Dozens of ideas and initiatives will undergo rapid prototyping for viability inside Seven West Media’s newly created AI Factory with personalised ad loads tailored to viewing patterns and time spent on 7plus among a handful of early use case tests with an embedded Databricks team at Seven HQ. 

Another crucial AI “sprint” underway currently is an ambition to re-engage 10 per cent of “sleepy users” with personalised communications and content among the circa 14 million IDs 7plus has in its database. 7plus typically has 4-5 million users in a 30 day period.

The AI Factory is tasked to identify a combination of market-facing innovation and faster internal delivery of business, content and audience insights and recommendations through bespoke large language models. Ultimately the hope is the machines will uncover and facilitate diversification of revenues beyond advertising, although ads are snapping much of the early focus. 

"The AI factory is about spinning up as many pilots as possible in the next 12 months,” Seven’s Director, Data and Growth, Andrew Brain, told Mi3. “We might not use 70 per cent of what we develop, but that's 70 per cent of learning to get where we need to be. It's about democratising data across our business – we've got the content team, marketing, sales, finance and it’s essentially all our people across the business to actually get data quicker so they can go 'I've got that insight now. That's what I needed. Now we can go and make a decision off the back of that'.”

Personalised adloads is one of the early tests in planning. “We want to deliver ad loads based on your engagement score,” Brain said. “If say your session duration is two hours, why would you get the same ad load experience to [someone who] only came in for half an hour?” Brain said the ad load sprint would determine if Seven had “the data points” and technical capability to build it. “Will we be delivering in two weeks? Absolutely not but we're going to spin that in a sprint.”

Reactivating sleepy 7plus users was another priority project for the AI Factory – Brain said Seven wanted to develop AI capabilities that could identify users who hadn’t engaged with 7plus in recent months, analyse active users and model their patterns to lookalike sleepers and build automated communications.

“We’re actually modelling on the audiences who are active,” Brain said. “We're looking at what they absolutely look like from all the different attributes.” It included optimal times to target the sleepers based on their previous viewing patterns. Brain said the intent was for AI to select content that is likely to appeal to a lapsed user based on their previous viewing habits and the habits of similar active users and assemble that in personalised communications or the experience they have when returning to 7plus.

“We want to bring back from a sleepy state around about 10 per cent of audiences,” Brain said.

If it can, “it's worth tens of millions of dollars for us."

Seven also plans to open the AI Factory for collaboration with agencies and advertiser projects. 

The AI Factory use case trials being planned for the next 12 months include:

• Predictive audience modelling: If a brand wants to run a campaign targeting women aged 25-34 over the next month, the system predicts audience demographics and engagement for the next 28 days with 94 per cent accuracy.
• Content scheduling optimisation: When Seven needs to maximise viewership for a new drama series, the machines analyse historical viewing patterns and predict the optimal time slots for releasing new episodes, potentially increasing viewership by identifying when the target audience is most likely to be engaged.
• Personalised ad loads: The system adjusts ad frequency based on individual engagement scores. The long-duration viewer might see fewer, more spaced-out ads, while the short-session viewer gets a more condensed ad experience, optimising both user experience and ad revenue.
• Audience re-engagement: Seven identifies a significant number of users who haven't accessed the platform in the last three months. AI analyses the viewing history and preferences of these lapsed users, then generates personalised content recommendations and targeted notifications to re-engage them, aiming to bring back 10 per cent of inactive users.
• Revenue forecasting: Sales teams need to set targets for the upcoming quarter – the machines analyse trends in direct bookings and SSP (Sell Side Platform) data, providing a detailed forecast of expected revenue, allowing for more accurate goal-setting and resource allocation.
• Content affinity prediction: Seven might be considering purchasing rights to a new international series. The machines dissect viewing patterns, content metadata and audience preferences to predict how well the new series will perform with Seven's audience.
• Dynamic programming guide: A user opens Seven's streaming app looking for something to watch – AI generates a personalised programming guide based on the user's viewing history, current popular content, and predicted interests, potentially increasing watch time and engagement.
• Automated Insights generation: A Seven exec needs quick insights on the previous night's primetime performance – using natural language processing, the executive can ask: ‘How did our 8pm show perform last night?’ and receive an instant report with key metrics and comparisons to historical data.
• Cross-platform viewing prediction: Seven wants to optimise its content strategy across both broadcast and streaming platforms. The machines tap viewing patterns across all platforms to predict how audiences will split between broadcast and streaming for different types of content, informing programming and advertising strategies.
• Churn prediction and prevention: Seven wants to reduce subscriber churn on its streaming platform – patterns that indicate a user is likely to churn are identified and personalised retention strategies such as targeted content recommendations or special offers are recommended.

The next 12 months will prove whether all of that actually moves the needle for the new regime.