Salesforce Agentforce Pricing: Digital Wallet & Credit Consumption Explained

💬 In plain words: Agentforce usage isn't free; it is strictly metered. Every conversation, background action, and document indexing job consumes credits. The Digital Wallet is where Salesforce tracks this consumption. As an architect or developer, you must establish a baseline of your credit balance before you build anything, so you can accurately predict your running costs.

📌 Real-Life Example: A team at Meridian decided to baseline their credit usage before starting an Agentforce project. After completing their initial document indexing and running their testing suite three times, they checked the Digital Wallet again. Because they had their starting and ending numbers, they were able to present leadership with an exact, data-driven "cost per conversation"—instead of relying on rough guesses that could blow up the IT budget.

🔑 Key Points: How Agentforce Consumption Works

To design scalable and cost-effective AI agents, you need to understand exactly what burns credits in the Salesforce ecosystem. Cost per conversation is a very real design constraint.

• The Digital Wallet: Accessible directly from the App Launcher, this dashboard gives you real-time visibility into your AI credit consumption.
• The Heaviest Consumers: Your highest credit burn rates will come from one-off massive tasks (like indexing thousands of Knowledge documents) and repetitive heavy tasks (like running your Agentforce Testing Center suites, voice conversations, and extended reasoning chains).
• Testing Center Costs: Test runs use Einstein Requests, which decrement your Digital Wallet balance. If you run a suite of 22 test cases 3 times, you just paid for 66 full AI sessions. Size your test suites deliberately!
• Internal vs. External Pricing: Employee-facing agents might actually be unmetered if the interacting Salesforce user holds the specific "Unmetered User-Based AI" permission set. Always verify your licensing model.
• Efficiency Equals Savings: Shorter, tighter reasoning chains are not only faster and more accurate—they are significantly cheaper. Every extra "turn" the LLM takes costs you credits.

Salesforce Digital Wallet (App Launcher → Consumption)
 
 Heaviest AI Credit Consumers:
 ├─ Indexing Documents        → (One-off but massive payload)
 ├─ Test Suite Runs           → (e.g., 22 cases x 3 runs = 66 paid sessions)
 ├─ Voice Conversations       → (Consumes credits much faster than text)
 └─ Long Reasoning Chains     → (Every extra LLM "turn" costs you)

🧠 Core Takeaway: Baseline before you build. You cannot accurately measure or optimize what you never recorded. Write the balance down on day zero.

🧭 The 360 Card Summary

Rule: Record your Digital Wallet credit balance before your very first build step, and again after each major testing phase.

Gain: You earn massive credibility by being able to accurately quote the "cost per conversation" to stakeholders—something almost nobody else can do early in a project.

Price: It requires two minutes of admin work per phase and the discipline to consistently log it.

Limits: Indexing Data Cloud documents and running automated test suites are exceptionally expensive operations. Architect your testing strategy so you aren't burning credits needlessly.

At Volume: Long reasoning chains multiply their costs across every single conversation. Tighter, shorter system instructions are cheaper and yield higher accuracy.

⚠ INTERVIEW TRAP: Do not say you will "check the costs after go-live." By the time the AI agent is in production, the bill will arrive before your understanding does. Great architects treat AI credits as a hard design constraint during the development phase.

💬 Core Q&A & Interview Prep

Q: What does it actually cost to run an Agentforce AI agent?

🎯 Say this first: "It costs credits per conversation, action, and indexing run. I track these in the Digital Wallet against a strict baseline I record before writing a single line of code."

I treat AI cost as a technical design constraint, not just a finance team question. Usage is strictly metered and visible in the Digital Wallet via the App Launcher. The heaviest credit drains are indexing documents, repeated automated test suite runs, voice-based conversations, and unnecessarily long reasoning chains. Because Testing Center runs consume Einstein Requests, a moderately sized test suite run a few times can quickly drain credits.

By recording the starting balance before building and checking it after testing phases, I can accurately calculate the cost per conversation. I also ensure I understand the licensing context—for example, knowing that employee-facing agents might be unmetered if the user has the right permission sets.

🔗 Connecting the Dots: Follow-Up Questions

Q1: How do long reasoning chains directly impact Agentforce pricing?

Large Language Models (LLMs) bill based on tokens processed per "turn." If your Agent's instructions are vague, the AI will require multiple internal reasoning steps (turns) to figure out which Action to call and how to format the response. Every extra turn consumes more tokens, draining your Digital Wallet faster. Writing shorter, highly specific instructions reduces the turns required, making your agent both cheaper and much more accurate.

Q2: What does an Agentforce "trace" give you that a standard reply does not?

A trace tells you why an event happened. While a reply just shows the final text sent to the user, a trace exposes the inner workings: the specific routing decisions, the Apex actions called, the payloads returned, and the exact reasoning branch the AI took to formulate its answer.

Q3: Why is "Deflection Rate" alone a highly misleading success metric?

Because an angry, frustrated customer who gives up and abandons the chat counts as "deflected" just like a happy customer who got their problem resolved. Always pair deflection metrics with CSAT (Customer Satisfaction) and escalation quality to get the true picture of your agent's performance.

Q4: Where exactly do you go to check this credit consumption?

You check it in the Digital Wallet, which can be found simply by searching for it in the Salesforce App Launcher.