Tips for passing Salesforce AI Specialist Certification

 

Motivation behind this

Salesforce has come up AI Specialist on 9th Sept,2024, which is next level to AI Associate certification. As Salesforce is talking about AI + CRM + Data Cloud all over, which drives me to brushing up the knowledge on this area and successfully pass this certification on the launched day. 

In addition to that, it is the 40th certification in my Salesforce career. Special for me!

Sharing my tips to prepare and pass this certification for trailblazers.

Exam Outline

Exam Guide

Content: 60 multiple-choice questions

Time allotted to complete the exam: 105 minutes

Passing score: 73% (44 answers to correct)

Registration fee: Free $200 plus applicable taxes as required per local law

Retake fee: $100

Happy to see me passed.

Topics

Below are the topics for AI Specialist exam.

Preparation

1) Trailmix: Drive Productivity with Einstein AI

2) Enablement Series:

• Watch: Protect Data Security & Privacy with the Einstein Trust Layer
• Watch: Inside the Einstein Trust Layer
• Watch: Enhance Customer Service and Sales with Generative AI
• Watch: Personalize at Scale with Prompt Builder
• Watch: Maximize Efficiency with Einstein Copilot
• Watch: Discover Salesforce Model Builder

Exam Structure

For each question there are only 3 options and only one option to choose. As there are no multi-choice answers, so it will be less confusing on choosing the answers.

Few Important points to cover for exam

Though this certification doesn't have any pre-requisite, but clearing AI Associate certification would help to build the foundation.

Einstein Trust Layer (15%)

Refer documentation

Prompt Journey

• Prompt : which comprise of clear instructions, contextual information, and constraints help which is being sent to LLM.
• Secure Data Retrieval with Dynamic Grounding: 
• Relevant info from Salesforce Record is merged with prompt template
• Respects user permission, role hierarchy based access, field level security
• Data Masking: Sensitive data is masked, it supports multiple regions & languages
• Prompt Defense: System policies help limit hallucinations and decrease the likelihood of unintended or harmful outputs by the LLM.

Response Journey

• Secure LLM Gateway: Prompt to leave secure trust boundary by passing through the Secure LLM Gateway to connected LLM.
• Zero Data Retention: Salesforce partnered with Open AI and Azure Open AI to enforce the policy:
• No data is used for LLM model training or product improvements by third-party LLMs.
• No data is retained by the third-party LLMs.
• No human being at the third-party provider looks at data sent to their LLM.
• Toxicity detection and demasking: It removes toxic, hateful, violent, sexual, identifiable, physical, and profane responses. It dehydrates same tokenized data.
• Toxicity scoring: Toxicity scores are logged and stored in Data Cloud as part of the audit trail. The score for safety category ranges from 0 to 1, with 1 being safest.
• Auditing: 
• Prompts, responses, and trust signals are logged and stored in Data Cloud.
• Feedback can be used for improving prompt templates.
• Provides pre-built reports and dashboards for analysis.

There are questions on all the points mentioned above and it covers 9 questions.

Generative AI in CRM Applications (17%)

Sales AI

Sales AI unique as it is built on 4 guiding principles: Trust, Relevance, Security and Ecosystem

Here are key features:

• Prepare with Einstein Copilot: for meetings by researching prospects, summarizing records, and delivering recommendations for planning your day. It provides selling guidance and close plans, auto-updates CRM data, and more.

• Create AI-powered generative emails: auto-create personalized and impactful emails at scale for prospecting, following up on meetings, and sending proposals. 
• Leverage AI-powered call summaries and call support: Your reps save time, improve follow-up, and make every interaction count with AI-powered insights, summaries, automated follow-ups, and sales signals. Sellers are provided with key mentions, transcripts, and action items automatically from recorded calls. 
• Build data foundation: Use automatic activity capture, opportunity and lead scoring, predictive forecasting, account insights and deal insights to make data-driven decisions.
• Coach Sellers:  Provide instant and actionable feedback through immersive simulations and personalized coaching programs.
• Guide selling and grow relationships: Use propensity to buy, next-best opportunity, and product recommendations alongside relationship maps, recommended connections, and email insights to sell the right solution to the right person at the right time.

I have prepared this below picture for better understanding.

Sales AI Capabilities (prepared by Santanu Boral)

For exam, pay attention to how to prepare sales emails from "Draft with Einstein", Call Explorer, Einstein Conversational Insights, Activity Capture, How to prioritize opportunities with the help of Einstein.

Service AI

Here are key features:

For exam, pay attention to following at-least:

• Einstein Case Classification: Predicts field values like Priority, Reason, or Type for classifying incoming cases based on the text a customer presents in the case Subject and Description.
• Einstein Work Summaries: Drafts a summary, issue, and resolution using generative AI, based on a Chat conversation between an agent and customer.
• Einstein Case Wrap up: Lets chat agents complete cases fast, with greater accuracy and consistency.

Check, what is the difference between case wrap up and work summaries.

• Einstein Conversation Mining: Transforms conversation data into service insights and builds bot intents.
• Einstein Service Replies: Drafts and recommends fluent, courteous, and relevant replies using generative AI for your agents to review, edit, and post.
• Einstein Reply Recommendations: Analyzes chat transcripts to recommend relevant replies during chat and messaging sessions.

Prompt Builder (37%)

There are almost 22 questions covering this area.

Prompt are the detailed instructions that helps LLMs generate great output.

Prompt Definitions

Lets take a look how prompt looks like:

Best Practices for designing effective prompt:

• Be clear and concise
• Include context and directives
• Consider output
• Provide Example response(s)
• Break up complex tasks
• Experiment and Refine

Prompt Template Types:

 There are questions on each template types and scenarios on when to use it.

Read Prompt Builder Basics on each prompt template types.

Here is example of prompt template with secure data retrieval and dynamic grounding.

There were questions on the followings:

• Standard Prompt templates
• Ground templates with Salesforce Resources:
• Record merge
• Flow merge
• Apex merge
• Related list field merge
• Prompt Builder Limits
• Prompt Builder Limitations
• Scenarios of creating Template-Triggered Prompt Flow.
• Save & Preview of template
• If the field values are not showing then how to troubleshoot.
• Saving and versioning of prompt templates.
• Enabling prompt builder for users

Einstein Copilot (23%)

Einstein Copilot is trusted conversational AI assistant for CRM that’s seamlessly built into the Salesforce interface. It handles questions and requests posed in natural language and provides relevant answers drawn from secure, proprietary company data.

Components of Copilot:

This is most important

Copilot Standard Actions

• Identity Record by Name
• Identify Object by Name
• Query Records
• Summarize Record
• Draft or Revise Sales Email
• Answer Questions with Knowledge

Copilot Custom Actions

Copilot Testing

• Think like an user
• Use dummy data
• Test on Salesforce mobile

There were questions on followings:

• Purpose of Enhanced Event Logs
• Extending Copilot with Custom Actions
• Standard Copilot Action References

Model Builder (8%)

Einstein Studio

• Use the model builder to create a new AI model with clicks, not code
• Connect to an existing AI model in an external platform such as AWS SageMaker, Google Vertex AI, or Databricks.
• Connect to an existing LLM (Large Language Model) from third parties like OpenAI or Azure

Model Type

• Created with Einstein models built from scratch in Einstein Studio.
• Connected models bring in outputs from a model hosted somewhere else such as AWS SageMaker, Vertex AI, or Databricks.
• Salesforce-Enabled models have been set up for you

Model Capability

• Regression models predict a number such as currency amount, count, or likelihood percentage
• Binary Classification model predict an outcome group with two options such as true/false, yes/no, or won/lost.

Also, check the scenario where BYOM should be proposed.

That's all. Hope it will help you to prepare and pass this exam.

Thanks for reading!

Note: The pictures and concepts have taken from Salesforce materials.

Further Reading

List of my blog posts on Salesforce Certifications