Creating bespoke customer engagement with ThinkAnalytics using generative AI on AWS

This blog was co-authored by Peter Docherty, Founder and Chief Technology Officer, ThinkAnalytics.

ThinkAnalytics is an AWS ISV partner specializing in personalized content discovery, metadata, viewer insight and targeted advertising for media and entertainment. ThinkAnalytics provides AI-powered personalized content recommendations for media companies. Its products are used by major broadcasters and streaming services globally. To provide even more personalized recommendations, tailored to individual users, ThinkAnalytics sought to add generative AI capabilities to deepen the personalization of its Think360 content discovery and viewer engagement platform. At the IBC show in Amsterdam 2023, ThinkAnalytics unveiled new generative AI features for personalized recommendations and other inbound and outbound interactions, powered by Amazon Bedrock.

The Think360 platform is an AI/ML-based personalized search, recommendation, and user engagement solution that puts control of content discovery in the hands of the business, revenue, and product owners. ThinkAnalytics has proven that personalization of the viewer’s experience delivers increased viewing, purchasing, higher satisfaction, and perceived value for money, reducing churn. For more details, see AWS Case Studies.

By integrating the content discovery platform with generative AI, ThinkAnalytics has further enhanced its personalization and recommendation capabilities. This makes the recommendations feel more natural and personal rather than generic. Users are then more compelled to view the suggested content. Think360 leverages large-language models (LLM) to create unique personalized experiences and deliver engaging end-user experiences.

New voice and chatbot capabilities

The Think360 engagement module uses generative AI to personalize recommendations and interactions. It takes generic recommendations and enhances it in two ways. First, it adds additional context to the answer. Second, it adjusts the tone based on who the end user is. For example, to appeal to younger viewers, the service converses in an animated, enthused tone, as shown in the screenshots below.

The image on the left shows a basic recommendation. The image on the right shows a generative AI–enriched response, personalized to the viewer.

Fig 1. Basic recommendation with limited interactions

Fig 2. Enriched recommendation based on target persona

Think360 enriches engagement beyond in-platform content recommendations. It also actively applies generative AI to create personalized outreach via email and social media, allowing customers to contact users at risk of leaving the service effectively. With generative AI–enriched engagement, Think360 sends those users communications tailored to their interests and preferences. This increases interaction, engagement, and ultimately helps drive retention and reduce customer churn.

ThinkAnalytics leverages Amazon Bedrock to create these generative AI–enriched interactions. Amazon Bedrock offers a fully managed, serverless platform to build generative AI applications using foundation models from leading AI providers via a single API. Amazon Bedrock was selected for its serverless experience, requiring no capacity management or provisioning of servers to perform inference. With Amazon Bedrock, customers are charged only for what they consume, which makes it well-suited to applications with varying levels of activity.

Interaction with Amazon Bedrock is via the InvokeModel API action. This action allows inference to be run against any supported model. It also allows for inference parameters to be set that influence the response generated by the model. By adjusting the Temperature and Top P parameters, ThinkAnalytics is able to create more varied responses.

ThinkAnalytics uses the Amazon Titan and Anthropic Claude models available within Amazon Bedrock to leverage industry-leading text-generation capabilities. Data privacy is an important consideration for ThinkAnalytics and its customers. Amazon Bedrock is a good fit, as training and inference data stays within the AWS network meaning ThinkAnalytics never has to send any data to third party model providers, providing confidential generative AI capabilities.

To provide a conversational interface, ThinkAnalytics used Amazon Lex, a serverless AI-powered chatbot that supports both text and voice input. It chose Lex because of its ease of integration with the rest of ThinkAnalytics’ serverless solution.

The end user interacts via text or voice with the Lex chatbot, which sends the request to an AWS Lambda function. This function then calls the Think360 personalization engine to generate detailed recommendations based on user segmentation and demographics. The data is sent to Amazon Bedrock, where a large language model enriches the response and creates a tailored, personal response.  This response is delivered to the end user. This is shown in this diagram:

Fig 3. Think360 flow diagram

Personalized cross-channel marketing

ThinkAnalytics personalizes the entire customer experience, from initial content discovery through customer retention and churn prevention. This blended approach allows consumers to interact with tailored, relevant content at every touchpoint.

To complement the enhanced in-product personalization, ThinkAnalytics has undertaken a new initiative to deploy personalized campaigns across different marketing channels. By analyzing user viewing behaviors captured by the Think360 platform, ThinkAnalytics identifies different user segments and demographics. These insights empower marketers to create targeted campaigns personalized to specific user groups across channels, such as email and social media. For example, documentary lovers receive an email about a documentary on dinosaurs, while drama fans see social media advertisements for a gritty new series.

This personalized cross-channel approach allows ThinkAnalytics and its customers to measure the incremental impact of personalization. ThinkAnalytics compares conversion rates for generic messaging against personalized outreach. Early results show over twice the engagement for personalized campaigns. This cross-channel personalization is powered by the same Think360 platform and generative AI capabilities. User data and segments identified within Think360 will feed into campaign management and marketing automation tools to create bespoke campaigns and customized messaging. Similar to the in-product personalization, Amazon Bedrock’s generative AI enriches communications with tailored tone and terminology.

Importance of first-party data

Third-party cookies have been commonly used in online advertising to track user behavior across websites. Advertisers place these cookies on users’ browsers when they visit sites to gather data and target advertisements based on browsing history. However, concerns around privacy and data collection have grown in recent years and regulations such as the General Data Protection Regulation (GDPR) now require explicit user consent for most cookie tracking. Major browsers are also increasingly blocking third-party cookies by default.

As third-party cookies are phased out, high-quality first-party data is increasingly critical for effective ad targeting and personalization. Amazon Bedrock’s strong data security and privacy controls allow customers to leverage their first-party data confidently. Training and inference data never leaves the AWS network, providing security, while multi-tenant isolation prevents any mingling of data across Bedrock users. These protections allow the use of sensitive first-party data about viewer behaviors and demographics to better personalize content and recommendations. Broadcasters and content creators achieve over a 100% increase in ad click-through rates with Think360’s first-party, data-powered generative AI capabilities.

Rather than relying on broad third-party segments, first-party insights reach more precisely defined and better targeted audiences. Think360 generates marketing communications dynamically to resonate with specific viewer demographics. It fine tunes content recommendations based on an individual’s viewing history and preferences. As third-party data usage faces increasing regulation, high-quality first-party data is key to engaging with end users effectively. With Amazon Bedrock, ThinkAnalytics customers can leverage their first-party data safely to deliver personalized, relevant experiences securely and at scale.

Conclusion

Enhancing personalization through AI is a key priority for ThinkAnalytics. By integrating Amazon Bedrock’s generative-AI capabilities into the Think360 platform, ThinkAnalytics takes personalization to the next level. Early results show strong promise for improving personalized experiences, engagement, ad-targeting, and customer retention. The enriched, tailored recommendations powered by generative AI help deliver content that resonates at an individual level. By extending personalization across channels, ThinkAnalytics reaches consumers with the right message at the right time. Personalized cross-channel campaigns have shown substantial lifts compared to generic, non-targeted messaging.

Together, ThinkAnalytics and AWS push the boundaries of what’s possible with AI-driven personalization. By using Amazon Bedrock, Think360 will enhance personalized experiences, improve marketing performance, and boost customer lifetime value. The future is bright for personalized, AI-enriched engagement.

The ThinkAnalytics Think360 platform is available in the AWS Marketplace. Providers can deploy it in their own AWS environment or use it as a fully-managed ThinkAnalytics SaaS service. To learn more, contact ThinkAnalytics at sales@thinkanalytics.com or connect with your AWS account representative.

About ThinkAnalytics

ThinkAnalytics is an AWS Partner for media and entertainment specializing in content discovery, viewer analytics, and contextual advertising solutions that empowers video service providers, studios, broadcasters, and media companies to deliver personalized experiences to their customers. The company has delivered content discovery and viewer insights to over 85 service providers serving approximately 475 million subscribers in 43 languages, with 8 billion recommendations per day. ThinkAnalytics is an EMMY Award Winner for Technology and Engineering. Privately held, ThinkAnalytics is jointly headquarted in Los Angeles and London.