Can generative AI unlock immense value? While organizations recognize the potential benefits of embracing technology, they remain cautious in their decisions regarding its implementation, seeking to optimize its adoption for maximum impact. With a dizzying array of styles, suppliers, and strategies to choose from, the sheer scope of options can be daunting. Investors naturally demand tangible returns on their investments, making it crucial for funds holders to identify practical ROI strategies that can justify the significant costs and organizational changes associated with embracing GenAI.
With a rich legacy of leveraging enterprise AI, Databricks has consistently harnessed its internal capabilities to drive innovative applications in areas such as fraud detection and financial forecasting. Our system ingests knowledge from multiple sources, integrating seamlessly with Salesforce and Metronome, and funnels it into our centralized logfood structure, where it is processed, refined, and made accessible to diverse personas, including data scientists and software engineers, for informed decision-making and actionable insights. This infrastructure comprises more than 10 petabytes of data and spans across 60 multi-cloud and multi-geographical locations, enabling us to manage over 100,000 daily tasks for our more than 2,000 weekly customers. As we work alongside clients on their AI development and journey, understanding how we leverage AI within our own organization becomes crucial; this enables us to uncover the tools, techniques, and best practices we employ.
A key approach to developing our AI methodology involves first implementing a robust AI governance framework, which involves close cooperation with designated teams of engineers, safety experts, and other relevant stakeholders. As we establish ourselves, we adopt a hybrid approach that combines tried-and-tested third-party solutions with internally developed AI-powered applications, rigorously testing their efficacy against traditional methods through A/B experimentation.
This framework and resolution methodology offers valuable insights for AI practitioners, as it showcases tangible achievements that can serve as a foundation for expanding into new use cases.
Databricks showcases its multi-step Generative AI vision in practice through compelling examples of clear wins and innovative approaches, highlighting its implementation in real-world applications.
Clear Wins
The adoption of GenAI by both internal and external community groups has been a resounding success for Databricks, with numerous organizations also capitalizing on its expertise to drive innovative outcomes. Initially, enhancing a company’s support operations is a crucial first step in implementing an AI strategy, whereupon we focused on empowering our support teams with more comprehensive documentation, enhanced information flow, increased autonomy to drive velocity or mitigate support issues, automated performance metrics, and further self-service capabilities for our customers. More than 40 engineering channels currently utilize our in-house Slackbot to streamline operations, alongside a community of over 3,000 active users. Across our organization, we’ve successfully automated responses to around 40,000 internal queries related to challenging topics such as difficulty assessments, script and SQL assistance, error code explanations, and architectural guidance.
The same Slackbot, boasting a vast and engaged user base, has successfully responded to over 1,200 queries in its external capacity. With regards to IT support, we have integrated GenAI with cutting-edge technologies to enhance our assistance and learning capabilities. Collectively, artificial intelligence (AI) chatbots and human-assisted help systems have successfully addressed an increased volume of inquiries, achieving a remarkable 30% deflection rate – a significant improvement from zero just two years prior. By the end of 2024, our ultimate goal is to achieve a success rate of at least 60%. Meanwhile, our BrickNuggets chatbot, now integrated with Area Sidekick, has introduced bite-sized learning modules tailored specifically for our gross sales team. Our general third-party chatbot is utilized globally by our teams to facilitate collaboration and provide tailored solutions to pressing questions, serving more than 4,700 active monthly users within our community.
The second compelling use case for GenAI lies in its application within software development programs. Through the strategic deployment of copilots, we have successfully enhanced the productivity of our engineers, thereby fostering the development and acquisition of valuable engineering intellectual property. The introduction of Copilot technology yields profound efficiency and productivity benefits, with an early adopter survey revealing that 70% reported increased productiveness, 73% achieved task completion more quickly, and 67% attributed time savings to focus on higher-priority tasks.
At Databricks, we harness the power of GenAI copilots to rapidly build instruments, dashboards, and machine learning models, including complex designs that would previously have required significant expertise or extensive development time. As a valued customer of Databricks IQ, we serve as trusted assistants to accelerate the development of knowledge engineers, streamlining tasks such as knowledge ingestion, reporting, and various data-related responsibilities. Extensive use of co-pilots enables the lengthening of language migration, ensuring case growth and promoting code clarity? Noteworthy improvements in productivity have resulted in significant gains, with increases of up to 30% in certain instances.
A spirit of experimentation
Databricks has demonstrated a proactive approach to experimenting with our AI technology, while simultaneously implementing necessary safeguards to ensure the integrity of our innovation efforts. From numerous Databricks hackathons, innovative concepts have evolved into prototypes or manufacturing-ready products, echoing the spirit of revolutionary thinking and acknowledging that we’re not simply integrating AI features but also building AI-driven infrastructure.
“One instance relates to the email era for our internal sales team.” Automating email efforts offers a convenient and eco-friendly approach to manage sales team workloads, yet executing this strategy can be challenging due to the need for contextual understanding of a specific industry, product, and customer base. By leveraging the intellectual capital within our organization’s knowledge repository, housed in our lakehouse, we’ve optimized its usage through the facilities of. We can seamlessly integrate open-source AI frameworks with our intelligence platform, which combines a knowledge warehouse, Databricks’ Unity Catalog governance, model-serving capabilities for execution, RAG Studio for augmented era, Mosaic AI, and unified knowledge units to refine structured and unstructured data, ultimately delivering high-quality response rates. The Role Alignment Gateway (RAG) plays a crucial role in our strategy, as it enables seamless integration of large language models (LLMs) with business knowledge while striking a perfect balance between quality and speed, thereby expediting the learning process effectively.
The outcome is a sophisticated email era feature that seamlessly integrates contextual information, including the purpose of the contact, the industry they represent, and relevant customer references, with email generation assistance, encompassing phrase count, tone, and syntax, as well as effective email best practices. We collaborated meticulously with our experienced enterprise growth specialists to craft tailored prompts, which served as a vital training ground for the models. This strategy has consistently delivered impressive results; AI-generated emails sent through our model demonstrate a response and engagement profile comparable to that of a seasoned sales or business growth consultant, yielding click-through rates ranging from 30% to 60% and reply rates between 3% and 5%. The price per email has dropped significantly, from $0.07 to just $0.005, thanks to the adoption of a refined, open-source model that has streamlined operations. Prior to sending emails to prospects, our Sales Development Representatives (SDRs) have complete creative control over the content. Our proprietary technology and meticulous editorial process incorporate robust safeguards to eliminate hallucinations and irrelevant information, guaranteeing that our email initiatives are laser-focused and impactful.
One highly promising software solution for inside gross sales representatives is our innovative AI-powered sales-based agent language model. By leveraging the ‘hover’ chatbot’s capabilities, this solution provides data-driven insights to sales teams regarding potential product offerings and usage scenarios for a specific company, enabling informed decision-making. Customers in Salesforce leverage the software to stay informed about the latest changes within an organization prior to a meeting, thereby enabling them to develop targeted and relevant interventions, such as migrating to a cloud platform or building a new data warehouse using structured insights from related firms. The key aspect of the mannequin’s performance lies in its ability to seamlessly integrate both structured Salesforce data and unstructured information from internal and external sources, while ensuring controlled access management and adhering to strict guidelines on knowledge confidentiality?
We are also exploring innovative methods in contract administration by developing a GenAI software tool that streamlines contract summarization processes. The organization may take into account unconventional expressions and scenarios that contradict established wisdom in Salesforce, thereby determining the scope of liability and permissible risk associated with a specific agreement. The implementation of auto-summarization capabilities enables expedited contract processing, thereby reducing the burden on our internal authorized teams. This initiative is underpinned by a comprehensive AI governance and security framework, jointly developed with our safety and privacy experts to ensure robust protection.
Key issues
Regardless of whether you’re navigating experimental use cases or building upon successes, several key considerations must be taken into account when developing GenAI.
- We leverage a blend of structured and unstructured knowledge, utilizing RAG-based frameworks to deliver actionable insights and mitigate hallucinations; increasingly, we rely on our dedicated Databricks RAG Studio platform to validate fashion efficacy, ensuring optimal return on investment (ROI) and minimized costs. By leveraging tailored prompts and combining them with business knowledge using the Databricks Intelligence Platform, organisations can streamline experimentation and accelerate learning from results.
These approaches provide stability to pace and ensure high-quality performance, which can be fine-tuned or integrated into a language model’s pre-training process. Measuring efficiency against distinct campaigns and styles underscores the profitability benefits for the corporation and its various stakeholders.
- Regularly assessing and refining employee expertise from the outset to throughout the technology’s lifespan guarantees optimal utilization and fosters seamless integration of the innovation. Collaborative efforts across various teams can lead to sustained and widespread changes throughout the organization. Establishing standardized protocols ensures that expertise is utilized consistently and effectively.
- As organisations scale their use of AI, challenges arise in complexity, but these are far from insurmountable. While acknowledging that knowledge can be chaotic and testing a challenge, many organizations can mitigate these issues by implementing various strategies. By harnessing the capabilities of lakehouses, implementing an incremental approach to database scaling, and developing a framework for evaluating organisational impact following process validation, key milestones can be achieved. Ensuring seamless transitions between ML Ops tiers, strategically allocating resources for focused intervals to deliver top-notch prompts, and ensuring that solutions yield tangible, data-driven insights are crucial considerations.
- This encompasses IT and governance capabilities, both of which can inform considerations around ROI.
Going forward, Databricks is driving innovation by applying GenAI to a wide range of high-value use cases within the company, including but not limited to: enhancing enterprise operations through AI-powered deal desk and IT support; boosting productivity with features like account alerts, content discovery, and meeting preparation; streamlining marketing efforts via AI-generated content and outbound prospecting; optimizing HR processes with ticket deflection and recruiting effectiveness; revolutionizing legal services with contract knowledge extraction; and unlocking business insights through self-serve, ad-hoc analytics queries. Notwithstanding, we’re not discounting the value of GenAI for our external customer base.
JetBlue leveraged its knowledge intelligence platform in tandem with sophisticated open-source large language models (LLMs) to develop a chatbot, empowering employees to access customized key performance indicators (KPIs) and data relevant to their specific roles. The impression of this answer has been to reduce coaching necessities and accelerate turnaround times for feedback, thereby simplifying access to insights across the entire team. EasyJet, a European provider, developed a conversational AI solution designed for non-technical users to ask voice-based questions in their natural language and receive insights that can inform the decision-making process. This answer has not only helped enhance the group’s knowledge techniques, but also provided customers with seamless access to information and LLM-driven insights, thereby sparking innovative concepts around various revolutionary GenAI use cases, including resource optimization, chatbots focused on operational processes and compliance, as well as personal assistants that offer tailored travel suggestions.
As organizations deploy General Artificial Intelligence (GenAI) solutions, it’s crucial to balance safety, governance, and ROI considerations, our expertise reveals that when companies harness GenAI’s cross-functional capabilities through iterative experimentation, the potential effectiveness gains can yield a significant competitive advantage for both themselves and their customers.
