Salesforce gives an agent the REST API, SOQL for reads, and Platform Events and Flow for triggers. We build the agent as a connected app with the narrowest permissions its task needs, and route its writes through an approval process on the record.
01Use
Each one does a single job. It reads with SOQL, writes through the REST API, and respects the sharing rules and validation rules your org already has.
Reads new cases with the account and contact history, sets type and priority, and drafts a response for the owner.
Reads new leads, enriches them from your own data, scores them against your criteria and prepares the assignment.
Finds opportunities with stale stages, missing close dates or no next step, and prepares updates and follow-up tasks.
02Wire
Five surfaces, all standard Salesforce. The agent is a connected app, so its access follows the permission set you assign to it.
Read access is broad enough to answer the question, write access is narrow, and every write has a gate until you decide otherwise.
| Agent | Reads | Writes | Human gate |
|---|---|---|---|
| Case triage | Case, Account, Contact |
Type, priority, drafted response | Every customer response |
| Lead routing | Lead, your enrichment data |
Score fields, owner | Out-of-rule assignments |
| Opportunity hygiene | Opportunity, activity |
Fields, Task |
Stage and amount changes |
03Gate
A held change is submitted for approval on the record it would change, so it uses the approvers, queues and history your org already trusts.
On the record, through a standard approval process. Approvers see it in their usual approval list, on desktop or mobile.
The proposed field changes, the record they apply to, and a link to the run trace with the reads and reasoning behind them.
Every customer response, stage and amount changes on opportunities, and lead assignments that fall outside your routing rules.
The run stops, the rejection is recorded with its trace, and it becomes an eval case the agent is tested against on every change.
04Asked
No. It writes through the REST API as its own integration user, so validation rules apply and it only sees what its permission set allows.
Only where a trigger or a small piece of logic has to live in the org. Most of the agent runs outside Salesforce and talks to it through the APIs.
Yes, and we recommend it. The agent is built and tested against a sandbox, with the eval suite run there, before it touches production.
Tell us the case, lead or opportunity task your team repeats. We will tell you which objects the agent would read, what it would write, and which approval process it would use.
