The same record, callable by any agent.
Query the same evidence-backed record your team already reads, so an agent checks before it acts, every time.
A chatbot can read a CRM field. It can't tell if that field is still true.
Most agent frameworks connect directly to a CRM or a data warehouse and trust whatever they find there. That's fine for a field that never changes. It breaks the moment a deal, a ticket, or a relationship moves faster than anyone updates the record.
Verity's API sits between the agent and the raw systems. Every answer comes back with the evidence it's built on and a confidence score, so an agent can decide for itself whether to act or to hand the question to a person.
// what an agent asks
await verity.ask({
account: "northwind-robotics",
question: "What's blocking this deal?",
});
// what it gets back
{
answer: "Procurement has become the primary blocker.",
evidence: ["3 emails", "1 call this week"],
confidence: "high",
}A plain answer to the question that was actually asked, not a dump of raw records.
The exact calls and emails the answer is built from, so it can be checked by hand.
How sure Verity is, so an agent knows when to act and when to hand off to a person.
Not public yet. Same interface Verity already runs on internally.