AI-assisted personalisation and monetisation, optimising CDNs, examining security solutions to safeguard emerging threats in streaming, dynamic broadcasting scalability and streaming flexibility were some of the topics regional CTOs discussed at the recent BroadcastPro summit. We bring you a summary.
AI has had a significant impact on the media and entertainment industry, driving production efficiencies, improving recommendations and providing advanced insights into audience metrics, among other benefits. New innovations push the boundaries of technology and regional players are adopting processes that enhance and streamline automation, adding value to the overall personalised consumer experience.
Spearheading a power-packed session on the latest and greatest in data and tech innovations at the BroadcastPro Summit were Suhail Ahmed, CEO of Mediability, alongside Brad Eliot, Chief Technology Officer, IMI; Faraz Arshad, CTO, StarzPlay; Israel Esteban, CTO, beIN Media Group; Romary Dupuis, Chief Data Officer, Viaccess-Orca; and William Sharp, CTO, Intigral.
Predicted to grow at over 25% a year to $99bn by 2028, AI within the media and entertainment industry is reshaping how content is delivered, personalised and enjoyed. Used for purposes from video editing and advanced effects to selecting winning scripts, organisations use AI-assisted processes to increase retention and revenue.
“We are trying to expand existing models to provide more personalised content, based on viewership and geography. Earlier, metadata was expensive, but now with advanced GPUs and the reduced cost of cloud, we’re able to put metadata into recommendations and surface a greater depth of the library to our customer,” said William Sharp, CTO, Intigral.
Technology is also driving the timeframe for serving ads – how many and when – as well as for being more reactive in real time and “dynamic in terms of monetising specific customer segments. In terms of content, do we want automated content recommendation carousels? Content editors would like to see their personal input for people to watch, and along with the LLM models bring forward a depth of content.”
“The move towards AI has brought about a noticeable shift in engagement and consumption, and our reporting, including the advanced churn prediction and ROI models, is influenced by this transformation. We are using AI while serving live sports engagement/predictions, compared to FAST and SVOD/ AVOD monetisation, with a more realistic and balanced approach,” said Faraz Arshad, CTO of StarzPlay.
“We developed a hyper-personaliser with an MVP consisting of a couple of carousels, and we can clearly see the shift in engagement, improving the LTV, which is the main driver for the entire OTT industry. Within three months of going live with AI-driven personalisers, we saw a 400+ percent uplift in consumption, which is phenomenal.
“Refreshed and fully personalised for each user session, data was gathered from usage behaviours such as clicks, impressions and stay events. Combining the metadata provided a new direction for search, engagement and targeting our customers.”
From corporate applications to broadcast user content recommendations, AI paves the path in production in terms of language, subtitling and dubbing. “We have content in all the continents in several languages; it hasn’t scaled production, but we are looking to tap into those areas,” said Israel Esteban, CTO, beIN Media Group.
“We are using for churn prediction, how to engage better with customers, and an important area is anti-piracy and content protection. I would say in that space we have been using AI earlier, helping teams catch and react faster to content leakage and theft.”
With free-to-air broadcast and ad-supported digital services, IMI is working on personalisation and recommendations across the group, said Brad Eliot, CTO, IMI. “Getting first-party data is a bit of a challenge, so we have rolled out our customer data platform and kicked off an initiative with the National, our English-language digital platform, on registrations to receive first-party data. And then from there, building up profiles to create personalised content. This is in addition to the content recommendations and personalisation services that are already running on our Sky News Arabia platforms.”
Having employed AI for audience measurement and segmentation, targeted advertisements and recommendations, there is also the risk of the black-box algorithm effect. “The user experience has to be more friendly to explain the usage of AI,” said Romary Dupuis, Chief Data Officer, Viaccess-Orca. “We used large language models to make it easier for customers, for their audience segmentation by natural language than complicated details based on generic description.”
Large language models are breaking down the barriers that were cumbersome in the early days of robotics when scientists were trying to build humanised interfaces. The past couple of years has shown AI as more of an end-user tool, raising its possibility as a personal assistant connected to APIs that will build an ecosystem towards the super application. “AI transforms into a user experience, which then evolves into a super assistant, ultimately becoming your super app. AI as a user experience is going to be truly exciting,” said Dupuis.
As the industry inches its ways towards hyper-personalisation in the delivery of content, companies have the opportunity to push the limitations of AI beyond existing behaviour and demographics, and span those frontiers that edge the emotional and mental state of the viewer.
The idea behind a hyperpersonaliser is to train different deep learning models using user consumption behaviour and to build a pathway towards engagement. The goal is to drive customer behaviour even from a cold start and be able to pinpoint exactly what they want to watch. Gradually, a mix of predicted content is introduced, associated with different monetisation models.
“Variations are different when writing deep learning algorithms to support this kind of personalisation, taking into account sentiments, obsession, behaviour and other factors, and that’s the beauty of these algorithms,” said Arshad.
Personalisation was the first step companies took towards driving AI and where the immediate shift in patterns was noted. It is also used heavily in “anything that can help to stay relevant – so yes, start with the safest spot where it is easier to measure the results on the KPIs, and then anything that can help build that profile to the user but staying relevant and cautious,” said Esteban. “Because you’re going to start with something and end up with verticals that may be ambiguous.”
As an aggregator, it is about taking data and adapting to the interpretations. “We have a lot of information based on profile, age group, what percentage come, search, browse and leave, and we try to train our models accordingly. It’s an evolution, a learning, and about making changes to adapt and keep persevering with data sets to try and get value,” said Sharp.
Regional perspectives come into play while customising for certain geographies. “With the large language models, we try to reduce toxic content and prepare the AI to adapt to different cultural points of view,” said Dupuis.
The lack of Arabic content needed to train the large-language models leads to bias. “Because we generate a lot of Arabic content, we’re working with research institutes to help address this and enhance the training of these models in the region,” said Eliot.
Several governments around the world, including in the UK, Australia and here in Dubai, have created AI officers and ministries across different entities as a step towards promoting investment in the sector. The focus in the media and entertainment industry has been on optimisation and enhancements, and balancing the efficiencies that AI brings.
“We are looking at how Gen AI can play a role in generating new content,” said Eliot. “We’ve had some success in creating short-format documentaries that are 100% AIgenerated, from script, video content and voiceovers to transcription.”
People are putting the right strategies in place and adopting AI for experience, acquisition, retention, search or any vertical they wish to pick. “The big concern right now is governance, which is why governments are establishing agencies to bring some competencies in place. Until then, private LLMs are important to building an IP, or you end up giving your recipes to the whole world,” said Arshad.
Security is vital around the data being used. “We need a process, approvals, prioritisation of use cases, etc, because in the end everyone is looking to invest. What is the ROI and what are we trying to achieve? So a governance framework is key, and how to implement that across organisations and different departments,” said Esteban.
Even from a development standpoint, governance is crucial since the key component for local OTT players is localisation, which is generally done through in-house development teams and SaaS platforms. “How we add value and differentiate is through localisation, which comes at a cost,” said Sharp. “If you can increase development output by 40-50%, reduce bugs, have automated monitoring, error correction, all of which is fantastic, but you have to have a level of governance and standardisation before we end up putting our whole code bases onto ChatGPT.”
In the context of a super app or a one-stop shop for everything with aggregated services such as WeChat in Asia, the region hasn’t clearly demarcated a definition. Platforms add on value with games, maps, news and podcasts, anything towards retention and identifying them as super apps. “It has got to be relevant between watching TV and doing something else. It’s a balancing act – should we merge our properties into one app, one interface, the features of which are beneficial, but will we dissuade customers?” said Sharp.
On the streaming side, aggregation via a super app versus stacking of various streaming services remains the big unanswered question. “Aggregation is nice and convenient, and sport is a great example of that, but only time will tell whether having everything in one place is what people want, versus dedicated streaming services for each sporting code or discipline,” said Eliot.
As more cloud and AI is leveraged, security remains one of the biggest challenges and threats, said Ahmed, and blockchains or NFTs to counter security issues or track rights management is a possibility to consider.
“We have been working for over 20 years on cryptography and moved from content protection to cyber security. There are decentralised content protection systems but whether it’s solving how hackers hack into regular cryptographic systems is not the solution, compared to the investment,” said Dupuis. “We are focusing on an efficient model to detect piracy quickly so as to remediate faster. But exhaustive amounts of data are needed to differentiate between a bug in the application that is generating licences or someone who’s really hacking you.”
The use of AI is a new exposure in terms of cyber security. As a recommendation engine or to automate operations or make a new user interface, any opportunity comes with risk. “But it is also an exciting area, with a bigger need to address data poisoning, prompt hacking and other security issues,” said Dupuis.
In terms of rights management for certain regions, information from ISPs about who’s accessing what content and being able to then work with authorities to do takedowns would be a step towards battling piracy. DRM solutions involving third-party tools and AI using multiple datasets and models are also making headway towards tackling piracy.
“And this is where we go back to governance,” said Esteban. “What and how do you use AI, and how do you expose it?”
The big focus, especially in Saudi Arabia, is around PDPL (personal data protection law) and the rules of GDPR. With the risk of security breach, “we’re seeing an introduction of cloud into KSA, where we identity management solutions and lock down one solution, and then have a hash and tokenised value going out to the rest of our partners and platforms to try and reduce some of that risk,” said Sharp.
With the growth of OTT platforms, everyone wants better acquisition and retention through the latest features and the greatest solutions. The future lies in reshaping strategies between aggregated OTT offerings for each user segment, which will help achieve economies of scale. “It’s all about how to make our aggregated product offerings attractive as an OTT service provider and whether it will be better priced for the customer,” said Arshad.
For telcos, broadcasters and content owners, it’s about data management to optimise operations. “The challenge is to make one consistent data lake that addresses everyone’s needs efficiently,” said Dupuis.
On the content side, using technology to optimise how you acquire, produce and distribute content will be “a key theme moving forward, and to increase the use of virtual and augmented reality in the sets and in the studios”, said Eliot.
While it’s exciting times for the personalised user experience and AI will probably become the ergonomics around applications, governance is at the centre of it all. With that a whole new world of AI development will take shape, and global players will pave the way towards a licensed cost-effective ecosystem of advanced processes that will transform the entertainment experience.
