How AWS Gross sales makes use of generative AI to streamline account planning

Yearly, AWS Gross sales personnel draft in-depth, ahead wanting technique paperwork for established AWS clients. These paperwork assist the AWS Gross sales crew to align with our buyer development technique and to collaborate with the complete gross sales crew on long-term development concepts for AWS clients. These paperwork are internally known as account plans (APs). In 2024, this exercise took an account supervisor (AM) as much as 40 hours per buyer. This, mixed with comparable time spent for help roles researching and writing the expansion plans for patrons on the AWS Cloud, led to important group overhead. To assist enhance this course of, in October 2024 we launched an AI-powered account planning draft assistant for our gross sales groups, constructing on the success of Discipline Advisor, an inner gross sales assistant instrument. This new functionality makes use of Amazon Bedrock to assist our gross sales groups create complete and insightful APs in much less time. Since its launch, hundreds of gross sales groups have used the ensuing generative AI-powered assistant to draft sections of their APs, saving time on every AP created.

On this put up, we showcase how the AWS Gross sales product crew constructed the generative AI account plans draft assistant.

Enterprise use circumstances

The account plans draft assistant serves 4 major use circumstances:

• Account plan draft era: Utilizing Amazon Bedrock, we’ve made inner and exterior information sources accessible to generate draft content material for key sections of the APs. This allows our gross sales groups to shortly create preliminary drafts for sections reminiscent of buyer overviews, trade evaluation, and enterprise priorities, which beforehand required hours of analysis throughout the web and relied on disparate inner AWS instruments.
• Knowledge synthesis: The assistant can pull related info from a number of sources together with from our buyer relationship administration (CRM) system, monetary reviews, information articles, and former APs to supply a holistic view of our clients.
• High quality checks: Constructed-in high quality assurance capabilities assist be sure that APs meet inner requirements for comprehensiveness, accuracy, and strategic alignment with our clients and enterprise.
• Customization: Whereas offering AI-generated drafts, the product permits AMs to customise and refine the content material by importing proprietary paperwork to match their distinctive buyer information and strategic method.

The account plan draft assistant hundreds when a consumer tries to create an AP, and customers copy and paste every part they wish to use of their remaining plan.

Our AMs report decreased time to write down these paperwork, permitting them to focus extra on high-value actions reminiscent of buyer engagement and technique growth.

Right here’s what a few of our AMs needed to say about their expertise with the account plans draft assistant:

“The AI assistant saved me at the very least 15 hours on my newest enterprise account plan. It pulled collectively an excellent first draft, which I used to be then capable of refine primarily based by myself insights. This allowed me to spend extra time really partaking with my buyer quite than doing analysis and writing.”

– Enterprise Account Supervisor

“As somebody managing a number of mid-market accounts, I struggled to create in-depth plans for all my clients. The AI assistant now helps me quickly generate baseline plans that I can then prioritize and customise. It’s a game-changer for serving my full portfolio of accounts.”

– Mid-market Account Supervisor

Amazon Q, Amazon Bedrock, and different AWS providers underpin this expertise, enabling us to make use of giant language fashions (LLMs) and information bases (KBs) to generate related, data-driven content material for APs. Let’s discover how we constructed this AI assistant and a few of our future plans.

Constructing the account plans draft assistant

When a consumer of the AWS inner CRM system initiates the workflow in Discipline Advisor, it triggers the account plan draft assistant functionality by means of a pre-signed URL. The assistant then orchestrates a multi-source information assortment course of, performing net searches whereas additionally pulling account metadata from OpenSearch, Amazon DynamoDB, and Amazon Easy Storage Service (Amazon S3) storage. After analyzing and mixing this information with user-uploaded paperwork, the assistant makes use of Amazon Bedrock to generate the AP. When full, a notification chain utilizing Amazon Easy Queue Service (Amazon SQS) and our inner notifications service API gateway begins delivering updates utilizing Slack direct messaging and storing searchable information in OpenSearch for future reference.

The next diagram illustrates the high-level structure of the account plans draft assistant.

Answer overview

We constructed the account plans draft assistant utilizing the next key elements:

1. Amazon Bedrock: Offers programmatic (API) entry to excessive performing basis fashions (FMs) together with vector search capabilities and metadata filtering utilizing Amazon Bedrock Data Bases. We populate an Amazon Bedrock information bases utilizing sales-enablement supplies, historic APs, and different related paperwork curated by AWS Glue jobs (see extra on AWS Glue jobs within the merchandise 4).
2. AWS Lambda: Helps two use circumstances:
   1. The async resolver Lambda perform interfaces with the front-end shopper CRM and orchestrates async job IDs for the shopper to ballot. This layer additionally handles enter validations, consumer request throttling and cache administration.
   2. Employee Lambda features carry out the precise heavy lifting to create AP content material. These features work concurrently to generate totally different sections of APs through the use of publicly accessible information, inner information, and curated information in Amazon Bedrock information bases. These features invoke numerous LLMs utilizing Amazon Bedrock and retailer the ultimate content material within the AP’s DynamoDB database corresponding to every async job ID.
3. DynamoDB: Maintains the state of every consumer request by monitoring async job IDs, tracks throttling quota (world request depend and per-user request depend), and acts as a cache.
4. AWS Glue jobs: Curate and rework information from numerous inner and exterior information sources. These AWS Glue jobs push information to inner information sources (APs, inner tooling crew S3 buckets, and different inner providers) and to Bedrock KBs, facilitating prime quality output by means of retrieval augmented era (RAG).
5. Amazon SQS: Permits us to decouple the administration airplane and information airplane. This decoupling is essential in permitting the info airplane employee features to concurrently course of totally different sections of the APs and make it possible for we will generate APs inside specified instances.
6. Customized net frontend: A ReactJS primarily based micro-frontend structure allows us to combine instantly into our CRM system for a seamless consumer expertise.

Knowledge administration

Our account plans draft assistant makes use of an Amazon Bedrock out-of-the-box information base administration resolution. By means of its RAG structure, we semantically search and use metadata filtering to retrieve related context from numerous sources: inner gross sales enablement supplies, historic APs, SEC filings, information articles, government engagements and information from our CRM methods. The connectors constructed into Amazon Bedrock deal with information ingestion from Amazon S3, relational database administration methods (RDBMS), and third-party APIs; whereas its KB capabilities allow us to filter and prioritize supply paperwork when producing responses. This context-aware method leads to increased high quality and extra related content material in our generated AP sections.

Safety and compliance

Safety and Compliance are paramount to AWS when coping with information concerning our clients. We use AWS IAM Identification Middle for enterprise single sign-on in order that solely licensed customers can entry the account plans draft assistant. Utilizing Discipline Advisor, we use numerous inner authorization mechanisms to assist be sure that a consumer who’s producing APs solely accesses the info that they have already got entry to.

Person expertise

We constructed a customized net frontend utilizing a micro-frontend method that integrates instantly into our CRM system, permitting AMs to entry the account plans draft assistant with out leaving their acquainted work setting. The interface permits customers to pick out which sections of APs they wish to generate, supplies choices for personalization, and notifies customers to create their APs on time by means of Slack.

Trying forward

Whereas the account plans draft assistant has already demonstrated important worth, we’re persevering with to boost its capabilities. Our aim is to create a zero-touch account planner that gross sales groups can use to generate a full AP for a buyer, incorporating finest practices noticed throughout our clients to supply gross sales groups best-in-class methods to have interaction with clients. This would come with:

•  Deeper integration with our bespoke purpose-built planning instruments and help with account planning, reminiscent of robotically producing worth maps and stakeholder maps.
• Enhanced personalization to tailor content material primarily based on trade, account dimension, and particular person consumer preferences.
• Improved collaboration options, in order that a number of gross sales crew members can work collectively on refining AI-generated plans.
• Expanded use of suggestions to supply what subsequent? concepts to our gross sales groups to higher serve our clients.

Conclusion

The account plans draft assistant, powered by Amazon Bedrock, has considerably streamlined our AP course of, permitting our AWS Gross sales groups to create increased high quality APs in a fraction of the time they at the moment want. As we proceed to refine and develop this functionality, we’re excited to see the way it will additional improve our capability to serve our clients and drive their success within the AWS Cloud.

In the event you’re excited by studying how generative AI can rework your gross sales perform and its processes, attain out to your AWS account crew to debate how providers reminiscent of Amazon Q and Amazon Bedrock will help you construct comparable options in your group.

In regards to the Authors

Saksham Kakar is a Sr. Product Supervisor (Technical) within the AWS Discipline Experiences (AFX) group targeted on creating merchandise that allow AWS Gross sales groups to assist AWS clients develop with Amazon. Previous to this, Saksham led giant gross sales, technique and operations groups throughout startups and Fortune 500 firms. Outdoors of labor, he’s an avid tennis participant and novice skier.

Vimanyu Aggarwal is a Senior Software program Engineer in AWS Discipline Experiences (AFX) group with over 10 years of trade expertise. During the last decade, Vimanyu has been specializing in constructing large-scale, advanced distributed methods at numerous Fortune 500 organizations. At the moment, he works with a number of groups inside the AFX group to ship technical options that empower the $100 billion gross sales funnel. Outdoors of labor, he likes to play board video games, tinker with IoT, and discover nature.

Krishnachand Velaga is a Senior Supervisor for Product Administration – Technical (PM-T) within the AWS Discipline Experiences (AFX) group who manages a crew of seasoned PM-Ts and a set of gross sales merchandise, utilizing generative AI to allow the AWS Gross sales group assist AWS clients throughout the globe undertake, migrate and develop on the AWS Cloud consistent with their enterprise wants and outcomes whereas bolstering gross sales effectivity and productiveness and decreasing operational price.

Scott Wilkinson is a Software program Growth Supervisor within the AWS Discipline Experiences (AFX) group, the place he leads a cross-functional engineering crew creating instruments that mixture and productize information to energy AWS buyer insights. Previous to AWS, Scott labored for notable startups together with Digg, eHarmony, and Nasty Gal in each management and software program growth roles. Outdoors of labor, Scott is a musician (guitar and piano) and likes to prepare dinner French delicacies.