George Davis, Ph.D., Founder and Chief Executive Officer at Frame AI.
getty
AI features are creeping into the products we use every day: writing assistants for office suites, copilots for coding and bots to guide you through how-to articles. Behind most of these sits a framework with the unassuming name “RAG”—retrieval augmented generation. RAG works like a Magic 8 Ball for users: You pose a question, an application shakes that question up with knowledge and context from the work at hand, and a large language model (LLM) formulates a valuable response.
Recent offerings from OpenAI (GPT and the Assistant offering), Amazon (Bedrock’s Q models) and Google compete to streamline this pattern for developers.
The launch of ChatGPT set the stage for RAG-driven enterprise AI strategies in late 2022, when, within two months of its release, ChatGPT had attracted an estimated 100 million monthly users. Its user-friendly interface and robust conversational capabilities drove this rapid adoption, and the popularity of ChatGPT among the general public bled into enterprise AI strategy.
However, RAG-based models only represent some of what is possible with AI. Despite RAG-based bots’ practical applications and user-friendly nature, their success has contributed to a somewhat myopic view of AI as a sophisticated bot.
Introducing STAG
AI-forward businesses are discovering that not every use case can be met with the same reactive query-response platform. For applications that require proactive analysis of streaming data, there is a new architecture to consider: stream-trigger augmented generation systems, or STAG. STAG systems are proactive. Where RAG exposes curated knowledge to dynamic queries, STAG tracks fixed queries against massive streaming data. Where RAG amplifies user knowledge, STAG amplifies user vigilance. While RAG and STAG operate distinctly, their power is most evident when combined.
RAG And STAG: Two Models For Querying Data
RAG and STAG are complementary architectures, making it easier for people to answer important questions. However, the biggest gap in the query-based RAG systems is that users must know which questions to ask. However, it won’t proactively offer insights on emerging trends or unanticipated user issues the user hasn’t thought to ask about.
Because the user must ask to receive, the RAG-based interaction pattern is definitionally reactive, which makes it difficult to meet the needs of environments that demand continuous data analysis.
In STAG-based systems, users don’t have to query to receive insights. Instead, STAG systems automatically and continuously query on behalf of the user.
In the example of a product leader working on feature enhancements, a STAG-based system might automatically surface a sudden spike in user engagement with a particular feature, indicating a shift in user preference or a newfound application of that feature. Alternatively, it could identify an emerging pattern of issues or complaints specific to a recent update, allowing the product leader to address these problems before they escalate quickly.
WIT
Businesses investing in generative AI face a decision about how to allocate resources between different AI technologies and business applications. Given the popularity of RAG-based systems like ChatGPT, the most obvious answer today is to deploy reactive bots. However, the people who can most benefit from AI often have the least time to engage with AI tools.
STAG Unlocks The Value Of Unstructured Data
A key challenge in the modern data landscape is the overwhelming presence of unstructured data—including documents, call notes and social media content. Surprisingly, over 99% of this data remains unanalyzed yet replete with vital insights.
Through its integration with advanced large language models (LLMs), STAG addresses this gap. It goes beyond basic analyses to understand nuances in unstructured data, identifying key patterns, trends and actionable insights that would otherwise remain hidden or broken into silos.
The true power of STAG lies in its ability to not only process but also contextualize unstructured data. Traditional alert systems based on structured data often provide limited insights. STAG, by contrast, utilizes its LLM to add context to data trends in relevant terms, much like a RAG bot would in response to a specific query.
This approach transforms simple alerts into comprehensive, actionable insights tailored to the user’s specific role. For instance, STAG would detect a spike in customer complaints, communicate the reasons for that spike to a CX manager, and do so in a way that offered constructive solutions for addressing those complaints.
STAG systems monitor for signs of adverse customer reactions or trends that can endanger the account and proactively alert the company so they can intervene. The company’s executive team now receives daily checklists from AI for better oversight and theme identification, significantly improving customer satisfaction and retention.
Additionally, STAG plays a critical role in bridging silos within organizations. It allows for proactively exploring unstructured data from multiple disconnected sources. This capability is particularly beneficial in large, complex organizations where crucial information is often scattered and underutilized.
To effectively utilize a STAG system, it’s essential first to identify the specific insights or events it should focus on and then align these with existing workflows. The process involves crafting well-defined queries that guide the STAG system to discover relevant and actionable insights within vast amounts of unstructured data.
RAG And STAG Are Complementary
The businesses that benefit most from generative AI won’t be those that choose between RAG and STAG architectures but those that deploy both frameworks in complementary ways. STAG’s autonomous data analysis capabilities and RAG’s interactivity can form virtuous feedback loops. For example:
• Data from an RAG-based customer service bot can be continuously monitored by an STAG system, which proactively suggests topics causing unnecessary effort and proposes content that would improve bot performance.
• A STAG-based system may be effective in surfacing new opportunities for a marketing campaign, while a RAG-based assistant helps marketing develop messaging and creative content for the campaign.
The meteoric rise of ChatGPT and the subsequent emphasis on RAG-centric AI systems have significantly influenced the understanding and application of AI in both the public and corporate sectors. However, the advent of STAG and the growing recognition of the need for a diversified AI strategy underscore the importance of a more balanced and comprehensive approach to AI integration in business.
Effective AI deployment should mimic successful human organizations, which balance standing responsibilities with strategic management. The combination of RAG and STAG supports the structure behind the most transformative human teams.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : Forbes – https://www.forbes.com/sites/forbestechcouncil/2024/02/26/beyond-bots-what-else-can-ai-do-for-business
