Overview
Agentforce consulting: AI agents that work inside your Salesforce org
Salesforce's AI used to be branded Einstein. Most of it now sits under Agentforce, Salesforce's product for building and running AI agents that act inside Salesforce, such as handling service cases or qualifying leads. Einstein is still the name for the AI built into the platform for insights, predictions and generated content, and for the Einstein Trust Layer that protects data sent to large language models. This page keeps its old Einstein address, but the work it describes is mostly Agentforce.
An agent is built from instructions, actions and guardrails. In Agent Builder you describe the jobs an agent does in natural language, then turn existing Flows, prompt templates, Apex and APIs into actions it can take. Salesforce's Atlas Reasoning Engine breaks each request into smaller tasks and plans the steps. Agents can also reach external tools through MCP servers, and several specialised agents can be orchestrated together.
Agentforce suits Salesforce customers with a frequent, well-defined process, such as case triage, order status questions or lead qualification, and data good enough to ground an agent. Great Minds is a Sydney-based systems integrator that works in the gaps between platforms, so we often connect agents to systems outside Salesforce: an ERP for order status, Jira for engineering escalations, or a knowledge base in Confluence or SharePoint.
Capabilities
What an Agentforce project covers
We start small: one agent, one process, measured against how the work is done today. We expand only when the first agent is doing its job.
Agentforce works across Sales Cloud and Service Cloud. For the service side in depth, including Agentforce in the contact centre, see our Service Cloud page.
Connected platforms
Agentforce connected to the systems outside Salesforce
Agents are only as useful as the data and actions they can reach. Agentforce supports MCP servers for external tools, and actions can call MuleSoft integrations or external APIs for systems such as ERP, ecommerce and service desks.
We build those connections and keep each one limited to the data the agent needs. Our MCP integration work covers the same approach for AI agents beyond Salesforce.
How we deliver
From the first use case to agents in production
We map one process end to end, measure its volume and handling time today, and agree what the agent may and may not do, and when it must hand over to a person.
Agent, actions and grounding built in a sandbox and tested against real past conversations and records, including the awkward ones, before a pilot with a small group.
A limited release to one team or channel, with transcripts reviewed, wrong answers fixed and results compared with the baseline before anything is widened.
More jobs, channels or agents once the first one works, with an agent register, a change process and monitoring that your Salesforce team owns after hand-over.
Australian considerations
Privacy, residency and governance for Agentforce in Australia
Agents read and act on personal information, so the Privacy Act applies to every prompt and action. The Einstein Trust Layer is Salesforce's main control here: Salesforce lists secure data retrieval, dynamic grounding, zero data retention and toxicity detection among its protections. We check how each applies to your configuration and document it for your privacy team. If an agent makes or supports decisions that significantly affect people, such as declining a refund, privacy policies must disclose that from 10 December 2026; our guide to automated decisions and the Privacy Act explains the change.
Salesforce runs Hyperforce in Australia, and Data 360 has been available on it since 11 December 2023. Generative features call large language models, and where those requests are processed depends on the feature and model, so we confirm it for each one before go-live rather than assume it stays onshore.
For APRA-regulated entities, an agent that supports a critical operation falls within CPS 230's operational risk requirements, and its access to data within CPS 234. Approval steps, logging and a fallback to a person are part of the design, not added later. Costs are in AUD: at the time of writing, Salesforce's Australian price list shows Agentforce for Sales as an add-on from AU$175 per user per month.
FAQ
Agentforce and Einstein questions
Is Einstein AI the same as Agentforce?
Not quite. Agentforce is Salesforce's product for building and running AI agents, and Salesforce says it was formerly called Einstein Copilot. Einstein is still the name for the AI built into the platform and for the Einstein Trust Layer. In practice, most new Salesforce AI work is Agentforce work.
What does Agentforce cost?
Salesforce offers three paid models: Flex Credits, Resolutions and per-user licences. It also says every customer can start with Agentforce for free under Salesforce Foundations. We model each option against your expected volumes before you commit, because the cheapest choice depends on how much the agent will be used.
Do we need Data 360 to use Agentforce?
Not always. Agents can be grounded in Salesforce records and knowledge articles. Data 360 matters when agents need data from outside Salesforce, or customer records unified across several systems. Our Data 360 page explains when it's worth adding.
Can Agentforce work with systems outside Salesforce?
Yes. Actions can call external APIs and MuleSoft integrations, and Agentforce's MCP support lets agents use external tools exposed as MCP servers. We keep each connection limited to the data and actions the agent actually needs.
How do we stop an agent giving wrong or unsafe answers?
Narrow instructions, grounding in approved content, Salesforce's agent guardrails and Trust Layer, testing against real past conversations, and a hand-off to a person for anything out of scope. After go-live, transcripts and the audit trail show what each agent did.
How long does a first Agentforce agent take?
It depends mostly on how ready your data and knowledge content are, and whether new integrations are needed. We give you a scoped plan and price before work starts.
People
Salesforce and AI consultants who stay accountable
Agentforce projects are led by senior consultants who understand both the Salesforce platform and the process being automated. They work with your Salesforce admins, service and sales leads, and privacy team.
AI agents help us build and test faster; people decide what an agent is allowed to do.
Case studies
The CRM and integration work agents depend on
We don't yet have a published Agentforce case study. The work below shows the Salesforce data, marketing automation and integration foundations that agents rely on.
4x faster campaign execution, 1M+ contacts migratedSalesforce Marketing Cloud to Klaviyo migration with 1M+ email marketing recipientsUnlocking personalised growth at scale through Klaviyo implementation and data migration
Marketo and Salesforce connected for 2 fund launchesMarketo and Salesforce, connected for a global fund launchMarketo and Salesforce integrated into one marketing automation engine, so Montgomery could launch two global equity funds to the right investors at the right time.
A 100-person firm running on Salesforce Sales CloudSalesforce CRM implementation for Hong Kong law firmDigital transformative journey of a 100-person law firm in Hong Kong powered by Salesforce Sales Cloud.Insights
Reading on AI agents, Salesforce and governance
The connected enterprise: integrating Salesforce, Atlassian, Adobe and ERPHow to connect Salesforce, Atlassian, Adobe and your ERP: system-of-record map, architecture options, API limits, APP 8 and CPS 230, and where to start.
Salesforce training and user adoption: a practical guideDesign for adoption, role-based training on realistic data, in-app guidance, measuring real adoption and preparing users for Agentforce.
Salesforce for fund managers: investor relations, distribution and DDOWhich Salesforce products fit, modelling advisers, platforms and investors, registry integration, DDO and AML/CTF support, and AI.