Values — How We Operate

This page states the values behind how we run this company and every agent we ship. Each one is grounded in a practice that exists today — most of them are spelled out in more detail on Trust & Operating Principles, and the promises we make to customers are tracked line by line in an internal obligations matrix so they can be verified, not just admired.

Tell the truth, especially about ourselves

We don't dress up estimates as results. Every performance figure on this site that comes from our deterministic dry-run simulator is labeled simulated, and our case-study tables stay empty until a real engagement fills them with measured data. We have no customer logos or invented track records on this page on purpose — we'd rather show you an honest empty table than a decorated fake one. Details: Trust & Operating Principles.

Humans decide; agents draft, score, and route

Our agents are structurally incapable of deciding on their own to contact a person, and they escalate to your humans when uncertain. Outreach goes through approval gates, and situations outside an agent's delegated authority — contract questions, unusual spend, potential security events — reach a human immediately instead of being improvised. Autonomous does not mean unsupervised.

Collect the minimum data we need

We use your data for three things only: running your agents, billing accurately, and keeping the service safe. We don't sell it, don't run ad trackers, and don't mix it across customers — each client's workspace is an isolated unit with its own knowledge base and memory, and nothing from one customer's workspace trains another's agent. We don't use your data to train AI models. The full picture is in the Privacy Policy.

Escalation over surprise

Agents fail — anything that runs 24/7 eventually does. What we commit to is response: issues surface immediately instead of festering, anything unusual sits outside delegated authority the moment it's unusual, and a human on our side is looped in without waiting for you to notice. Boring and predictable is the point. If something does go wrong, we tell you the truth about what happened.

Guardrails live in code, not in prompts

Business rules that must never be broken — booking hours, slot conflicts, spend caps, contact policy — are enforced deterministically in the agent's action layer. A language model can interpret a request; it cannot override a hard guardrail. Every blocked action carries an explicit machine-readable reason.

Plain language everywhere

Our contracts, policies, and pages are written to be read, not to be survived. If a sentence wouldn't survive being said out loud on a sales call, it doesn't belong on the site. Where a document has to be precise (like the Terms of Service), we keep it short enough that you can actually read it — and where your signed agreement and a web page ever disagree, the agreement you signed wins.