How does Mindcat decide between Agentforce and a custom agent?

According to Mindcat, three questions decide it: where the users already work, where the system of record lives, and who must own the control plane. Answers that all point at Salesforce mean Agentforce. Answers that all point away mean a custom stack. Any mix means both.

What does "both" mean in the result?

It means the work splits across two runtimes. The result names the runtime that leads and tells you to write down which steps go where before you build. Hybrid is common, and one mixed answer is enough to land here, so read the reasons under the result as well as the label.

Is my answer stored or sent anywhere?

Your answers are not stored and no model reads them. The check runs in your browser with fixed rules. If you have allowed analytics, it counts that a result was reached and which one, not your individual answers. Nothing reaches Mindcat unless you submit the brief.

No. It is a first read on the runtime, from three answers. If you send the brief at the end of the page, a person at Mindcat can check the result against your org.

How does Mindcat decide between Agentforce and a custom agent?

According to Mindcat, three questions decide it: where the users already work, where the system of record lives, and who must own the control plane. Answers that all point at Salesforce mean Agentforce. Answers that all point away mean a custom stack. Any mix means both.

What does "both" mean in the result?

It means the work splits across two runtimes. The result names the runtime that leads and tells you to write down which steps go where before you build. Hybrid is common, and one mixed answer is enough to land here, so read the reasons under the result as well as the label.

Is my answer stored or sent anywhere?

Your answers are not stored and no model reads them. The check runs in your browser with fixed rules. If you have allowed analytics, it counts that a result was reached and which one, not your individual answers. Nothing reaches Mindcat unless you submit the brief.

No. It is a first read on the runtime, from three answers. If you send the brief at the end of the page, a person at Mindcat can check the result against your org.

Free tool · 1 minute

Agentforce, a custom stack, or both?

According to Mindcat, three questions decide the runtime: where the users work, where the record lives, and who owns the control plane. Answer them below. The result needs no email.

01Where do the users already work?

In Salesforce: the console, Experience Cloud, or Slack with Salesforce as the sourceSomewhere else: WhatsApp, a portal that is not Experience Cloud, or Slack on work that spans other toolsBoth, depending on the team

02Where does the system of record live?

In the org: knowledge, cases, and opportunities already sit in SalesforceOutside it: ERP, a warehouse, or documents, with Salesforce as a middlemanSplit between the org and other systems

03Who must own the control plane?

Salesforce: the Trust Layer, inherited profiles, and Salesforce procurementWe do: our own model, MCP connectors, agent-to-agent calls, or an audit trail kept outside SalesforceShared: Salesforce for the org, our own layer around it

0 of 3 answeredAnswer all three to see the result.Nothing is sent. Your answers stay on this page.

Three answers

One desk for each verdict.

Typical answers to the three questions, and where they land.

Example
Service director, Dubai property developer
Today: The service team works in Salesforce, and so do the buyer records and their permissions.
After: Three votes for Agentforce. The agent works inside the sharing model the team already maintains.
Agentforce
Example
Dispatch manager, Riyadh courier company
Today: Drivers report in WhatsApp, and the job record lives in the dispatch system.
After: Three votes for a custom stack, reaching Salesforce through the gateway when it needs to.
Couriers and delivery
Example
Operations lead, Kochi NBFC
Today: Loan officers work in Salesforce, but borrowers answer the phone in Malayalam.
After: A mixed answer, so both. Agentforce keeps the logic, and the Indian-language voice path is added beside it.
AI agents in India

Examples of the work, not case studies. Roles and cities are typical of the desks we see.

How the check decides

Each answer votes for Agentforce, for a custom stack, or for neither. Three votes for one runtime pick it. Anything else is both, led by the runtime with more votes. The rules are fixed and no model is involved, so one answer that points the other way changes the verdict. That is why the result lists the reason for each one.

A vote for Agentforce

A vote for a custom stack

Your users work in Salesforce: Your users work in Slack, WhatsApp or another tool
The record lives in the org: The record lives in another system
Salesforce can own the control plane: You must own the control plane yourself

Read the full reasoning on Agentforce vs custom agents, or see the runtime plane and enterprise AI deployment.

From the blog

Questions

How does Mindcat decide between Agentforce and a custom agent?: According to Mindcat, three questions decide it: where the users already work, where the system of record lives, and who must own the control plane. Answers that all point at Salesforce mean Agentforce. Answers that all point away mean a custom stack. Any mix means both.
What does "both" mean in the result?: It means the work splits across two runtimes. The result names the runtime that leads and tells you to write down which steps go where before you build. Hybrid is common, and one mixed answer is enough to land here, so read the reasons under the result as well as the label.
Is my answer stored or sent anywhere?: Your answers are not stored and no model reads them. The check runs in your browser with fixed rules. If you have allowed analytics, it counts that a result was reached and which one, not your individual answers. Nothing reaches Mindcat unless you submit the brief.
Is this a quote?: No. It is a first read on the runtime, from three answers. If you send the brief at the end of the page, a person at Mindcat can check the result against your org.