AI driven Sales Productivity Enhancement Using Domain Specific Web Search, Data Analysis, ML and Large Language Models.
This story talks about one of our customer projects where we created an ETL data pipeline that included Web Data Extraction, Intelligent search and Classification, NER , Semantic Matching and LLMs to meet all of the functional requirements for this B2B Business .
We will look into step by step process of how we made it happen !
The Idea :
The idea was to create a platform for self-updating high probable opportunities, relevant stakeholders along with other account insights and targeted content for customer interaction .
The Solution Pipeline :
We started with web search specific to news and articles relevant to companies which would map as Events to the list of Sales Opportunities for Consulting firms .
This would result in Reduction in effort spent in research/ insights gathering by the salesperson of the firms.
The Solution Pipeline was tailored for Technology services/ solutions/consulting organizations which deal with relatively high value consultative sales.
Once data gets collected from web as Events, we run NER [ Named Entity Recognition] over the content to get various aspects such as :
- Company Name [ which would later help us in mapping it to role in this company and look for the right people to be contacted for this opportunity] ,
- Specific Event [ which would map to the opportunity for the sales person in the consulting/services firm ]
- Other Content Elements[ which would help in formulating email later using LLM]
We persist this in the database from where it gets consumed by downstream applications to be combined with other data sources .
Other data sources include :
- Company specific data such as list of business events for this particular firm. [ ex. Service Catalog for this firm]
- Transactional Data [ past interaction record for this firm for training and learning of ML Algorithms to render better results for upcoming events.
- Master Data of the companies [ this is commercially acquired by our customer ] — comes in form of table containing details about companies such as company’s mapping to its verticals which helps our intelligence to map events in companies to their corresponding roles in the company to be contacted for business opportunities.
- Master Data of people [ this is commercially acquired by our customer ] — comes in form of table containing details about people belonging to all companies where opportunities can be offered via contact information of these profiles.
Steps :
Combine various data sources to create connection between a business event which has come in news/article and an opportunity for the salesperson to contact the relevant profiles of relevant company/companies with details of opportunities via auto-generated email having entire context curated according to the event and opportunity.
To enable this, the backend data is prepared through the following steps :
step1. Article/event picked from the web
step2. Data cleaning and Data Quality Validation.
step3. NER creates set of entities of interest → Event saved in event table
step4. Company information mapped to Master data of companies → saved for getting roles for profiles to be contacted.
step5. Events information mapped to service catalog of consulting firm → mapped to Events transaction data of the firm to map to opportunities.
step6. Article content saved for downstream process of summary generation for this opportunity
step7. context creation by combining role, opportunity and verticals [ from company master data] → to filter profiles from people master data and get their LinkedIn profile and contact information.
step8. Email generation using LLM
All this information is stored to be viewed by the end user, here the registered salesperson from the UI. The salesperson goes through the following steps in order to use the provided functionalities :
step1. registration → provides his/her name, company name, email ID.
step2. Some recommended customer accounts are shown after registration based on collected information
step3. Salesperson chooses the target company and data is rendered on UI about news and events around that company . Ex. for GE — Merger And Acquisition [ M&A ] is an event. A news by Financial times is presented on the UI.
step4. User can also chose ‘Theme’ [ or vertical] from this company to get specific profiles information belonging to those roles.
step5. Pick the Usecase [ or ‘Opportunity’] from within that vertical to track them whenever an event occurs. The ‘Leads’ tab gives information on Leads for opportunities .
step6. View the list of lead for the specific usecase.
step7. The user has the option to view the client persona of the selected profile :
step8. Select the profile and send email for the selected event and opportunity.
This marked the end of set of features that we developed for our customer .
