The agents are the easy part.
Here's the rest.
What a team needs sits under each agent: chaining, scale, scheduling, grounding, fallback, routing and ownership.
What actually runs under an agent
An agent is the part you can see. Underneath it sits the machinery that makes it usable by a whole team, and this is our own map of it. Each row below stands on 1 node of it.
Eva runs it, and helps you build it
Ask her in words and she launches an agent or a workflow on the records you meant. She also writes the things that run: the agents themselves, the personas and verticals they read, the workflows they sit in. She works inside the permissions the person asking already has.
Specialists, not 1 model with a long prompt
Each agent does 1 job and carries its own settings: what it may conclude, how hard to look, what it costs per run. They are grouped the way the library groups them, so the roster reads as a team rather than a feature list.
You launch once and the schedule handles every run after
A workflow is a chain of agent steps. Launch it across the records your filters select, or put it on a daily, weekly or monthly schedule and leave it. A safety cap pauses the automation and tells you rather than quietly working part of the list.
Accounts that qualify carry on, and you see which branch
A Condition step sends each record down its own branch by re-reading fields on the record, not the previous agent's answer. So the accounts that qualified carry on to contact finding and the rest stop, and you can see which branch a record took.
Every answer carries its source
Each criterion you wrote gets an answer, a date and the reasoning behind it, on the record's Qualification & Research tab. The verdict is not a score: 1 failed qualification criterion is a No, and the No names that criterion.
No single source has everyone
When a source comes back with nothing the next is tried, inside the same run, and the report shows it happening rung by rung. Phone intent scoring is a switch on the step itself, and the score then sits on the contact's own number for reps to sort by.
Your CRM stays the system of record
A connected CRM shows what syncs and in which direction, per object. Each agent's answer arrives as its own field, mapped per direction in one bulk action rather than one field at a time, so what the agents learn lands where your team already works.
You set the effort. We keep the model current.
Agents run on models from OpenAI, Google and Anthropic, and you never pick one. You set effort and speed on the agent itself, and where an agent takes those settings they are what its price follows, not the model that happened to serve it.
See what a rule will take before it takes it
Rules run in priority order and the first match wins. The preview counts what each one would claim and writes nothing, so you can reorder before anything moves. A record that matches no rule is left alone rather than parked with somebody.
The material is yours, and the agents read it
Your persona cards and the agents that run against them sit in 1 place. Publish a change to the value proposition and every persona is queued to be rebuilt against it. Research agent instructions are written when you generate them, so regenerate those to pull a later rewrite in.
Your material stays yours
Under the MSA, everything you upload and everything the agents produce from it is yours.
Never used to train models
Your CRM data, research outputs and agent activity are never used to train or fine-tune models. Not by us, not by our providers.
Your workspace is isolated
Each workspace is logically isolated, so your data is never accessible to another customer.
You control retention
If you cancel, your data is returned or deleted, your choice. If you make no request, all personal data is permanently deleted within 30 calendar days.
SOC 2 audited. ISO 27001 certified. GDPR compliant. The detail, the sub-processors and the DPA are on Security and trust.
See the layer running
We will walk your RevOps lead, and whoever they pull in, through the chain, the fallback and the ownership rules on a real workspace.
Request a demo