AI that knows: using RAG to work smarter

At the Pieces AI Productivity Summit, Salesforce Solutions Engineer Spencer Gallardo shared his journey with AI in the enterprise, highlighting how Retrieval Augmented Generation (RAG) has transformed his work. 

Speaking from years of experience, Gallardo offered a first-hand look at how a high-growth environment like Salesforce integrates RAG into daily operations to drive productivity and reduce friction.

From manual context switching to AI acceleration

Spencer opened by describing the challenges he faced in his day-to-day workflow: managing multiple client calls, building custom demo environments, digging through fragmented documentation, and juggling open browser tabs. These distractions often derailed productivity. 

Searching for the right information or recreating resources from scratch was time-consuming and mentally draining.

The introduction of large language models (LLMs) like GPT was a turning point. At first, they helped with general coding and research tasks. 

But Spencer needed more than just autocomplete capabilities – he needed answers grounded in Salesforce-specific knowledge, something broader models couldn’t deliver on their own.

That’s when RAG became a core part of his toolkit.

What is RAG and why it matters

RAG, or Retrieval Augmented Generation, is an AI architecture that combines pre-trained models with external, domain-specific data. 

It retrieves content from trusted sources, structured or unstructured, and integrates it into the LLM’s response. 

This ensures outputs are tailored and relevant to a company’s internal systems, documentation, and customer scenarios.

For Salesforce, this means combining generative AI with internal data like Confluence pages, Slack conversations, PDFs, and knowledge articles. 

The result is faster access to relevant information with better precision.

Spencer explained how asking internal questions like “How do I activate Agent Force for a customer in the healthcare industry?” used to take hours. 

Now, with RAG, responses are immediate and accurate because they’re grounded in Salesforce’s proprietary data.

Real-world use cases inside Salesforce

Spencer detailed several use cases where RAG and internal AI agents have improved workflows. 

Salesforce’s internal AI, Agent Force, is configured to surface insights from a wide array of internal data streams. 

It can answer complex questions, trigger internal automations, and retrieve exact documentation or links – all from a natural language query.

He also outlined how this system enables support for enterprise-scale products, which often evolve quickly and span multiple teams.

 With RAG, the AI can handle customer-specific questions about product features, compliance requirements, and use-case recommendations with depth and accuracy.

The system is designed to understand a user’s intent, retrieve vectorized knowledge, and deliver a grounded response. 

Spencer highlighted this with a demo showing how a simple query about resetting an Okta password could trigger a guided workflow, leveraging internal documentation and issuing a direct action – all within seconds.

From internal tools to personal projects

Spencer emphasized that RAG’s value extends beyond the enterprise. 

He uses it for personal projects, including health tracking and YouTube content development. By connecting his personal data, like meal logs and content ideas, to a cloud backend and AI assistant, he’s able to generate weekly meal plans and analyze trends without manual tracking.

Tools like Notebook LM, a Google-powered UI for RAG, have been especially helpful. 

With no advanced coding required, users can upload documents, URLs, or personal notes and get AI-powered answers grounded in those inputs.

 He demonstrated how even simple use cases, like installing a software product, benefit from personalized RAG-driven responses.

Enabling teams and customers alike

In his role, Spencer also supports enterprise clients building similar AI-enabled solutions. 

He mentioned companies using RAG to enhance IT support workflows – automating password resets or surfacing relevant helpdesk content as well as customer-facing implementations for industries like sports and entertainment.

These organizations are using RAG to improve self-service experiences, enabling customers to ask complex questions and receive contextual answers without human intervention. 

It reduces response time and increases confidence in AI as a reliable support layer.

Building with RAG is a practical path forward

Spencer closed with a few steps for teams or individuals looking to adopt RAG:

• Start with a clear use case – FAQs, note syncing, or support queries are good starting points.

• Identify where knowledge lives – whether in Google Docs, Notion, Slack, or other platforms.

• Use available frameworks and UIs – like Notebook LM or your own agent architecture.

The emphasis was on starting small, validating results, and scaling from early wins. For Gallardo, what once took hours now takes minutes. 

Tasks like demo prep, previously a two-hour effort, can now be completed in under 15 minutes with the help of RAG.

The bottom line

Spencer’s presentation was a practical walkthrough of how RAG is more than a buzzword, it’s an enabler for productivity, both in professional environments and personal projects. 

By combining enterprise data with generative models, Salesforce has streamlined complex tasks, improved internal collaboration, and empowered employees to work smarter.

As AI continues to integrate into daily workflows, Spencer’s talk served as both a blueprint and a motivation for others looking to harness the power of contextual AI.