the ad-tech market such AdCP, arTF.
Read, collaborate, share
In the first phase much of the content you will see will be from Ozoners, sharing our work - hypotheses and opinions, whitepapers and studies, prototypes and benchmarks - which at times it may seem a little unstructured, but will all be focused around our three themes of audiences, content and advertising.
As we get feedback and start working with customers and partners, we will start to share collective work - summaries and outputs from joint projects and hackathon days.
As much as possible we will open source anything we think that would be useful for others to use, fork, build upon. We hope this will be where others can share back, fostering a collaborative culture that we all benefit from.
Any updates to projects will be updated on our Ozone Labs blog, so if you want to stay up to date on the work we are doing here sign up to the blog/newsletter.
We are excited by what’s possible.