Discover Adobe Experience Platform AI Assistant
How AI Assistant works
AI Assistant combines generative and non-generative components to unlock the power of data — from both Adobe and customer sources — in Adobe Experience Platform, with user permissions and access controls honoured. Here are three key components:
AI Assistant user interface
Users interact with AI Assistant through a user prompt. AI Assistant can interpret natural language and interact with a dialogue management service, which guides other models to provide engaging and natural responses. It can rephrase a given question based on conversational context and determine the type of questions being asked to provide a useful answer.
Generative experience models
AI Assistant uses a collection of experience models designed to address the context of each specific AI Assistant use case and allow for quick data navigation as needed.
There are three key dimensions of generative experience models that make it so powerful:
- Base models are foundational models to AI Assistant and are applied for all customers. They consist of large language models (LLMs), linguistic models and task-specific language models that understand natural language within prompts from users. These base models are grounded in Adobe data so that AI Assistant users can ask open-ended questions and get guidance and insights to help them progress through tasks and answer the questions they may have while using Adobe applications.
- Customised models augment base models and are grounded in customer data to give customer-specific context to AI Assistant. These allow customers to answer questions about their data, discover data insights and understand trends. Customised models can also power predictive use cases, such as forecasting and making recommendations relevant to a customer’s business. Because these customised models use customer data, they won’t be shared outside of that specific enterprise and role-based access control limits what each user can access.
- Decisioning services are layered on top of models and data to help inform what AI Assistant should serve up to the user based on current and historical context. This could be in the form of next steps, recommendations or a response to previous questions that they asked so the conversation is multi-turn in nature. For example, if the customer asks a question around a merge policy, AI Assistant could make a recommendation for the user to ask how many segments are associated with that merge policy.
AI Assistant is powered by data
Data is the vital ingredient that drives personalised insights and engagement. That’s why we launched Adobe Experience Platform to unify siloed data across digital and enterprise sources to build unified customer profiles.
Adobe Experience Platform provides the foundation for AI Assistant to gather a complete view of the business. AI Assistant is grounded in Adobe data, including our product documentation, community forums and industry or use case playbooks. The tool is also grounded in the customer’s enterprise data and metadata stored within Adobe Experience Platform.
Generative experience models can query these data shops and generate outputs used to answer a question. The data can remain organised, pre-joined and indexed into customer-specific knowledge bases so language models can interact with it in an open-ended way.
Putting trust at the centre
AI Assistant has been developed with enterprise-grade trust, governance and customer data stewardship in mind. The generative experience models architecture allows Adobe to ensure that the following principles are respected:
- Customer data is always safe. The models developed on top of it are custom for each customer so there is never a data leakage between customers.
- No LLM is trained or fine-tuned using any of the customer interactions or customer data. Furthermore, logging for LLMs has been disabled as an extra precaution.
- A series of filters are applied to the prompt and answer pipeline to ensure that the conversation is safe. We leverage 3rd party LLM’s content filtering service to moderate sensitive or dangerous content. We have also developed other filters to scrub personally identifiable information and filter out sensitive inputs. Responses are only provided to the user if they pass both checks.
- No third-party sources are used to provide responses back to the customer.
- Every answer provided by AI Assistant has appropriate layers of verifiability.
- All Adobe generative AI features go through Adobe’s AI governance process and are align with Adobe’s AI ethics.
As part of generative experience models, Adobe has developed a series of models that help with intent classification, natural language to query expressions, citations and more. These are internal models that operate within the Adobe ecosystem and allow for controls to continuously improve the correctness of the answers. It also allows Adobe to be very transparent about the internal architecture and keep customers informed.
What AI Assistant can do for you
AI Assistant is designed to make your work easier and more productive, whether you’re a new user just getting familiar with Experience Platform-based applications or you’re an expert trying to streamline your workflows and drive better outcomes. Here are some of the ways AI Assistant is helping drive efficiency for teams in the enterprise.
Campaign and channel marketing teams
AI Assistant helps marketers quickly understand available tools and features within the application they are using to enhance the quality of their work. It can also help them jumpstart the creative process and accelerate campaign creation, content generation and publishing.
Take, for example, a campaign manager who just recently started using Real-Time CDP and may have got stuck trying to activate an audience. They can ask AI Assistant a question like, “What are the next steps to activate an audience?” and AI assistant will provide a tailored answer based on Adobe documentation.
Marketing operations teams
Marketing operations teams spend a lot of time measuring the performance of their campaigns as well as understanding how to configure journeys, segments and audiences within their marketing workflows. AI Assistant helps them streamline their work by providing a fast and centralised place to get answers to questions about their marketing workflows. Operations teams will be able to troubleshoot faster based on the insights they uncover and get product guidance to unblock them when they encounter unfamiliar concepts.
An example is a marketing operations analyst who needs to understand if there are audiences that have been created but are not being utilised or if there are audiences with duplicate definitions. Instead of having to wait for a data specialist to help, they can just ask AI Assistant to provide the list of such audiences. AI Assistant can also visualise the answers to insights questions.
Data and IT teams
AI Assistant helps streamline routine tasks and accelerate data exploration and insight discovery that can inform and improve the accuracy of processes like pipeline management, data hygiene management and data lineage impact analysis.
Take, for example, a data engineer who is trying to understand the lineage of their data and needs a graph that shows where a specific field is being used. They can ask AI Assistant to provide a list of segments, schemas and destinations where a particular data field is in use. They’ll even get the source query used by AI Assistant to generate the answer.
What’s coming and how AI Assistant will evolve
Adobe is focused on bringing the potency of AI Assistant to a range of use cases beyond the ones described above. The use cases that are generally available or in the works are focused on product knowledge and guidance, troubleshooting, content generation and insights and help use cases. In the near future, AI Assistant will be able to perform workflow automation and goal-driven tasks for users, surfacing predictive analytics and recommendations that will help to optimise business outcomes.
Learn more about what AI Assistant can do for your business.
Akintunde Ajayi is a senior product marketing manager for Adobe Experience Platform. He focuses on bringing innovative Experience Platform capabilities to market and driving awareness and adoption around Adobe’s Personalised Insights and Engagement offerings. He has over 14 years of experience in consulting, systems integration and product marketing. Akin joined Adobe in 2019 after obtaining his MBA from Kellogg School of Management at Northwestern University. Akin is a devoted husband, father of one and an ardent fan of Liverpool FC football and Utah Jazz basketball.
Horia Galatanu is a director of product management for Adobe Experience Platform. His focus is on building functionality in the areas of artificial intelligence and machine learning, data and experimentation. Horia joined Adobe in 2007 and has been involved in multiple initiatives over the years, including Adobe Primetime, Adobe Campaign and Adobe Journey Optimizer. Horia lives in the San Francisco Bay Area with his family and loves hiking, photography and watching Arsenal FC play football.
Recommended for you
https://business.adobe.com/fragments/resources/cards/thank-you-collections/generic
