Agentforce.
AI agents you can trust.
Salesforce Agentforce consulting and implementation for sales, service and marketing. Build agents grounded in your data, with clear guardrails and human handoffs.
“Can you help with my renewal?”
Understand. Reason. Act.
What is Salesforce Agentforce?
Agentforce is Salesforce’s platform for building AI agents that work inside your Salesforce environment. Each agent is given a role, instructions and a set of approved actions.
Unlike a scripted chatbot, an agent uses your CRM and Data 360 data to decide how to respond and what to do next, such as updating a record, drafting a quote or handing the conversation to a person. Agentforce consulting is the work of choosing the right use case, preparing the data and setting the limits so that it does this reliably.
AI agents for sales, service and marketing teams
Pick a team to see the kind of work agents are scoped to do. We start with one narrow, repeatable job and expand from there.
Service agent
Answers customer and employee questions from your knowledge and records, and takes approved actions on cases and orders.
- Answer product, policy and order questions from your knowledge base and Salesforce records
- Open, update and route cases with the full context attached
- Escalate to a person with a summary when the request is sensitive or the answer is unclear
Sales agent
Works inbound leads and follow-ups using account and opportunity context, so your reps spend time on conversations that matter.
- Respond to new leads and capture the details your team needs
- Draft follow-ups and meeting requests grounded in CRM history
- Route qualified conversations to the right rep with notes
Marketing agent
Supports campaign and journey decisions with engagement data, within the brand and compliance rules you set.
- Draft and adapt campaign copy within brand and compliance guidelines
- Use engagement signals to suggest the next step in a journey
- Hand high-intent contacts to sales with context
How we build an Agentforce agent, layer by layer
Six workstreams take an agent from a use case on paper to a governed agent in production. Select a layer to see what we do in it.
Discover and map the use case
We start with the job: which task, for whom, with what data, and what “done” looks like. The goal is one focused first use case that is worth automating and safe to automate.
- Use-case shortlist and scoring
- Data and permission check
- Success measures and a named owner
Ground the agent in your data
An agent is only as good as the data it can reach. We connect CRM records and knowledge articles and, where the use case needs it, Data 360 so that answers come from your own information.
- CRM and knowledge grounding (retrieval)
- Data 360 for unified, multi-system data
- A clean-up list for the gaps we find
Configure topics, instructions and reasoning
We define the agent’s topics and instructions in plain language and use Agentforce Builder, with Agent Script where a process needs tighter control.
- Topics and plain-language instructions
- Tone, scope and refusal rules
- Test conversations for every topic
Build custom actions
Actions are what turn an answer into an outcome. We build them with Flow and Apex, and connect your other systems through APIs where the use case needs it.
- Flow and Apex actions
- Record updates, case creation, quotes and approvals
- Connections to your other systems
Add guardrails and human handoffs
Sensitive actions should not run unchecked. We add approval steps and escalation paths, and test the agent against edge cases before anyone outside the project team sees it.
- Approval steps for sensitive actions
- Escalation to people with context attached
- Pre-launch testing on edge cases
Deploy, observe and tune
We put the agent where your customers and teams already work, then review conversations so instructions, topics and actions keep improving.
- Web, email and messaging channels, and Slack where it fits
- Post-launch conversation review
- A prioritized improvement backlog
Is your Salesforce org ready for an AI agent?
Tick what is already true. It takes a minute and shows where an Agentforce project would start.
Start with discovery
Tick what is already true. Whatever your score, the first step is the same: pick one use case and check the data behind it.
Book a Strategy SessionFrom first use case to production agent in five steps
- 01
Discover
Choose the first use case, the data it needs and how success will be measured.
- 02
Ground
Connect CRM records, knowledge and, where needed, Data 360 so the agent has something reliable to work from.
- 03
Build
Configure topics, instructions and actions, including Flow and Apex where the agent needs to do things.
- 04
Test and govern
Run realistic and awkward conversations, add approvals and set escalation paths before launch.
- 05
Launch and tune
Deploy to your channels, review real conversations and improve the agent over time.
Guardrails are part of the build, not an afterthought
An agent that can act needs limits. We design them with you before the first conversation goes live.
Defined scope
Each agent has a clear job and a written list of things it must not do.
Approvals where it matters
Sensitive actions, such as discounts, refunds or record changes, route to a person first.
Respect for permissions
Agents work within the access and data rules configured in your Salesforce org.
Tested before launch
We run realistic and awkward conversations before anyone outside the team sees the agent.
Reviewed after launch
We review conversations so instructions, topics and actions can be tightened.
Clear ownership
Someone on your side owns the agent after launch, with a way to see what it is doing.
Agentforce works best with the rest of your Salesforce stack
Agentforce questions, answered plainly
What is Salesforce Agentforce, and how is it different from a chatbot?
Agentforce is Salesforce’s platform for building AI agents that work inside your Salesforce environment. A scripted chatbot follows fixed paths. An Agentforce agent is given a role, instructions and a set of approved actions, then uses your CRM and Data 360 data to decide how to respond and what to do next, such as updating a record, drafting a quote or handing the conversation to a person.
What can an Agentforce AI agent do for my team?
Common starting points are service agents that answer questions and manage cases, sales agents that respond to and qualify inbound leads, and marketing agents that support campaign and journey decisions. The right first use case is a narrow, repeatable task where the data is already in Salesforce and the rules for escalation are clear. We map candidates in discovery and recommend one to start with.
How much does Agentforce cost?
Salesforce lists a free starting point for building agents, Flex Credits at $500 per 100,000 credits (you pay per action), per-user add-ons from $125 per user per month for Sales, Service, and Field Service and Operations, and editions with Agentforce included from $550 per user per month. These figures are from Salesforce’s pricing page as of November 10, 2026 and can change, so confirm them before budgeting. Implementation and integration work is separate and depends on scope.
Do we need Data 360 (formerly Data Cloud) to use Agentforce?
Not always. An agent can work with data that already lives in your Salesforce org. Data 360 becomes valuable when the data an agent needs is spread across several systems or has to be unified and cleaned before the agent can rely on it. We check this in discovery so you only add what the use case needs. See our Data 360 services.
Is Agentforce secure, and how do you keep agents within limits?
Agents act within the permissions, topics and actions you configure in Salesforce. We define what an agent may and may not do, add approval or escalation steps for sensitive actions, test against edge cases before launch, and review conversations after launch so instructions can be tightened. Your own security and compliance requirements should be reviewed with your internal teams.
Will Agentforce replace our employees?
Agentforce is designed to take on defined, repeatable tasks and hand off the rest. We design the handoffs so that people stay in charge of judgment calls, exceptions and customer relationships, and so that your team can see what the agent did and why.
How long does an Agentforce implementation take?
It depends on scope: how many use cases, how ready the data is, which systems need to connect, which channels you use, and how much testing and approval your organization requires. Starting with one focused use case keeps the first release manageable, and we set a realistic timeline in discovery once those factors are known.
What does an Agentforce implementation partner actually do?
A partner helps you choose the right first use case, prepare and connect the data, configure topics, instructions and actions, build custom actions with Flow and Apex, test and add guardrails, deploy to your channels and tune the agent after launch. We explain this in more detail in our guide to Agentforce implementation partners.
Tell us the first job you want an agent to do.
In a Strategy Session we look at your use case, your data and the guardrails you would need, and tell you honestly whether Agentforce is the right fit and what a focused first release would involve.