<img src=/content/images/image_43e2d5fddea38189a3b42c7f75b9f634.jpg><h2>Client Background</h2><p>KDDI Corporation is one of Japan’s leading telecommunications companies, headquartered in Chiyoda-ku, Tokyo. KDDI offers a wide range of mobile and fixed-line personal communication services through its au brand, as well as business solutions. As part of its efforts to implement generative AI in its operations, KDDI aims to create new value by partnering with startups that possess advanced capabilities in developing state-of-the-art AI solutions*.</p><p><i>*KDDI CEO Message, May 2024, https://www.kddi.com/english/corporate/kddi/president/.</i></p><h2>Challenge</h2><p>KDDI promotes its services through online advertising campaigns, which must be adapted for different platforms and user behaviors. At the same time, ad creatives must adhere to strict brand guidelines, ensuring consistency in logos, colors, and messaging while staying true to the brand’s values. This process requires skilled professionals who invest significant time and effort to get it right.</p><p>Although tools exist to streamline ad banner creation, they rarely leverage first-party advertising data (e.g., click-through rates). As a result, creative decisions often rely on the experience of designers and marketers rather than clear, data-driven insights, making it difficult to measure effectiveness.</p><h2>Recursive’s Solution</h2><p>To address these challenges, Recursive, in collaboration with the KDDI group company, developed an AI-powered, data-based ad creative generation system. It enables the semi-automatic creation of ad banners while ensuring alignment with KDDI’s brand guidelines and the four visual identity styles defined for the au brand. Once the banners are generated, the system automatically selects those with the highest AI-predicted advertising effectiveness based on past campaign performance.</p><p>The system’s pipeline integrates two AI models: the <i>Image Generation Model</i> and the <i>KPI Prediction Model</i>.</p><h3>Image Generation Model</h3><ul><li><b>High-Quality Image Generation:</b> The model leverages the latest image diffusion technology to generate realistic human images based on text descriptions while avoiding exaggerated or stereotypical elements associated with Japan that are commonly found in models trained on Western datasets (e.g., kimonos, anime-style).</li><li><b>Visual Identity Customization:</b> The model detects trigger words in prompts and applies style-specific adjustments to elements like lighting, contrast, and color. This ensures that images align with one of the au brand’s four visual identity styles—clean, friendly, playful, and advanced.</li><li><b>Automatic Prompt Translation:</b> Since the model’s API only supports English input, Recursive integrated automatic Japanese-to-English prompt translation using a large language model (LLM) to ensure seamless functionality.</li><li><b>Image-to-Image Functionality:</b> In addition to text-to-image generation, the model can modify elements of an input image—such as adding, removing, or changing accessories, clothing, hairstyles, etc.—based on the user’s prompt.</li><li><b>Faster Processing:</b> Recursive developed a request handler with asynchronous processing to efficiently manage multiple tasks. This significantly enhances response speed and ensures smooth performance for multiple users.</li><li><b>Background Removal:</b> Since the images are designed for banners, the model includes automatic background removal, which can be activated by adding a specific trigger word to the prompt. This makes post-editing easier and more efficient.</li></ul><img src=/content/images/image_be9fdbad3e1bef788ad3934446e7f756.png><h3>KPI Prediction Model</h3><ul><li><b>Accurate CTR Prediction:</b> Using a custom neural network architecture, the model predicts click-through rates (CTR) for different audience demographics (e.g., age, gender). This provides the client with data-driven insights into audience engagement and ad effectiveness.</li><li><b>Real-Time Performance Estimates:</b> The model generates predictions in less than one second, enabling the client to assess how a finalized but unreleased banner will likely perform in an ad campaign.</li><li><b>High Accuracy Results:</b> Recursive trained the model on a wide variety of historical data from different social media platforms provided by the client. This extensive training enabled the model to predict content performance levels with 72% accuracy.</li><li><b>Cost-Effective Deployment:</b> The model offers efficient processing time, even when running on standard CPU hardware, facilitating easier deployment and reducing operational costs.</li></ul><img src=/content/images/image_d5bef342f008ca86f55d8b3db9af1ca8.png><div class="callout"><span></span><h2>Results</h2><p>The deployment of the ad creative generation system is set to deliver a significant impact:</p><p><b>1. Workload Reduction:</b> Beta testing confirmed that the system reduces the time spent by 50% on the following key tasks:</p><ul><li>Design planning (layout design, font selection, image selection, etc.)</li><li>Draft creation (background creation, color selection, font and image placement/size adjustment, etc.)</li><li>Finalization (adding effects such as shadows, image retouching, etc)</li></ul><p><b>2. Improved CTR:</b> Ad creatives are selected based on AI-predicted performance, ensuring maximum impact across platforms.</p><p><b>3. Expanded Creative Possibilities:</b> AI-generated, high-quality human images offer greater variety in ad creatives without increasing costs. Marketers can experiment with different styles and concepts, making campaigns more diverse and engaging within the same budget.</p></div>

Recursive, in collaboration with the KDDI group company, developed an AI-powered, data-based ad creative generation system.

“When translating corporate issues into concrete solutions, Recursive can offer a wide range of perspectives as they have not only engineers with expertise in information engineering but also experts with extensive experience in the marketing and business fields.”

Transforming Ad Creatives Generation for a Leading Telecommunications Company

<h2>Client Background</h2><p>NeXT FOREST Corporation., a joint venture between IHI Corporation and Sumitomo Forestry, aims to protect tropical forests in West Kalimantan, Indonesia, through the sustainable management of tropical peatlands—unique ecosystems composed of accumulated organic matter from dead plants.</p><p>Tropical peatlands are highly effective carbon sinks, storing twice as much carbon in the soil as all the world’s forests combined. They support the growth of massive trees that form rainforests, sustain diverse ecosystems, and provide habitats for many endangered species, such as the orangutan.</p><h2>Challenge</h2><p>Tropical peatlands are highly sensitive to human activity and climate change, relying on specific environmental conditions to remain stable—particularly proper groundwater levels. When these levels drop too low, peatlands can dry out, becoming highly flammable and susceptible to forest fires, which release vast amounts of stored carbon back into the atmosphere. Additionally, as peat decomposes in drier conditions, it further increases CO2 emissions, contributing to climate change.</p><p>One striking example occurred in 2015, when a forest fire in Indonesia burned 2.6 million hectares of forests, releasing approximately 1.3 billion tonnes of CO2 into the atmosphere—more than the annual emissions of many countries. This was a global-scale environmental disaster.</p><p>Traditionally, managing water levels in peatlands has involved experienced technicians manually surveying the land and creating detailed topographic and contour maps at 50-centimeter intervals. This labor-intensive process can take up to five years to complete.</p><h2>Recursive's Solution</h2><p>To address these challenges, Recursive utilized over 10 years of groundwater data collected by Sumitomo Forestry in the forests of West Kalimantan. By combining this data with expert knowledge on topography, vegetation, waterways, and meteorological conditions, Recursive developed a fully custom AI model from scratch. This model integrates machine learning with physics equations to create a high-resolution mesh map of the area and accurately forecast groundwater levels seven days in advance, with a precision of 6 cm.</p><p>The model is designed to predict when groundwater levels approach thresholds associated with potential hazards, enabling stakeholders to take prompt preventive actions. This proactive approach helps reduce peatland degradation and lowers the risk of devastating forest fires.</p><img src=/content/images/image_fd630d65b8a590ba4429b218c3c67fa7.webp><p>Following the initial development phase, Recursive’s engineering team visited West Kalimantan, Indonesia, to further enhance the AI model. On-site observations and direct feedback from engineers and workers operating the irrigation system played a crucial role in refining the solution.</p><p>This trip will enable the Recursive team to incorporate additional parameters that were previously unavailable or incomplete, such as variations in soil quality based on altitude, data on dam types, and the dynamics of water flow. These improvements in future stages will ensure our solution continues to evolve to meet customers’ needs. Learn more about this trip in our<a href=https://recursiveai.co.jp/en/blog/borealis-project-background-report/> blog post</a>.</p><img src=/content/images/image_a27c358f661b50a3df7fdce56ed4f6b3.png><p><i>Recursive team is preparing to get groundwater level measurements in West Kalimantan, Indonesia.</i></p><div class="callout"><span></span><h2>Results</h2><p>Recursive's AI-powered groundwater management solution delivered impactful and measurable results:</p><ul><li>1,200 km² of tropical peatlands are efficiently monitored and managed, enhancing oversight and reducing degradation risks.</li><li>Prevented the release of approximately 300 million tons of CO₂, contributing significantly to climate change mitigation.</li><li>Operations are now significantly faster, reducing reliance on lengthy manual surveys and decreasing the workload for employees.</li><li>AI simulations have optimized dam placement and irrigation planning, ensuring better resource management and sustainable water usage.</li></ul></div>

NeXT FOREST Inc., a joint venture of IHI Corporation and Sumitomo Forestry, uses Recursive’s AI-physics model to predict groundwater levels in Indonesia’s tropical peatlands. 

“Recursive has a strong track record and excellent engineers. Their knowledge of biophysics, in addition to their engineering perspectives, was a tremendous strength and a great source of reassurance.”

Case studies

Transforming Ad Creatives Generation for a Leading Telecommunications Company

AI Plays a Part in Sustainable Forestry

Leading enterprises trust Recursive

Document download