The five objections we hear before the discovery call — answered plainly.
Autonomy and guardrails. The audit log. The named-human handoff. Insurance notification triggers. The flat subscription. Each answer below is consistent with the same claims across the home page and /pricing— no new numbers, no new promises.
Autonomy without losing the loop.
What stops an autonomous agent from going further than you wanted?
Every enrolled endpoint runs four stations — Watch, Patch, Contain, Escalate — in sequence, with explicit guardrails between them. Suspicious processes are quarantined, identity tokens revoked, and network egress restricted the moment confidence drops below threshold. The agent never closes the loop alone: every action — patch, quarantine, token revocation — produces a row in the audit log before it produces a notification, and a reversal is its own row.
Out-of-window changes are held, summarised, and only applied after explicit sign-off. Patches only apply inside the maintenance windows you configure at enrolment — one window on Starter, two on Standard, three on Complete.
Watch
EDR, identity, and SaaS signal stream into a triage agent that ranks and correlates in seconds.
Patch
Routine OS and third-party patches apply inside the maintenance window you configured.
Contain
Crossing the confidence threshold triggers quarantine, with a full actions trail attached.
Escalate
A named technician is paged only for incidents that exceed agent confidence — never for noise.
An audit log a regulator will accept.
How do you prove what the agent actually did, after the fact?
Each action produces a row in the audit log BEFORE it produces a notification. Row contents are fixed: agent ID, evidence, and outcome. The log is Merkle-anchored, append-only, and readable by both the customer and the auditor from the same source of truth. The Merkle head is anchored off-platform so it cannot be quietly rewritten.
The export shape depends on your tier — same log, different reader. Starter exports CSV on demand, Standard groups rows by containment lineage and ships with a 15-minute acknowledgment SLA, Complete adds a weekly attestation pack for the underwriter.
Merkle-anchored
Each row is hashed into the chain; the head is anchored off-platform.
Append-only
A reversal is a new row, never an edit — the timeline stays provable.
Customer-read & auditor-read
Same source of truth; the underwriter reads what your CISO reads.
A named human, only when it counts.
When does a real person actually get paged?
Escalations are reserved for incidents that exceed the agent confidence threshold. The rule is narrow: a human is paged only when (a) an incident crosses the confidence threshold, (b) a governance review requires sign-off, or (c) a cyber-insurance notification trigger fires. Routine triage, correlation, and patching never page an analyst.
The shape of the human handoff depends on the tier — the rule is the same, the depth is calibrated.
Starter
A named operator above the confidence threshold. No published acknowledgment SLA.
Standard
A named operator with a 15-minute SLA to acknowledge the handoff.
Complete
A named technician on call during the maintenance windows you declare.
Cyber-insurance notification triggers.
Which events page the customer AND their underwriter?
A notification to the customer and their underwriter fires when the audit log records one of the trigger events below. The goal is a paper trail that matches the policy language — no surprise phone calls, no missing rows.
Confirmed containment lineage change
When an alert crosses the confidence threshold and shifts into the containment lineage, the row fires the trigger.
Complete tier escalation
A named-technician handoff on the Complete tier — the customer and underwriter both see it.
Audit anchor divergence
If the anchored Merkle head fails to reconcile at the next anchor cycle, the divergence itself is a row and a trigger.
Identity-graph / token-graph mutation
A revoke or rotate on the identity graph (Okta / Entra ID) writes a row that underwriters expect to see.
Flat subscription, not hourly surge.
Why per endpoint, and why no surge pricing when an incident lands?
Pricing is $10–$25 per endpoint per month across three tiers (Starter, Standard, Complete) — a flat subscription, not a meter you only notice after it ticks over.
The rationale is the same as the audit log itself: fees don't grow because an attacker probed harder. The flat model removes the perverse incentive against declaring an incident and aligns with the predictability a mid-market security budget needs to plan around. Onboarding the wider org does not re-license (no per-user fees). An incident does not change your invoice — that is by design.
The smaller questions, in the same register.
Scope, onboarding, billing windows, export shapes — the follow-up details behind the five answers above. If yours isn’t on the list, write to us — a real person reads it.
Answers covered, fleet next. Enrol one tier, watch the audit log fill in.
curtainwall@polsia.app