How GoDaddy constructed Lighthouse, an interplay analytics resolution to generate insights on help interactions utilizing Amazon Bedrock

This put up is co-written with Mayur Patel, Nick Koenig, and Karthik Jetti from GoDaddy.

GoDaddy empowers on a regular basis entrepreneurs by offering all the assistance and instruments to succeed on-line. With 21 million prospects worldwide, GoDaddy’s world options assist seamlessly join entrepreneurs’ id and presence with commerce, resulting in worthwhile development. At GoDaddy, we take satisfaction in being a data-driven firm. Our relentless pursuit of useful insights from knowledge fuels our enterprise choices and works to realize buyer satisfaction.

On this put up, we focus on how GoDaddy’s Care & Companies staff, in shut collaboration with the  AWS GenAI Labs staff, constructed Lighthouse—a generative AI resolution powered by Amazon Bedrock. Amazon Bedrock is a completely managed service that makes basis fashions (FMs) from main AI startups and Amazon out there via an API, so you possibly can select from a variety of FMs to search out the mannequin that’s greatest suited on your use case. With the Amazon Bedrock serverless expertise, you may get began shortly, privately customise FMs with your individual knowledge, and combine and deploy them into your functions utilizing the AWS instruments with out having to handle infrastructure. With Amazon Bedrock, GoDaddy’s Lighthouse mines insights from buyer care interactions utilizing crafted prompts to establish prime name drivers and cut back friction factors in prospects’ product and web site experiences, resulting in improved buyer expertise.

GoDaddy’s enterprise problem

Information has at all times been a aggressive benefit for GoDaddy, as has the Care & Companies staff . We notice the potential to derive significant insights from this knowledge and establish key name drivers and ache factors. On the earth earlier than generative AI, nevertheless, the expertise for mining insights from unstructured knowledge was computationally costly and difficult to operationalize.

Resolution overview

This modified with GoDaddy Lighthouse, a generative AI-powered interactions analytics resolution, which unlocks the wealthy mine of insights sitting inside our buyer care transcript knowledge. Fed by buyer care interactions knowledge, it allows scale for deep and actionable evaluation, permitting us to:

• Detect and dimension buyer friction factors in our product and web site experiences, resulting in enhancements in buyer expertise (CX) and retention
• Enhance buyer care operations, together with high quality assurance and routing optimization, resulting in enhancements in CX and operational expenditure (OpEx)
• Deprecate our reliance on expensive vendor options for voice analytics

The next diagram illustrates the high-level enterprise workflow of Lighthouse.

GoDaddy Lighthouse is an insights resolution powered by massive language fashions (LLMs) that enables immediate engineers all through the corporate to craft, handle, and consider prompts utilizing a portal the place they’ll work together with an LLM of their alternative. By engineering prompts that run in opposition to an LLM, we are able to systematically derive highly effective and standardized insights throughout text-based knowledge. Product subject material consultants use the Lighthouse platform UI to check and iterate on generative AI prompts that produce tailor-made insights a couple of Care & Companies interplay.

The beneath diagram reveals the iterative course of of making and strengthening the prompts.

After the prompts are examined and confirmed to work as supposed, they’re deployed into manufacturing, the place they’re scaled throughout 1000’s of interactions. Then, the insights produced for every interplay are aggregated and visualized in dashboards and different analytical instruments. Moreover, Lighthouse lets GoDaddy customers craft one-time generative AI prompts to disclose wealthy insights for a extremely particular buyer situation.

Let’s dive into how the Lighthouse structure and options help customers in producing insights. The next diagram illustrates the Lighthouse structure on AWS.

The Lighthouse UI is powered by knowledge generated from Amazon Bedrock LLM calls on 1000’s of transcripts, using a library of prompts from GoDaddy’s inside immediate catalog. The UI facilitates the choice of LLM mannequin primarily based on the consumer’s alternative, making the answer impartial of 1 mannequin. These LLM calls are processed sequentially utilizing Amazon EMR and Amazon EMR Serverless. The seamless integration of backend knowledge into the UI is facilitated by Amazon API Gateway and Amazon Lambdas capabilities, whereas the UI/UX is supported by AWS Fargate and Elastic Load Balancing to take care of excessive availability. For knowledge storage and retrieval, Lighthouse employs a mixture of Amazon DynamoDB, Amazon Simple Storage Service (Amazon S3), and Amazon Athena. Visible knowledge evaluation and illustration are achieved via dashboards constructed on Tableau and Amazon QuickSight.

Immediate analysis

Lighthouse provides a novel proposition by permitting customers to guage their one-time generative AI prompts utilizing an LLM of their alternative. This function empowers customers to jot down a brand new one-time immediate particularly for analysis functions. Lighthouse processes this new immediate utilizing the precise transcript and response from a earlier LLM name.

This functionality is especially useful for customers in search of to refine their prompts via a number of iterations. By iteratively adjusting and evaluating their prompts, customers can progressively improve and solidify the effectiveness of their queries. This iterative refinement course of makes certain that customers can obtain the highest-quality outputs tailor-made to their particular wants.

The pliability and precision provided by this function make Lighthouse an indispensable instrument for anybody attempting to optimize their interactions with LLMs, fostering steady enchancment and innovation in immediate engineering.

The next screenshot illustrates how Lighthouse lets customers validate the accuracy of the mannequin response with an analysis immediate

After a immediate is evaluated for high quality and the consumer is happy with the outcomes, the immediate might be promoted into the immediate catalog.

Response summarization

After the consumer submits their immediate, Lighthouse processes this immediate in opposition to every out there transcript, producing a sequence of responses. The consumer can then view the generated responses for that question on a devoted web page. This web page serves as a useful useful resource, permitting customers to evaluate the responses intimately and even obtain them into an Excel sheet for additional evaluation.

Nevertheless, the sheer quantity of responses can generally make this job overwhelming. To handle this, Lighthouse provides a function that enables customers to move these responses via a brand new immediate for summarization. This performance allows customers to acquire concise, single-line summaries of the responses, considerably simplifying the evaluate course of and enhancing effectivity.

The next screenshot reveals an instance of the immediate with which Lighthouse lets customers meta-analyze all responses into one, lowering the time wanted to evaluate every response individually.

With this summarization instrument, customers can shortly distill massive units of knowledge into simply digestible insights, streamlining their workflow and making Lighthouse an indispensable instrument for knowledge evaluation and decision-making.

Insights

Lighthouse generates useful insights, offering a deeper understanding of key focus areas, alternatives for enchancment, and strategic instructions. With these insights, GoDaddy could make knowledgeable, strategic choices that improve operational effectivity and drive income development.

The next screenshot is an instance of the dashboard primarily based on insights generated by Lighthouse, exhibiting the distribution of classes in every perception.

By way of Lighthouse, we analyzed the distribution of root causes and intents throughout the huge variety of each day calls dealt with by GoDaddy brokers. This evaluation recognized probably the most frequent causes of escalations and components most probably to result in buyer dissatisfaction.

Enterprise worth and impression

So far (as of the time of writing), Lighthouse has generated 15 new insights. Most notably, the staff used insights from Lighthouse to quantify the impression and price of the friction inside the present course of, enabling them to prioritize needed enhancements throughout a number of departments. This strategic method led to a streamlined password reset course of, lowering help contacts associated to the password reset course of and shortening decision occasions, in the end offering important price financial savings.

Different insights resulting in enhancements to the GoDaddy enterprise embrace:

• The invention of name routing flows suboptimal to revenue per interplay
• Understanding the basis reason for repeat contact interactions

Conclusion

GoDaddy’s Lighthouse, powered by Amazon Bedrock, represents a transformative leap in utilizing generative AI to unlock the worth hidden inside unstructured buyer interplay knowledge. By scaling deep evaluation and producing actionable insights, Lighthouse empowers GoDaddy to boost buyer experiences, optimize operations, and drive enterprise development. As a testomony to its success, Lighthouse has already delivered monetary and operational enhancements, solidifying GoDaddy’s place as a data-driven chief within the trade.

Concerning the Authors

Mayur Patel is a Director, Software program Growth within the Information & Analytics (DnA) staff at GoDaddy, specializing in knowledge engineering and AI-driven options. With almost 20 years of expertise in engineering, structure, and management, he has designed and carried out modern options to enhance enterprise processes, cut back prices, and enhance income. His work has enabled firms to realize their highest potential via knowledge. Enthusiastic about leveraging knowledge and AI, he goals to create options that delight prospects, improve operational effectivity, and optimize prices. Exterior of his skilled life, he enjoys studying, climbing, DIY tasks, and exploring new applied sciences.

Nick Koenig is a Senior Director of Information Analytics and has labored throughout GoDaddy constructing knowledge options for the final 10 years. His first job at GoDaddy included listening to calls and discovering developments, so he’s significantly proud to be concerned in constructing an AI resolution for this a decade later.

Karthik Jetti is a Senior Information Engineer within the Information & Analytics group at GoDaddy. With greater than 12 years of expertise in engineering and structure in knowledge applied sciences, AI, and cloud platforms, he has produced knowledge to help superior analytics and AI initiatives. His work drives technique and innovation, specializing in income technology and bettering effectivity.

Ranjit Rajan is a Principal GenAI Lab Options Architect with AWS. Ranjit works with AWS prospects to assist them design and construct knowledge and analytics functions within the cloud.

Satveer Khurpa is a Senior Options Architect on the GenAI Labs staff at Amazon Internet Companies. On this position, he makes use of his experience in cloud-based architectures to develop modern generative AI options for purchasers throughout various industries. Satveer’s deep understanding of generative AI applied sciences permits him to design scalable, safe, and accountable functions that unlock new enterprise alternatives and drive tangible worth.

Richa Gupta is a Options Architect at Amazon Internet Companies specializing in generative AI and AI/ML designs. She helps prospects implement scalable, cloud-based options to make use of superior AI applied sciences and drive enterprise development. She has additionally introduced generative AI use instances in AWS Summits. Previous to becoming a member of AWS, she labored within the capability of a software program engineer and options architect, constructing options for giant telecom operators.