7Targets is a SaaS B2B company in the Sales and Marketing domain focused on lead nurturing. It has a flagship product ‘AI Sales Assistant’ which has the capability to write emails and do follow-ups with the leads. As well, to read the received responses and act accordingly.
- AI Sales assistants that help sales teams win more customers
- AI assistant automatically follows up with all your leads to convert more leads into Hot leads so that you can focus on closing more deals
Being an email-based AI Assistant, 7Targets wanted to make sure that a scalable and working solution is created to send and receive the emails. The received email is read, categorized, understood and entities extracted from. So that further action can be taken by the assistant.Specific project-level challenges were:
- Ability to send email evenly distributed over time to avoid sudden spike
- Ability to read the email received and decide the next action to be taken by the ‘AI Sales Assistant’ after proper categorisation of the received email.
- Extract entities like date, phone number, title, etc from the email received.
- Set up a pipeline for continuous deployment with proper testing in the pipeline.
Applied AI Consulting worked on a solution
- To create the step functions (state machines process-lead, process-lead-response, etc) with various lambda functions for each step.
- Where the received email is accepted in the s3 bucket using SES.
- Process the received response and categorize based on sentiment and certain properties available in the response like date, etc.
- Handle any bounce and complaint emails responses too
This has helped solve the problem of sending the email and then the follow-up email at the right time too. The right categorization of the received email, based on which lead state was updated and the action ahead was decided by a state machine.Along with categorization from the email text, there was also a need for entity detection like date, phone number, etc. which was done as part of this implementation. For e.g., if an email was received and categorized to be ‘Out of Office’ then the return date was identified from the email so that the assistant can follow-up after the leads return to the office instead of following upin the Out of office period.
- Debounce and Zerobounce API is used to validate the emails of each lead. This is the first step in the process-lead step function
- 3rd party CRM like Hubspot, ZohoCRM is used and the 7Targets AI Assistant solution interacts with those solution
AWS Stepfunctions, AWS Lambda, AWS RDS, AWS Comprehend, AWS S3, AWS SES, AWS ECR, AWS Fargate, AWS SNS, AWS SQS, AWS API Gateway, AWS Cloudwatch, AWS ELB, AWS Cognito, AWS WAF, AWS CloudFormation, AWS DMS, AWS Route53
Able to grow per day email sending from 1K to like 10K and per day and can still grow more with the present solution.Able to process the response at a similar level but the responses are not as high as email sending.Will proper limits and thresholds be set and with continuous monitoring of failure for lambda andstep functions the limits and thresholds are changed.The container in Fargate is auto configured to scale up and down.The Continuous deployment pipelines are helping with automated zero downtime deployments to QA and then production setup.
Architecture diagrams of the specific customer deployment
Pipeline for deployment
AAIC is a technology services company that provides
high-end engineering services and is proficient in
enterprise-class for IT delivery.
AAIC focuses on Automation in Cloud and DevOps