With the introduction of ChatGPT, holding a promise to revolutionize digital communication by enabling users to interact with AI that understands data, intent and context, the integration of governed data to such generative AI models becomes a pivotal point in reaching the potential. Bringing governed data to life with generative AI makes it accessible, relevant, meaningful – and responsible.
Many organizations are taking to generative AI to find a way to quickly generate content and compelling marketing copy to promote their products in their online stores and social media. This ties closely into better and more efficient product experience management (PXM) processes. Using generative AI seems, at first glance, easy, appealing and very productive. However, when setting up so-called prompts, good data is essential to direct the algorithms to provide appropriate responses.
Responsible AI usually refers to the fact that it prioritizes ethical and societal considerations, while minimizing potential harms such as bias and privacy violations. It involves ensuring transparency and accountability in AI systems. However, the responsible output relies on the input being governed according to certain rules. Translating this process to generative product information, means that data governance is required for the input in order to ensure a trusted and valid product description for the output.
Generative AI, which relies on machine learning, is a powerful tool that can produce content in a matter of seconds, including product descriptions that are crucial for driving superior customer experiences. With such powers in their hands, companies must ensure that their use of generative AI is responsible and trustworthy. The key to responsible AI lies in prompt engineering.
Prompt engineering is the process of creating inputs or prompts that guide generative AI, therefore, the quality of these prompts can greatly affect the output. Companies must ensure that their prompts are accurate and relevant to their target audience in order to produce engaging and trustworthy content.
There are many pitfalls of using AI-generated product information indiscriminately:
Companies cannot rely on generative AI and chatbots as their single source of truth. While generative AI can produce compelling content quickly, it does not have the ability to govern data. In addition to the chatbot, at least two further components are needed in the content production: (1) A single source of truth based on data governance, (2) clerical review to ensure brand compliance and approve content produced by generative AI.
To ensure the accuracy and trustworthiness of AI-generated content, as well as to make the clerical review easier, you need to drive the AI query with governed data. This means that the data used to prompt the generative AI must be accurate, relevant and subject to strong data governance capabilities.
Master data management (MDM) supports governance of product data and is designed to provide a single source of truth. By interfacing with a generative AI, such as ChatGPT, you can ensure that the data used to prompt the AI is accurate and trustworthy.
Via an API and a simple configuration of the MDM, master data management and generative AI can work in conjunction to combine trusted data with the speed of content production. This allows you to produce content quickly while ensuring its accuracy and trustworthiness.
Automated content creation embedded in an MDM interface with a Generate Product Description button. The AI is prompted with a product name, specifications and a product image:
Let’s unfold the use case:
The product manager, Julie, needs a market-specific product description. She will use the AI assistant for that. She defines the target criteria in the MDM to prompt the AI for suggestions. The MDM contains pre-defined prompts to query the AI in the most efficient way. These prompts contain directives such as: context; target audience definition (demographics, psychographics, revenue); purpose; writing style and guidelines (tone, structure, length, word choice).
The prompt engineering is embedded in the MDM. This means the structured and quality-oriented data and processes of the MDM make it more efficient to query a generative AI.
The AI uses the clean data from the MDM and the prompt, then returns a few suggestions. In a workflow, the product manager reviews the suggestions, refines the prompt, and asks for specific quotes from local celebrities to be added. She receives new suggestions in response. She selects one and forwards it to a copywriter.
What happens is that a conversation takes place between the AI and the product manager.
The copywriter reviews the suggestion, refines the wording to make it more brand compliant. He notifies the data governance team of his feedback, so that they can refine the pre-defined prompts. He finds that the quote doesn't belong to the celebrity named by AI. He finds another quote, modifies the text and sends it back for final validation to the product manager.
Julie approves the final product description.
To be concluded: The AI accelerates the team’s work, but an expert’s review is indispensable to refine and validate.
AI needs to be asked the right question with the right data. By improving data quality and consistency, MDM empowers you to craft better prompts, leading to accurate, relevant, and reliable chatbot responses.
In a global world, trust and security are essential. MDM ensures legal compliance and data security, allowing you to confidently operate generative AI within data privacy regulations.
AI excels when integrated into everyday processes. MDM enables you to seamlessly integrate chat AI into workflows and foster cross-functional data collaboration for enhanced efficiency.
Generative AI is a powerful tool that can provide a tremendous tailwind to your product description creation, removing a great deal of tedious and repetitive work. However, in order to remain ethical and correct, the AI must be guided by a single source of truth containing governed product master data.