This is the third blog in the series “Design Your First Salesforce AI Agent with Agentforce : From Sales Persona to Automation“.
💡Why This Matters
Not every task is worth automating — and not every high-value workflow is easy to build. This step helps you separate the noise from the opportunity. By scoring each job by friction, impact, and effort, you create a focused, realistic roadmap that delivers value fast and scales over time.
🔑 What You’ll Learn
How to evaluate jobs using a 3-part scoring model: Friction, Impact, and Effort
How to identify Quick Wins vs. Strategic Priorities
How to balance value, complexity, and feasibility in your build plan
How to set up a prioritization matrix to guide your roadmap
With a clear view of the rep’s day-to-day workflows, the next step is to evaluate which jobs are most valuable and realistic to automate — and in what order.
This step is critical. It’s where discovery turns into direction. By scoring each task through a structured lens, you lay the groundwork for building a focused, outcome-driven AI agent roadmap — one tailored to the real-world needs of your sales persona.
We evaluate each job using three dimensions:
- Friction – How repetitive, manual, or mentally taxing is the task today?
- Impact – How much does this task drive core business outcomes like revenue, pipeline velocity, or customer retention?
- Effort/Complexity – What’s the technical lift? How many systems are involved? How much process or behavior change is required?
This model surfaces the tasks that are both painful and important — and lets you distinguish between:
- Quick Wins – High-impact, low-effort opportunities that drive fast value
- Strategic Priorities – High-value, higher-complexity use cases that require more planning and integration
You’ll use this prioritization to design your Agentforce roadmap in a way that’s modular, scalable, and adoption-ready — starting small but solving real problems from day one.
📝 Note on Scoring and Use Case Prioritization
The friction, impact, and effort/complexity scores provided here vary from customer to customer depending on:
Existing Salesforce configuration and data availability
Maturity of underlying processes (e.g., meeting follow-up, task creation)
Available integrations or automation already in place
A use case that appears high-effort in this example may be straightforward for a customer who has already invested in related infrastructure. Similarly, what’s considered high-friction for one team may be well-optimized in another.
Many of the lower-effort use cases shown in Phase 1 — such as meeting prep, follow-up task creation, or logging notes — can serve as foundational components for future capabilities. These early wins often help improve data quality, build rep adoption, and create a strong base for expanding automation in later phases.
👉 For the purposes of this blog, we will use the scores below. They are based on a representative sales persona and a typical Salesforce setup, reflecting how a modern sales team works day to day. Each activity is scored across four dimensions: friction, impact, business value, and effort/complexity.
- Friction reflects how manual the process is and whether it requires navigating multiple systems.
- Impact measures the productivity gain from automating the task.
- Business value is calculated by multiplying friction and impact to highlight which activities deliver the most return when automated.
- Effort/complexity indicates the time and technical lift required to automate the task in Salesforce.
This framework provides a consistent way to compare use cases and see where Agentforce agents can deliver quick wins versus where they represent longer-term, high-value investments.
Build Agent road map
💡Why This Matters
Trying to build everything at once is a fast track to failure. A phased roadmap gives you a clear, achievable path — one that delivers value quickly while setting up long-term success. By sequencing use cases by complexity and business value, you create momentum, reduce risk, and build trust with users.
🔑 What You’ll Learn
How to turn your use case scoring into a prioritized build plan
How to separate quick wins from high-effort strategic projects
What use cases to defer — and why they’re better in later phases
How to structure your roadmap into 3 phases: MVP, expansion, and intelligence
With each sales activity scored for friction, impact, and implementation effort, you now have the inputs to create a clear, phased roadmap for your first AI agent.
This roadmap helps you:
✅ Start with high-impact, low-effort tasks that deliver fast value
🟡 Identify mid-stage use cases that may require moderate integration
🔴 Plan for complex, high-value automation that spans multiple systems
Each Job to Be Done becomes a candidate for automation.
By sequencing use cases based on business value and technical complexity, you can roll out AI capabilities incrementally — minimizing risk, aligning with rep workflows, and delivering visible value early.
✅ This incremental approach also helps you unlock higher-complexity use cases over time.
Phase 1 automation — like improving data quality, enriching CRM records, and standardizing follow-up — creates the foundation needed for more advanced Phase 2 capabilities such as deal risk detection, upsell recommendations, or AI-driven pipeline planning. Each step builds system readiness, user trust, and the operational scaffolding needed to handle more powerful automation later.
🧩 What Data Will the Agent Need?
For each use case, assess the data required to reason and respond. In the examples provided, only standard CRM objects like Events, Opportunities, and Accounts are needed.
But your projects may require more:
- External data sources
- Unstructured content (emails, call notes, documents, knowledge articles, policy and procedure documents)
- Real-time or batch inputs from other systems
🧠 Tools to Help:
🧩 Salesforce Data Cloud
- Connects to external systems with zero-copy access
- Ingests both structured and unstructured data
- Creates vector search indexes with no-code retrievers, enabling semantic lookup across notes, emails, and documents
🛠️ Prompt Builder
- A no-code tool to instruct the LLM how to respond
- Uses CRM records, related lists, data graphs, and retrievers to generate grounded, reliable answers
- Perfect for crafting guided, on-brand agent responses without writing custom code or APIs
🤖 Agent Builder
- A no-code platform to define how your LLM-based agent should behave
- Leverages a powerful reasoning engine that processes natural language instructions, structured prompts, and dynamic context
- Enables you to design full agent workflows with conditional logic, tool usage, and step-by-step task execution
These tools enable smarter, more context-aware agents without requiring complex development up front.
⚙️ What Actions Must the Agent Perform?
Next, identify what actions the agent needs to take — and whether they are native to Salesforce or require external calls:
- If native, use Flows, Apex, prompts (help) or ADL(Agentforce Data Library)
- If external, options include:
- Agent-to-Agent (A2A) calls
- Model Context Protocol(MCP)
- MuleSoft for integration across systems
MuleSoft is especially useful for connecting to external services, enabling agents to trigger complex workflows or access data not stored in Salesforce.
🛠️ Dream big … start simple
Start by grounding your first agent experiences in native Salesforce data and actions, then gradually expand to external data sources and integrations as your use cases evolve.
This staged approach ensures fast wins, reduces risk, and sets the foundation for scalable AI automation.
🔒 What to Avoid in First Build
While high-value, some use cases may not be ideal to start with due to system complexity or readiness. For the first build, the goal is to get an agent up and running quickly to drive user engagement. These include:
- Machine learning–driven lead scoring
- Automated upsell detection from product usage
- Cross-system close plan generation
- Real-time risk scoring across the book of business
These are better suited for later phases once you’ve validated the agent pattern and have the right integrations in place.
🔽 Phase 1: Quick Wins (High Value, Low Effort)
Typical Timeline: 0-4 weeks
This phase focuses on high-value tasks that are simple to implement — ideal for proving early value and driving adoption. These use cases work using basic automation patterns. They’re designed for fast deployment, require minimal configuration, and deliver immediate productivity gains. Most importantly, they help reps see the benefit of AI in their day-to-day work from day one.
📅 Managing Client Engagement
🏗️ Managing Opportunities
⚙️ Phase 2: Medium Complexity, High Value
Typical Timeline: 4-8 weeks
These jobs deliver strong business impact but require deeper systems access, multi-source data stitching, or more structured agent logic. They’re best tackled after Phase 1, once foundational agent behaviors are established. In this phase, agents begin to handle more complex workflows using document intelligence, knowledge retrieval, and cross-system integration — enabling smarter, end-to-end task support.
🏗️ Managing Opportunities
📅 Managing Client Engagement
🔧 Phase 3: Strategic Builds (High Value, High Effort)
Typical Timeline: 8+ weeks
This phase focuses on high-impact use cases that involve advanced AI capabilities, cross-system coordination, and deeper integration into strategic business planning. By this point, your agent evolves from a task assistant to a strategic partner, supporting decision-making across pipeline health, territory coverage, and revenue risk. These jobs leverage predictive modeling, real-time prioritization, and strategic decision support — helping sales leaders and reps act earlier, faster, and smarter.
📒 Managing the Book of Business
🏗️ Managing Opportunities
📅 Managing Client Engagement
📊 Roadmap Summary
In the next step, we’ll take this prioritized list and determine the scope of our first agent. Checkout the next blog in the series here.
Written by
