Runtime and Salesforce Agentforce
Runtime vs Salesforce Agentforce for fintech and payment teams: what each does, where they overlap, and how Runtime agents investigate cases across your ledger and write back to Salesforce.
TL;DR: Agentforce is the natural choice for agents that live inside Salesforce: service, sales, and CRM-centric financial services work. Runtime is built for the payment operations work that happens outside the CRM, in your ledger, processor, and bank portals, and it reads and writes Salesforce as one of many systems. Many fintechs will run both.
| Feature | ||
|---|---|---|
| What it is | Agent harness for payment operations | AI agent platform inside Salesforce |
| Home system | Your ledger, processor, bank files, tickets | Salesforce CRM and Data 360 |
| Core use cases | Exceptions, recon, risk, disputes, bank RFIs | Service, sales, contact center |
| Financial services offering | Payments and fintech only | Agentforce for Financial Services |
| Where agents run | Isolated computers in your cloud | Salesforce platform |
| Models | Any model, including open-weight in your cloud | Salesforce-managed, plus bring your own LLM |
| Portals with no API | Built-in browser on the agent computer | Via MuleSoft and APIs |
| Approvals on money movement | Built in, in Slack, Teams, or dashboard | Configured in flows and actions |
| Pricing | Free, Teams from $99/seat/month | Flex Credits or per-user add-ons |
Runtime and Salesforce Agentforce both get called "AI agent platforms," so they end up on the same shortlist. At most fintechs they do different jobs. Salesforce is often the system of record for merchants, accounts, and cases, and Agentforce puts agents on top of that record. Runtime agents work the payment operations behind the case, across systems Salesforce doesn't hold, and treat Salesforce as one of the tools they read and write.
What each does
Salesforce Agentforce is Salesforce's platform for building and running AI agents on top of its CRM. Its Atlas Reasoning Engine plans and takes actions, Agent Builder and Agent Script let admins configure agents with Flows, prompts, Apex, and MuleSoft API connectors, and Data 360 grounds agents in CRM and external data. Salesforce leads with service agents that resolve cases, sales development agents that engage prospects, employee agents for HR and IT, and contact center voice. For banks, wealth managers, and insurers it offers Agentforce for Financial Services, with prebuilt templates for client onboarding, servicing, underwriting, and claims.
Runtime is the AI agent harness for payment and fintech teams. Anyone on payment ops, risk, compliance, finance, underwriting, onboarding, or support builds agents from their SOPs. Each agent works on its own isolated computer in your cloud, reaches tools through APIs, databases, CLIs, and MCP servers, and uses a browser for processor, network, and bank portals with no API. Agents start read-only, stop for approval before anything moves money, and record every run end to end.
Who each is for. Agentforce fits teams whose work and data already live in Salesforce. Runtime fits payment teams whose hardest work happens in the ledger, the processor, and the bank file, and who need that work governed for a sponsor bank.
Where they overlap
Be honest about it: there is real overlap.
- Customer service. Agentforce service agents answer customers and resolve cases. Runtime support agents also work "where's my payout" tickets. If your support work is mostly answering from knowledge and updating CRM records, Agentforce may cover it.
- Financial services workflows. Agentforce for Financial Services covers onboarding and servicing for banks and insurers. Runtime also handles onboarding and underwriting reviews, aimed at merchants, payfacs, and program managers rather than wealth and insurance clients.
- Bring your own model. Salesforce lets you connect your own LLM providers through Model Builder, routed through its Einstein Trust Layer. Runtime goes further on where the model runs (open-weight models served in your own cloud), but both give you a choice.
- Building agents from procedures. Both let operators describe the work and configure agents without writing a full application.
The difference is the center of gravity. Agentforce is built around the Salesforce record. Runtime is built around the transaction, which touches the ledger, processor, bank, tickets, and the CRM.
How they work together
A common fintech setup: Salesforce holds the merchant and the case, Agentforce handles first-line service, and Runtime agents do the investigation and back-office work.
- A case opens in Salesforce. A merchant writes in that a payout never arrived. Agentforce, or your support team, logs the case and answers the customer.
- A Runtime agent picks it up. It is triggered from the case, Slack, or the dashboard, and starts read-only on its own computer in your cloud.
- It investigates across systems. The agent queries your ledger and processor, checks the bank file and returns, reviews prior tickets and the merchant's risk history, and logs into a portal with no API if it has to.
- It drafts a decision with evidence. For example: the payout was held by a reserve rule, the hold can be released, here are the queries and records that show it.
- A person approves in Slack or Teams. Releasing the payout, applying a reserve, or booking a ledger entry waits for a named approver.
- The agent writes back and closes the loop. It updates the Salesforce case with the finding and resolution, notifies support, and the full record (trigger, every query and tool call, approval, cost, result) stays with you, ready to export for a sponsor bank or examiner.
The same stuck payment usually touches four teams: support gets the ticket, payment ops traces it, finance sees a recon break, risk may review it. In this setup it's one investigation with one record, and Salesforce still shows the customer-facing story.
When you need both
- Salesforce is your system of record for merchants or accounts, and you want Agentforce on service or sales
- The work behind the case lives in your ledger, processor, bank files, and data warehouse
- Card data or PII must stay in your cloud, or you want to run open-weight models there
- You need approvals before money moves and a run-level audit trail your sponsor bank will accept
- Risk, finance, payment ops, and compliance need agents too, not just teams that work in Salesforce
When Agentforce alone is enough
- Most of the work, and the data it needs, lives inside Salesforce
- Your use cases are service, sales, and contact center conversations
- You're a bank, wealth manager, or insurer and the Agentforce for Financial Services templates match your workflows
- Your Salesforce admins will build and run agents, and you prefer one vendor for CRM and agents
Pricing
Agentforce publishes its pricing. Flex Credits cost $500 per 100,000 credits, with a standard action using 20 credits and a voice action 30. Customer-facing agents can be priced at $2 per conversation. Per-user add-ons are $125 per user per month for Sales, Service, and Field Service ($150 for Industries clouds), an Agentforce user license is $5 per user per month plus Flex Credits, and Agentforce editions start at $550 per user per month with credits included. Salesforce Foundations includes Agentforce Builder at no extra cost.
Runtime has a Free tier ($0, one session), Teams from $99 per seat per month, and custom Enterprise pricing that includes self-hosting. Rain replaced $250k in vendor spend with Runtime.
See Runtime on the work behind your Salesforce cases
Bring one SOP. A forward-deployed AI engineer builds the first agent with your team, inside your cloud, writing back to Salesforce.
Frequently asked questions
Does Runtime replace Salesforce Agentforce?
Usually not. Agentforce is strongest for agents that work inside Salesforce, such as service and sales agents. Runtime agents handle payment operations work across your ledger, processor, bank portals, and data warehouse, and they read and write Salesforce cases as part of that work. Many fintechs run both.
Can Runtime agents read and update Salesforce cases?
Yes. Runtime agents reach Salesforce through its APIs or an MCP server, alongside your processor, ledger, Zendesk, Jira, Snowflake, and other tools. An agent can pull a case, investigate it across your systems, and write its findings and resolution back to the record after a person approves.
Is there an Agentforce product for financial services?
Yes. Salesforce offers Agentforce for Financial Services for banking, wealth, and insurance, with prebuilt agent templates for work such as client onboarding, servicing, underwriting, and claims. It is centered on the client relationship held in Salesforce.
How much does Agentforce cost?
Salesforce publishes several options: Flex Credits at $500 per 100,000 credits, with a standard action using 20 credits; $2 per conversation for customer-facing agents; per-user add-ons from $125 per user per month; and Agentforce editions from $550 per user per month.
Can card data stay out of the model provider with Runtime?
Yes. Runtime agent computers can run in your own AWS, GCP, or Azure account, card numbers and SSNs are stripped from prompts and logs, and you can serve open-weight models in your own cloud for PCI and PII work.
Related comparisons
Salesforce Agentforce vs Sierra
Salesforce Agentforce vs Sierra for financial services and fintech: product scope, CRM fit, channels, compliance, pricing, and where Runtime fits for payment operations behind the conversation.
Runtime vs Sierra
Runtime is one agent harness for every payment team, including support. Sierra builds conversational AI agents for customer experience. Compare scope, deployment, models, and fit.
Runtime vs Decagon
Runtime is one agent harness for every payment team, including support. Decagon builds AI agents for customer support across chat, email, and voice. Compare scope, data, and fit.
Runtime and Sift
Runtime vs Sift: Sift scores fraud and decides in real time. Runtime agents work the review queue Sift creates, investigating across ledger, processor, and tickets, then write the decision back.