What a Salesforce AI agent actually does
Reads a case, lead or opportunity inside Salesforce, pulls context from CRM and Data Cloud records, and resolves it directly or escalates to a rep with full context attached.
A Salesforce AI agent sits inside your org and runs a loop: read an incoming case, lead or opportunity update, retrieve related CRM and Data Cloud context, decide which workflow applies (service resolution, lead qualification, pipeline update), call a Salesforce action, whether that's a native Agentforce action, a custom Apex method, or a Flow, to actually do something, and escalate to a human rep when confidence is low or the action crosses an approval threshold. The useful ones ground every answer in your real records, not a generic guess, and log every action for review.
Salesforce spent 2026 consolidating this category under one roof. It closed its roughly $3.6B acquisition of Fin (the company formerly known as Intercom) on September 10, 2026, with Fin reportedly nearing $100M in annual recurring revenue, and the plan is to fold Fin directly into Agentforce. That's a clear signal Salesforce wants every customer-facing agent workflow, support, sales and service, billed through one Flex Credit model instead of scattered across separate products.
Where agents still fail: cost forecasting and product-roadmap churn. Agentforce's Flex Credit pricing ($0.10 per standard action, $0.15 per voice action, $2 per customer-facing conversation) is genuinely hard to predict before you've run a real month of usage, since a workflow that quietly calls five actions per case costs five times what it looks like on paper. And Salesforce's own SDR agent was recently rebuilt into the Lead Nurturing agent with no migration path, a reminder that even the native product lineup is still being reshuffled.
What the Salesforce agent handles (and what it shouldn't)
| Task | Agent handles it? | Systems it touches | Human role |
|---|---|---|---|
| Case or ticket resolution grounded in CRM data | Fully | Service Cloud, Data Cloud | Reviews flagged low-confidence cases |
| Inbound lead qualification and meeting booking | With approval | Sales Cloud | Rep reviews before final handoff |
| Outbound sequence drafting and sending | With approval | Sales Cloud, email | Rep approves tone and targeting |
| Opportunity and pipeline data updates | Fully | Sales Cloud | Rep reviews for large deals |
| Order status and account lookups | Fully | Service Cloud, Commerce Cloud | None for routine lookups |
| Call intelligence and coaching recommendations | Fully | Sales Cloud, call recordings | Manager reviews coaching notes |
| Pricing and discount approval | No | CPQ | Rep or manager approves |
| Contract negotiation terms | No | CPQ, Sales Cloud | Rep or legal negotiates |
Off-the-shelf AI agents on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
Salesforce Agentforce (native)
Companies already on Salesforce wanting native agent capability across service, sales and data
Salesforce Foundations gives a free starting point, including Agent Builder
Flex Credits at $500 per 100,000 credits ($0.10 per standard action, $0.15 per voice action); $2 per customer-facing conversation; Agentforce User License $5/user/mo plus credits; Agentforce Coworker license $50/user/mo
Sales/Service add-ons at $125/user/mo, Industries at $150; Agentforce 1/Max edition from $550/user/mo, including 2.75M Flex Credits per org per year
Pros
- +Deepest native fit, since it runs directly on your existing Salesforce data and Data Cloud
- +Fin folding into Agentforce after Salesforce's roughly $3.6B acquisition consolidates support, sales and service agents into one product
- +Agent Builder and Agent Script let admins configure workflows without a full dev project
Cons
- −Flex Credit costs are genuinely hard to forecast until you've run a real month of usage
- −The legacy SDR agent was rebuilt as the Lead Nurturing agent with no migration path
- −Genuinely custom logic still needs Apex or Flow development on top of Agent Builder
Fin (owned by Salesforce)
Teams wanting Fin's chat, email and voice support agent specifically, ahead of its full Agentforce integration
14-day trial, no card required
$0.99 per resolved outcome; a 50-outcome monthly minimum applies when running standalone on another helpdesk
Custom outcome-volume discounts for larger teams
Pros
- +Now backed by Salesforce's resources since the acquisition closed September 10, 2026
- +Works standalone on Zendesk as well as Salesforce Service Cloud
- +Handles chat, email and voice support from one agent
Cons
- −Its outcome-based pricing is being folded into Agentforce, so its standalone future as a separate product is unclear
- −Independent teardowns report 45-53% resolution in production versus the vendor's claimed 67%
- −Overlaps functionally with Agentforce's own service agent
Qualified (Piper AI SDR)
Companies wanting dedicated, Salesforce-native website chat qualification separate from Agentforce
Demo only
Contact sales; Vendr reports a median of about $62,000/year across 185 purchases, ranging $35,000-$117,000; a Premier list price runs about $68,000/year for 25 users
Custom
Pros
- +Built Salesforce-native from the ground up, not retrofitted onto a generic chat widget
- +Strong buyer-intent and website visitor-identification data feeding qualification
- +Meeting-booking and real-time Slack alerts for reps when a target account visits
Cons
- −No public self-serve pricing, with a reported median well into five figures annually
- −Narrower scope than Agentforce, focused specifically on inbound website chat qualification
- −Overlaps with Agentforce's own lead-qualification capability if you're running both
Outreach
Sales teams wanting dedicated outbound sequencing and call intelligence synced into Salesforce
Demo only
Contact sales; third-party estimates report Amplify Core around $100, Plus around $130, Pro around $160 per user/month; Kaia call intelligence adds $30-50; implementation reportedly runs $5,000-$25,000
Custom
Pros
- +Prospecting Agent, Research Agent and Deal Agent cover outbound sequencing, account research and deal-risk flagging in one platform
- +Kaia call intelligence adds coaching on top of sales engagement
- +Native Salesforce and Dynamics integration for syncing activity back to the CRM
Cons
- −Per-user pricing stacks with Kaia's add-on fee and a real implementation cost
- −Annual contracts are standard, not month-to-month
- −Primarily a sales-engagement layer, not a replacement for Agentforce's broader service and data capabilities
The AI stack
A production Salesforce agent needs a reasoning model, grounded CRM and Data Cloud context, and real write access gated by field-level security.
Reasoning model
Classifies the request, retrieves CRM context, and decides whether to act or escalate
Claude Haiku 5.5
$0.10/M input tokens (up to 100K), $0.50/M outputOrder status and simple account lookups
Claude Sonnet 5.5
$2/M input, $10/M outputThe main case and lead-qualification loop
GPT-6.1 Sol
$2/M input, $10/M outputTeams already standardized on OpenAI
Our pick: Haiku 5.5 for routine lookups, Sonnet 5.5 or GPT-6.1 Sol for case resolution and lead qualification.
Knowledge & retrieval
Grounds every answer in your actual CRM records and Data Cloud, not a generic guess
Native Salesforce Data Cloud and Agentforce grounding
Included in the Agentforce licenseThe buy-SaaS path
Custom RAG over exported CRM data
Developer time plus model token costCustom builds
Our pick: Use native Data Cloud grounding on the buy-SaaS path; build custom RAG only when you need to blend Salesforce data with systems Agentforce doesn't reach.
Integrations / tools
Lets the agent actually update records and take action instead of just drafting text
Agentforce actions (Agent Builder, Agent Script)
Included in Agentforce pricingBuy-SaaS
Salesforce REST/Bulk API via MCP or custom tools
Developer timeCustom builds and n8n
Our pick: Scope every write action to the specific objects and fields the agent needs, the same discipline you'd apply to a human rep's permission set.
Observability & evals
Tracks real Flex Credit consumption and catches a wrong record update before it compounds
Agentforce usage dashboards
IncludedBuy-SaaS
Langfuse
Free up to 50K units/mo, then $29+/moCustom builds
Our pick: Track actual Flex Credit consumption per workflow before scaling it org-wide; the per-action cost adds up faster than it looks on paper.
Reference architecture
The hardest engineering problem is credit-cost visibility: because Agentforce bills in Flex Credits per action rather than per conversation, a workflow that quietly calls five actions to resolve one case costs five times what it looks like on paper, so track actual credit consumption per workflow before scaling it org-wide.
Event arrives
Service Cloud, Sales Cloud or Data Cloud triggerA new case, lead or opportunity update fires the agent
Classify and retrieve
Reasoning model plus CRM lookupThe agent reads the record and pulls related account, contact and history context
Confidence check
Guardrail layerLow-confidence or policy-sensitive items route to a human rep
Take action or draft
Agentforce action or Apex/Flow callApproved actions update the record directly; otherwise the agent drafts a response for review
Human handoff
Rep or manager queueEscalated items land with a human along with the full context and the agent's reasoning
Record update
Salesforce objectThe case, lead or opportunity is updated once the action is confirmed
Logging and audit
Agentforce usage dashboard or LangfuseEvery action and its Flex Credit cost is logged for review
Estimated cost per request
~$0.10-$0.15 per standard or voice Flex Credit action, or $2 per customer-facing conversation, before any underlying model cost on a custom build (roughly $0.01-$0.03 per interaction)
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
Estimate your monthly Flex Credit cost at your own interaction volume.
Estimated monthly cost
$345
≈ $4,140 per year
Calculator notes
- Agentforce's official rate is $0.10 per standard action and $0.15 per voice action; this calculator assumes a blended 3 actions per interaction, mostly standard rate, as a starting estimate.
- The $2-per-customer-facing-conversation rate applies to a different billing mode than per-action Flex Credits; check which model your edition uses before budgeting.
- A higher share of voice actions raises the real cost above this calculator's blended estimate.
- Building custom instead of using Agentforce swaps Flex Credits for your own model and infra cost, shown as a separate line for comparison.
Build it yourself with vibe-coding tools
n8n's Salesforce node gets a basic record-lookup agent running in an afternoon: a trigger on new cases, an AI Agent node scoped to read and write specific objects, and a human-handoff path for anything uncertain.
Time to MVP
1-3 days for a basic record-lookup agent, 2-4 weeks for a full workflow with write actions
Total cost to MVP
$0-$70 (n8n Cloud Starter at roughly 20 EUR/mo plus model API costs)
You'll need
Starter prompt
Set up a Salesforce trigger on new or updated cases using n8n's Salesforce node. Feed the case into an AI Agent node using Claude Sonnet 5.5 as the model, with the Salesforce node configured as a tool scoped to read the case, related contact and account records, and a second, narrowly scoped action limited to updating the case status or adding a case comment. Give the AI Agent node this system prompt: 'You are a Salesforce case assistant for {Company Name}. Read the case and look up the related contact and account using the lookup tool before responding. If you are at least 90% confident in a resolution grounded in the account's actual history, add a case comment with your resolution and update the status. If you are not confident, or the case involves a refund, a contract dispute, or anything outside documented policy, do not act: reply with ESCALATE and a one-sentence summary for a human rep. Never invent account history or policy details that are not in the retrieved records.' Add a Switch node after the AI Agent: route ESCALATE cases to the assigned rep's Slack channel, and let resolved cases update directly through the Salesforce node. Log every action and its outcome to a Google Sheet or Postgres table for weekly review.
Paste this into n8n
Follow-up prompts (run in order)
- 1
Add a lead-qualification branch that scores inbound leads and books a meeting through Salesforce Sales Cloud
- 2
Add a Data Cloud lookup for a unified customer profile before responding
- 3
Add a daily digest summarizing what the agent escalated and why
- 4
Add a fallback model call if the primary model times out or errors
Expected output
A working agent that resolves routine cases and updates Salesforce records within seconds, and escalates anything uncertain to the assigned rep's Slack channel with full context attached.
Known gotchas
- !Salesforce API rate limits (daily org-wide call caps) can throttle the agent during high-volume periods
- !OAuth connected-app tokens need refresh handling, or the agent silently fails
- !n8n bills per workflow execution, so a surge of cases can hit the Starter plan's monthly cap
- !The model will answer from general CRM knowledge instead of the actual record unless explicitly forced to call the lookup tool first
- !Write actions need field-level security review so the agent can't update fields a human rep couldn't
Compliance & risk reality check
A Salesforce agent inherits whatever compliance obligations already apply to your org's data, plus a few agent-specific risks.
Data access governance and least privilege
An agent with broad Salesforce API access can see or edit more than it should across accounts, contacts and opportunities.
Mitigation: Scope the connected app's OAuth permissions to only the specific objects and fields the agent needs.
Industry cloud overlays (Health Cloud, Financial Services Cloud)
If your org runs Health Cloud or Financial Services Cloud, HIPAA or GLBA obligations apply to any agent touching that data, on top of standard Salesforce security.
Mitigation: Apply the relevant BAA or GLBA vendor oversight to the agent's data access, not just the base Salesforce contract.
EU AI Act Article 50 transparency
Applies from August 2, 2026 to any agent interacting directly with people in the EU, including an Agentforce-powered chat or voice agent.
Mitigation: Disclose "I'm an AI assistant" at the start of any EU-facing interaction.
Vendor training on your CRM data
As with any AI vendor, confirm whether Salesforce or a third-party agent vendor can train on your customer data.
Mitigation: Review your Salesforce and any third-party agent vendor's data-use terms, and add a no-training clause where available.
Build vs buy: the real math
3-6 weeks for a RapidDev build; 1-3 days for a basic n8n version
Custom build time
$13K-$25K for an agency build (up to $35K for multi-cloud work); $0-$70/mo for DIY
One-time investment
Roughly the point where your workflow needs genuinely custom logic beyond Agent Builder's standard actions, or where Flex Credit consumption at scale outpaces a custom build's model cost
Breakeven vs buying
Turning on Agentforce for 5 users costs about $25/month in licenses plus Flex Credits, which for most standard service and sales workflows beats a $13K-$25K custom build in year one. The math flips once a workflow needs logic Agent Builder's standard actions don't support, or once high-volume Flex Credit consumption, at $0.10-$0.15 per action, exceeds a custom build's marginal model cost of roughly $0.01-$0.03 per interaction.
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship through our AI app development service.
Discovery call (free)
30 minWe map your exact Salesforce AI agent use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
3-6 weeks for a RapidDev build; 1-3 days for a basic n8n versionOur engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
3-6 weeks for a RapidDev build; 1-3 days for a basic n8n version
Investment
$13K-$25K for an agency build (up to $35K for multi-cloud work); $0-$70/mo for DIY
vs SaaS
ROI in Roughly the point where your workflow needs genuinely custom logic beyond Agent Builder's standard actions, or where Flex Credit consumption at scale outpaces a custom build's model cost
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does an AI agent for Salesforce cost?
Native Agentforce starts at $5/user/mo plus Flex Credits ($0.10 per standard action, $0.15 per voice action, or $2 per customer-facing conversation). A custom RapidDev build runs $13K-$25K upfront (up to $35K for multi-cloud work), plus $150-$600/mo in model and infra costs. A DIY n8n version costs $0-$70/mo but takes your own time to build and maintain.
How long does it take to set up?
Configuring Agent Builder against your existing org takes 2-6 weeks. A custom agency build takes 3-6 weeks because it has to be wired into your exact Salesforce objects and approval rules. A basic n8n version can look up and resolve simple cases within 1-3 days, though a full write-enabled workflow takes 2-4 weeks.
Can RapidDev build this for us?
Yes. RapidDev has shipped 600+ apps and offers a free 30-minute consultation to scope your Salesforce org, field-level permissions and which actions need human approval.
Will Fin and Agentforce stay separate products?
Probably not for long. Salesforce closed its acquisition of Fin on September 10, 2026, and plans to fold it into Agentforce, so expect the product boundary between the two to keep shifting.
What happens when the agent can't find enough context in Salesforce?
A well-built agent scores its own confidence and escalates to a human rep with the case and its reasoning attached rather than guessing, the same discipline that prevents an agent from inventing account history or policy it can't actually ground in your records.
Can it work with Service Cloud, Sales Cloud and Data Cloud together?
Yes. Agentforce is built to work across all three natively, and a custom build can reach any of them through Salesforce's REST and Bulk APIs, including custom objects none of the standard agent actions cover.
Is Agentforce included in my existing Salesforce license?
No, it's billed separately through the Agentforce User License ($5/user/mo) plus Flex Credits, or bundled into the pricier Agentforce 1/Max edition starting at $550/user/mo. Salesforce Foundations gives a free starting point to try Agent Builder.
Want the production version?
- Delivered in 3-6 weeks for a RapidDev build; 1-3 days for a basic n8n version
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.
