IBM Creative Assistant and unified search: How IBM uses watsonx tools to manage marketing assets and search institutional data

Many organizations find it hard to produce high-quality marketing assets at scale. Creating content that abides by brand guidelines, suits compliance requirements and connects with various audiences is a resource-heavy endeavor that often puts enormous pressure on creative teams. For instance, designers might spend hours adapting a single campaign across multiple media formats. Similarly, copywriters often struggle to tailor messages for diverse audience segments while ensuring alignment with brand voice.

As the digital landscape continues to evolve, the demand for marketing is increasing due to factors such as the growing importance of personalized customer experiences, the proliferation of digital channels and heightened competition across industries. Businesses are under constant pressure to engage with customers in meaningful ways, often requiring a high volume of customized content delivered at speed. This pressure limits the ability to focus on strategic, high-value work. This challenge is a clear opportunity for innovation through AI-driven asset generation.

Creating assets demands significant time and effort to ensure alignment with branding, customization and compliance standards. For this reason, AI has the potential to aid copywriters and designers with noncore components of their work, maximizing the impact of their expertise and labor hours.

IBM® Creative Assistant, built with IBM watsonx™, uses the power of generative AI to jumpstart marketing content creation. With Creative Assistant, you can access a comprehensive database of IBM product information directly within the tool. It is a versatile and intelligent AI-powered marketing asset generator. Creative Assistant uses AI technology and preapproved IBM templates to help produce polished, on-brand assets in a fraction of the time traditionally required.

IBM Creative Assistant has already generated over 4,000 assets, and the company is looking to expand its capabilities further.

While IBM Creative Assistant efficiently produces content, it lacks prompt-to-image and video generation capabilities. This next exciting step in asset generation is not straightforward to implement. While generative models for images and videos exist, ensuring that asset generation follows IBM’s design and brand compliance is a complex challenge. IBM assets must adhere to stringent design, legal and ethical specifications.

To address this, IBM is considering an intermediary step: training an AI model to check images for brand compliance. This approach is easier than generating brand-compliant images from scratch and can be applied to both stock and AI-generated images to ensure high-quality, acceptable outputs. Creative Assistant already incorporates similar capabilities for text, such as legal compliance, which has proven much more straightforward to implement.

Expanding these capabilities might significantly boost efficiency while reducing costs. With the time saved, IBMers can focus on writing briefs and guiding the creative vision while AI takes on the heavy lifting of asset creation. This shift optimizes working hours and ensures that designers can be more innovative and focus on high-value tasks, amplifying their impact.

Several specific capabilities have been identified as priorities by our team. Fast, high-quality asset generation for both internal and external use, adhering to IBM’s brand and design standards, would deliver significant time savings given the volume of assets that IBM must produce. Also, expanding into automated text, video and image generation presents wider opportunities for personalization to different markets and customer segments, yielding content that resonates more effectively with target audiences.

The goal is to enable AI to generate assets with minimal designer input, streamlining the creative process. However, implementing these initiatives remains complex, primarily due to the scale of the organization and the legal requirements for AI technology. By taking a phased approach and using intermediary solutions such as brand-compliance checking, IBM is well positioned to navigate these challenges and unlock the next wave of efficiency and innovation in asset generation.

Large organizations often struggle to unify disparate data across systems and databases. Inefficient data retrieval systems lead to team silos, scalability challenges and reduced productivity. Also, with fragmented systems, running tasks requires domain expertise to locate specific data. This requirement increases the learning curve for new hires and hinders employees from efficiently finding the information they need.

Any workable solution must synthesize data from these sources without requiring wholesale migration of the data itself. It must also maintain security and confidentiality—revealing appropriate data only to the appropriate authorized user.

To address these challenges, IBM has developed a unified search capability over the past 18 months, drawing data from diverse sources, including Slack and internal repositories. Textual data is stored in vector databases, with an advanced AI model layered on top for processing, allowing for free-text search of all sources. Internal Application Programming Interfaces (APIs) enable IBM teams to access both the raw vector data and the model. This functionality has already been integrated into internal search tools, replacing the older Lucene syntax with free-text, contextually aware querying. The result is a far more effective and intuitive search experience for IBM employees.

A key feature of this search system is that it has been provided with context about the employees who are querying it. When processing queries, the AI considers the user’s role, permissions and identity, returning only data that the individual is authorized to access. This mechanism ensures confidentiality and compliance with internal access controls while delivering highly relevant results. It is pivotal to ensure not returning sensitive data, so guardrails for access based on user identity are crucial to implement.

IBM is working to expand these capabilities to include images and videos. This expansion requires proper textual tags for existing visual assets, which is crucial for an AI’s understanding of these media types. While IBM’s Digital Asset Management system stores these digital assets, many are not correctly tagged. IBM plans to use AI to autotag all digital assets to address this gap. After completing the tagging process, the search model can evolve into a multimodal system capable of returning relevant images, videos, text and source documents through a single query.

This unified search system can dramatically improve efficiency by giving employees universal access to essential information regardless of format. It can also eliminate the need to store documents in multiple vector stores, streamlining IBM’s data infrastructure. Moreover, other internal products—such as Creative Assistant—can integrate unified search as part of their offering rather than building and maintaining separate implementations of similar technology.

AI-driven innovations will continue to transform workflows, enabling teams to focus on strategic, high-value tasks while AI handles the heavy lifting. Unified search and LLMs for findability will empower employees to find information with ease, while advanced image and content generation will enable teams to create high-quality, brand-compliant assets like never before. By integrating these capabilities, IBM not only enhances its internal operations but also sets a benchmark for delivering similar transformative value to its clients, redefining how organizations harness AI to drive impact.

IBM partnered with NYU’s Stern School of Business to deploy 4 MBA consultants for a semester-long project supported by Stern’s experiential learning program. Samir Ali-Darwish, Auston Ricca, Tomer Friedman and Narissa Moonsammy of Stern worked closely with IBM’s Marketing, Communication and Corporate Social Responsibility team to provide an external perspective on the organization. The MBA consultants researched IBM employee feedback data, conducted interviews with IBM leaders and used IBM’s own watsonx platform to synthesize the information obtained, culminating in a prioritized AI roadmap for the organization. To help close the global AI skills gap, IBM is committed to training 2 million learners in AI by the end of 2026. Through collaborations such as this one with NYU Stern and free coursework from IBM SkillsBuild, we are reaching learners around the world.