Securing the Anchors: Foundations for an AI-Powered Future
Determining the optimal Large Language Model (LLM) for specific business tasks is critical in AI adoption. Options range from training custom models to utilizing pre-built solutions through APIs, adding the proper instructions, providing grounded data to prompts to enable in-context learning and supplying the information required to learn “in context”.
Generative AI comes in many forms today, such as general AI assistants like ChatGPT, specialized AI assistants like Salesforce Sales GPT and Marketing GPT, and fine-tuned open-source and pre-trained models. Furthermore, Salesforce offers a unique combination of tools, including Prompt Builder, which designs and builds prompts that dynamically interact with CRM data (e.g., sharing account or product details), and Agent Builder, which automates tasks and integrates with existing flows and external systems.
Platforms like Salesforce's Einstein can also incorporate OpenAI or Azure APIs to connect with models and generate desired responses. Users can interact with Salesforce via natural language processing (NLP), and the more grounded the data added to prompts, the better the outcomes. For instance, in industries like telecommunications, customer narratives or prompts for new order capture can be translated into low-code/no-code programming, as shown below. These can then be deployed on external websites or internal portals developed using Salesforce Experience Cloud, significantly boosting productivity.
Browse Product
Select Product
Quote/ Contract
Change/ Upgrade
Assisted/ Unassisted
Deboarding
Content/Marketing related summary
Analytics/Recommendation Extract and Summarize the Price and Rule associated with the Product
Credit check/Smart Contract Content generation/Meeting transcription, proposal generation
Auto-renewal/Change orders Document Summarization
Self Help/Advisors/Agents Knowledge creation, Article creation from Root Cause Analysis, Agentic AI response
Disconnect/ Bill/settlement FAQ (Question & Answer), Document Summarization
The traditional e-commerce journey of shopping, buying, and support can be fundamentally transformed through AI. By simulating human behavior, AI agents can autonomously browse products, make purchases and even provide support. The potential of generative AI in this domain is vast, extending to automatic configuration, code generation for order processing, and AI-powered content creation for marketing and recommendations. This convergence of AI and e-commerce heralds a new era of autonomous commerce.
Here are a few more use cases that showcase the power of AI in real-world scenarios and help build the foundation for an AI-powered future
Optimizing Retail Energy Usage with AI for Cost Savings—Suppose a consumer uses an oven to bake a cake during peak energy hours from 12 to 1 p.m.; he will incur higher costs. By analyzing data on electricity costs, appliance usage patterns, and customer preferences, AI can recommend shifting energy-intensive activities to off-peak hours. This can lead to substantial savings for consumers, such as reducing electricity bills by up to $5 by avoiding peak usage times. These straightforward use cases can be implemented early on using datasets that include time, appliance type, energy consumption rates, and usage priorities.
AI-Powered Optimization for Industrial Energy Usage-For industrial consumers, such as a pharmaceutical firm with its waste recycling plant operating during peak hours from 12 to 2 PM, AI can predict energy cost savings and carbon footprint reductions. AI recommends shifting operations to after 5 PM when energy costs are lower, and the energy source is more sustainable, such as wind turbines. This adjustment not only saves money but also reduces environmental impact.
