AI Agents

💬 In plain words: Think of an AI Agent like a smart, reliable team member rather than a basic search bar. If you tell a sharp employee, "Prepare for our Q4 customer account review and handle the follow-up tasks," you do not have to micromanage every single mouse click. They understand the big goal, check what data is missing, query your CRM, read the results, figure out the next logical step, pause to ask for your approval before making risky changes, and finish the job. That iterative cycle— Goal → Reason → Act → Observe → Decide → Repeat —is the heart of an Agentic Workflow . ⚡ 1-Minute Summary: Goal-Driven Architecture: An AI Agent is built to pursue a business goal using dynamic reasoning, live context, external Tools , and step-by-step decision-making. The Agentic Loop: Whether you are working with Anthropic's Claude via Model Context Protocol (MCP) or Salesforce Agentforce (powered by the Atlas Reasoning Engine ), the...

💬 In Plain Words: The 1-Minute Summary Tool calling in the Model Context Protocol (MCP) is not a one-off button press—it is a complete 6-step conversation loop. Think of an AI model like Claude or a Salesforce Agentforce agent as a smart new hire sitting at a desk. The AI cannot magically reach into your database on its own. Instead, it looks at a menu of approved software tools, picks the right one, fills out a request form, and asks the host application to run it. Once the system hands back the data, the AI reads the result and decides what to do next.

⚡ 1-Minute Summary (Key Points) The Model Context Protocol (MCP) is an open standard that lets AI models like Anthropic's Claude and Salesforce Agentforce safely plug into external business databases and apps. Every MCP server is built on three foundational building blocks (called primitives): Tools , Resources , and Prompts . Tools are executable actions controlled by the AI model (e.g., creating a Salesforce Lead or running a live SOQL search). Resources are read-only files and context controlled by the application (e.g., reading an Account profile or PDF contract). Prompts are pre-built instruction templates controlled by the user (e.g., a slash command that formats an executive sales briefing). The Golden Memory Trick: Tool = DO (Model-Controlled), Resource = KNOW (App-Controlled), Prompt = GUIDE (User-Controlled). With Salesforce's native MCP support in Agentforce and MuleSoft, enterprise architects...

💬 In plain words: Think of connecting AI to Salesforce like running a high-security bank. Claude is your smart financial advisor sitting in the lobby who knows how to solve customer problems. MCP (Model Context Protocol) is the pneumatic tube system and armored courier service between the lobby and the vault. Salesforce is the main bank vault holding all your customer accounts, rules, and security locks. When you ask a business question, Claude doesn't walk into the vault itself. Instead, it writes down a request, sends it through the MCP tube, waits for Salesforce to verify your ID badge and send back the exact folder you are allowed to see, and then reads that folder to give you a clear, helpful answer.

💬 In plain words: A variable securely stores a specific fact for the rest of an AI conversation. Without variables, you are blindly trusting the AI model to remember things accurately. By using variables, you lock that fact in place so it never drifts, hallucinates, or disappears.

💬 In plain words: Because Large Language Models (LLMs) generate probabilistic output, your test suite scores will bounce around slightly even if you haven't touched a single line of code. To find your noise floor , run your exact same test suite three times without making any changes. The mathematical spread between those three scores is your normal variance. If a score change happens inside that spread, it means nothing. ⚡ Key Points AI test scores vary naturally; a drop in your score is not always a bug. Without measuring your baseline variance (the noise floor), your test scores are uninterpretable. Your Continuous Integration (CI) gate threshold must be set just outside the observed spread. Always re-measure your noise floor when you add a significant chunk of new test cases to your suite.

💬 In plain words: Testing Center is your dedicated AI validation hub. Located directly inside Agentforce Studio (right next to Agent Builder and Observability), it replaces manual trial-and-error testing. It can automatically generate test cases, simulate full multi-turn conversations using different user personas, and mathematically score the results against rules you define. ⚡ Key Points Location: Testing Center is a native, dedicated tab in Agentforce Studio, completely removed from traditional Salesforce Setup. Expected vs. Actual: Every test case requires you to define an expected subagent and an expected action, allowing for direct comparison. Scale with AI: Instead of writing hundreds of tests by hand, use AI-generated cases to instantly build synthetic interactions. Custom Evaluations: Score your agents using strict deterministic rules (Strings/Numbers) or an LLM-as-a-Judge. Cost Awareness: Running these automated test suites c...

⚡ 1-Minute Summary Agent testing is statistical, not pass/fail. Never assert on exact wording. Assert on which subagent handled the request, which action ran, the decided value, and whether it correctly refused. Measure your noise floor : run the same suite three times unchanged and record the variance.

💬 In Plain Words: Imagine if every AI app needed its own custom connector for Salesforce, GitHub, your databases, and your files. That's a lot of code to build and maintain. The Model Context Protocol (MCP) fixes this. It's an open standard that works like a "USB-C port for AI." Build one MCP connector for a system, and any MCP-compatible AI app can discover and use its data and tools.

⚡ 1-Minute Summary In Salesforce Agentforce, agents are divided into two main categories based on who interacts with them: Employee Agents and Service Agents . Employee agents assist your internal staff natively within Lightning , Slack , and the Salesforce mobile app. Service agents interact directly with your customers across web chat, Experience Cloud sites, messaging platforms (like WhatsApp and SMS), and voice. The magic of this architecture? The exact same Agent Script powers both. You build the logic, topics, and actions once, and deploy them across any surface. When you switch to a channel like voice, the physics change (latency, spoken numbers, short conversational turns), but the underlying business logic remains untouched.

⚡ Key Points: 1-Minute Summary Assistants vs. Agents: AI Assistants primarily answer questions and provide information. AI Agents understand goals, plan out steps, use tools, and take autonomous action. The Core Engine: True AI Agents combine contextual reasoning, decision-making logic, and external tool integration to complete multi-step tasks. Workflow Automation: Agents shift the paradigm from requiring users to drive every step manually to having the AI manage and execute the workflow. Claudeforce Vision: By merging Anthropic's Claude models with Salesforce's data layer (often using the Model Context Protocol or MCP), enterprises can build intelligent, autonomous business workers. 💬 In plain words: An AI Assistant gives you answers when you ask a question. An AI Agent receives a high-level goal and actually works to complete it. For example, you ask an Assistant, "Which customers need follow-up?" and it gives you a l...

💬 In Plain Words: Think of Claude as a very smart employee sitting in an empty room. It understands problems and can suggest great solutions, but it can't look anything up or change anything. Tools give that employee a computer. With tools, Claude can check your Salesforce data, read customer history, create tasks, and take other approved actions. Tool use turns an AI from a chatbot that answers into an agent that acts .

💬 In Plain Words: When building enterprise AI agents, you must decide in advance which parts of a conversation the Large Language Model (LLM) is allowed to influence, and which parts are strictly off-limits. Remember this golden rule: An instruction is just a request; a filter is an absolute control. When money, compliance, or eligibility are on the line, you cannot rely on polite instructions in a prompt. You must use hardcoded determinism mechanisms to force certainty.

💬 In plain words: There are two ways an AI agent can hand off work to another subagent, and they behave completely differently. A transition is a one-way trip—the first agent says goodbye. A delegation is a round trip—the first agent asks a question, gets an answer, and comes back. If you choose the wrong one without understanding the difference, your AI will lose the context of the conversation.

💬 In plain words: When building AI agents in Salesforce Agentforce, before_reasoning and after_reasoning are two optional logic blocks inside a subagent. One runs exactly before the AI thinks, and one runs exactly after. They contain hard logic (code), not prompt suggestions. This is where your absolute, unbreakable rules must live. 📌 Example: A utility company verification agent checks if a caller actually owns the meter in the before_reasoning block. Later, it escalates any refund request above $100 inside the after_reasoning block. Neither of these critical business decisions is left up to the AI language model.

💬 In plain words: An Agent Script is a collection of building blocks, where each block contains either specific data or procedural logic. There are exactly eight types of blocks available, but as a developer, you will spend the vast majority of your time working inside just four of them. 📌 Example: Let's say you are building an internal employee assistant for a company called Meridian. This agent requires a config block with agent_type explicitly set to AgentforceEmployeeAgent . It also needs a variables block to store session data like the employee's meter ID and outage status, five separate subagent blocks to handle specific tasks, and a start_agent block to act as the traffic cop.

In plain words: There are two action lists inside a subagent in Agentforce, and they serve completely different purposes. One list runs strictly because your logic explicitly called it. The other acts as a menu of tools that the AI model can choose to use if it thinks they are needed. Confusing these two is the most common mistake developers make when writing Agent Script. 📌 Example: Imagine a utility company's agent (like Meridian). It runs an "outage lookup" deterministically every single time by placing it in the actions block. However, it offers a "book a crew visit" tool inside reasoning.actions , so the model only offers to schedule a crew when the context of the conversation actually requires it.