Project Kick Off - VISL AI

AI Driven Inbound Case Classification
with Salesforce Einstein CoPilot

Agenda
● Our Understanding of Requirements
● Out of Scope
● Assumptions
● Technical Solution Approach
● Release Plan
● Proposed Engagement Timeline

Our Understanding of Requirements
● AI Based Case Classification
○ New Case for every new email from sales@vashi.com considered for AI classification. New email address onboarding will
require data set training & configuration.
○ Case Classification is performed when
■ Case type is tagged into following categories only – Order / Inquiry / Complaints / Invoice / Other. Business
precedence for Case classification – Order & Inquiry
■ Brand field = Respective Brand Name or Multiband
■ Owner = Case Type Specific User / Brand Queue
○ Existing emails on ongoing threads excluded from AI Case Classification. Thread ID to be used as reference to differentiate
existing threads versus new email threads
○ Manual override privilege for Case & Brand classification available to pre-identified Salesforce users only
● Data Cloud Config & Grounding with RAG

○ RAG and Data Stream config for dynamic context and grounding
○ Batch Sync Mode based monthly refresh
● Dataset Training
○ The VISL team has agreed to provision the Case and Email Message data required for AI Dataset training from a rolling
period of the last 12 months.
○ Any change in the volume of available data set i.e. number of Case and Email Message records and the quality of the
content, for training will directly influence the accuracy of the generated content by the Salesforce Einstein CoPilot based AI
automation engine.
3
Volumetric Analysis & Considerations
Email Message
• Inbound Emails = Daily 3K / Monthly 90K / Annually 1.08 Million Emails
Case
• Cases (Type = Order) = ~Daily 300 / Monthly 9.4k / Annually 82k
• Cases (Type = Inquiry) = ~ Daily 600/Monthly 20k / Annually 1.7 Lakh
• Cases (Type = Complaint) = ~ Daily 7/ Monthly 240 / Annually 1.8k
• Cases (Type = Invoice) = ~ Daily 80 / Monthly 1.2k / Annually 23k
• Cases (Type = Other) = ~ Daily 600 / Monthly 18.6k / Annually 1.9 Lakh
4
Case Classification Process Flow
5
Out of Scope
1. Reading of content from inbound or outbound email attachments using the Salesforce Einstein CoPilot based AI engine is considered
outside the current scope of implementation.
2. Any process re-engineering is outside the current scope
3. Summarization of content from inbound and outbound email attachments and presenting on the Case record as a part of a field
using the Salesforce Einstein CoPilot based AI engine is considered outside the current scope of implementation.
4. Inbound emails originating from noncustomer Vashi email addresses will not be accounted for by Salesforce Einstein driven
automated case classification.
5. Any changes to the objects, workflows or code that are not related to the SOW agreed upon by both parties.
6. Any additional activities requiring skill sets not pertaining to the current SOW.
7. Any Local/ International travel for the resource(s)
6
Assumptions
• Data cloud credits for licensing will account against development, testing and actual production activities. This is governed by the
Salesforce licensing policy.
• Data Cloud Credits will be utilized based on the data volume and Frequency in Data Stream
• Salesforce Data cloud credit consumption will apply against the Salesforce allotted credits as per the Salesforce licensing contract.
Refer
https://www.salesforce.com/content/dam/web/en_us/www/documents/platform/september-17-data-cloud-platform-services-rat
e-sheet.pdf for further details
• Client users are expected to use relevant Salesforce licenses
• All Salesforce enforced Governor Limits will apply and will hold true to the implementation
• Client will be responsible to carry out the User Acceptance Testing on the UAT environment within the agreed timeframe.
• InfoBeans will ensure relevant and clear documentation for the implementation.
• InfoBeans will require having discussions with the client team and/or end users to understand the requirement and existing system in
detail.
7
Technical Solution Approach
8
Release Plan
9
Project Timeline
● Engagement Start Date – Sep 19, 2024
● Engagement End Date – Oct 25, 2024
● Potential Duration - ~ 5 weeks
10
Asks from Client
● Project Kickoff
○ Business rule hierarchy used for Case Type & Brand Classification
○ Classification of existing live Closed Case into Order, Inquiry,
Complaints, Invoice, Other for rolling period of 12 months
○ Classification of existing live Email Messages associated with closed
Cases into Order, Inquiry, Complaints, Invoice, Other for rolling period
of 12 months
● User Acceptance
○ Live production Case and Email Message records in CSV format from
prior business day
11
Let’s begin the
WOW! journey
THANK YOU
Q&A

Project Kick Off - VISL AI

Uploaded by

Copyright:

Available Formats

Project Kick Off - VISL AI

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Project Kick Off - VISL AI

Uploaded by

Copyright:

Available Formats

AI Driven Inbound Case Classiﬁcation

with Salesforce Einstein CoPilot

● Technical Solution Approach

● Proposed Engagement Timeline

● Data Cloud Conﬁg & Grounding with RAG

• Inbound Emails = Daily 3K / Monthly 90K / Annually 1.08 Million Emails

• Cases (Type = Order) = ~Daily 300 / Monthly 9.4k / Annually 82k

• Cases (Type = Inquiry) = ~ Daily 600/Monthly 20k / Annually 1.7 Lakh

• Cases (Type = Complaint) = ~ Daily 7/ Monthly 240 / Annually 1.8k

• Cases (Type = Invoice) = ~ Daily 80 / Monthly 1.2k / Annually 23k

2. Any process re-engineering is outside the current scope

7. Any Local/ International travel for the resource(s)

• Client users are expected to use relevant Salesforce licenses

● Engagement End Date – Oct 25, 2024

● Potential Duration - ~ 5 weeks

You might also like