A human stays in the loop
You decide what an agent can do on its own and what needs a person. Anything sensitive — refunds, discounts, big commitments — pauses for approval before it happens.
Trust & security
Trust starts with knowing what runs, where, and under whose control.
Handing your phone and your follow-up to AI only works if you stay in control. So every build documents account ownership, the guardrails you approve, the records each workflow creates, and the conditions that require a human handoff.
Guardrails
Guardrails define what may run automatically.
Visibility and control aren’t features we bolt on at the end — they’re how every system is built from day one. You climb the autonomy ladder at your pace, and anything sensitive waits for a person.
It only does what you scoped
Each agent runs to a plain-English playbook you sign off on: its job, its limits, and what it must never touch. Hard limits keep it out of anything outside its lane.
It escalates, it doesn't guess
We define uncertainty and risk thresholds during setup. When a connected workflow supports handoff, the agent pauses or routes the available context to the approved human path.
The audit log
Know what each connected workflow records.
The portal gives you a current operational view, while detailed records remain in Cognautic or the connected source system named in the scope. We document what is logged, where to find it, and how it can be handed off instead of implying one dashboard already contains every source-system record.
Operational records by design
Configured workflows write timestamps, status, and available decision context to Cognautic or the connected source system. The exact record set is documented in the build scope.
A portal view of connected operations
The current portal surfaces plan and wallet status, recent activity, and connected system summaries. Transcripts, call logs, and customer records remain in their source systems unless a scoped portal surface is added.
Changes follow a release path
We test changes and use approvals, versioned configuration, or rollback procedures where the selected integration supports them. The implementation plan states what can be reversed automatically and what needs manual recovery.
Ownership
You can’t delegate to agents you can’t see. So we built you the cockpit.
Customer-owned accounts and source data stay in your name. Managed runtime is identified separately, with a written exit path for what remains, what is exportable, and what must be migrated if service ends.
Your accounts, your keys
Where the scope uses your phone, CRM, payments, or provider account, it stays in your name. Managed Cognautic infrastructure is identified separately before the build starts.
Your source data stays yours
Call logs, customer records, and transcripts stored in customer-owned source accounts remain available through those providers' export tools. Any Cognautic-hosted handoff or export is defined in the exit plan.
Never sold, never trained on
Cognautic does not sell customer data or use it to train public models. Data handling by selected infrastructure and model providers follows the account, settings, and terms approved for the build.
A documented exit path
You keep the customer-owned accounts, phone numbers, data, and exportable assets named in the scope. Cognautic-hosted runtime and managed features may stop when service ends, so the exit plan identifies what stays and what must migrate.
Reliability
Fallbacks and monitoring are scoped around the dependencies you use.
Fallback routing
Where the selected phone and workflow providers support it, we configure fallback routes and define what happens when a dependency is unavailable.
Monitoring that matters
We monitor the systems and failure signals included in your plan, with escalation paths documented for phones, follow-up, bookings, and other scoped workflows.
A backup path to a human
Critical flows can include an approved human destination and whatever summary or context the connected systems make available.
Compliance posture
Consent-first calling and messaging, by design.
Consent for AI calls
Our AI callers identify themselves, and automated outreach only goes to contacts with proper consent on record. Consent-first isn't a setting — it's the default.
Opt-out handling is scoped by channel
Built-in email outreach honors suppression records. Text, voice, and cross-channel opt-outs require the selected provider controls and synchronization to be configured and tested before activation.
TCPA-aware messaging
Quiet hours respected, consent records kept, and messaging flows designed around TCPA rules. We review your specific use case on the consult, not after launch.
Straight talk on certifications:we’re a small, hands-on team, and we don’t currently hold our own SOC 2 or HIPAA certificate. What we do is build on serious, audited infrastructure — Supabase, Stripe, Twilio, OpenAI, Anthropic, Google — and follow their security best practices, on your accounts. If your business needs a specific compliance posture, tell us on the consult and we’ll scope it honestly up front, rather than wave a badge we haven’t earned.
The things owners ask before handing over the phone
Access is tied to approved Cognautic portal users and to permissions in customer-owned source accounts. Portal owners can invite teammates, assign owner or member access, resend invitations, revoke access, and reactivate members; account-changing actions and the bounded portal export remain owner-only. Cognautic access is limited to the people and service credentials needed to deliver and support the scoped system.
Cognautic does not sell customer data or use it to train public models. Calls, contacts, and customer records are processed only as needed for the scoped automations and support. The selected providers, account ownership, retention settings, and any masking requirements are documented during implementation.
The playbook defines when the agent must pause, ask for approval, or use a human handoff. Operational records and recovery options depend on the connected workflow, so the implementation plan identifies what is logged, where it appears, and whether recovery is automatic or manual.
Yes. Portal owners can run a self-service bounded portal export of Cognautic-hosted records — activity events, AI usage ledger, competitor intelligence, agent memory, and fleet module state — with per-section limits stated in the file and credentials always stripped. You can cancel under the service agreement and keep the customer-owned accounts, phone numbers, data, and exportable assets listed in the scope; source-system exports use those providers' tools. Cognautic-hosted runtime and managed features may stop, so the written exit plan identifies any data handoff or migration needed before service ends.
Cognautic doesn't hold its own SOC 2 or HIPAA certificate today, and we won't pretend otherwise. We build on audited infrastructure — Supabase, Stripe, Twilio, OpenAI, Anthropic, Google — and follow their security best practices on your accounts. If you need a specific compliance posture, we scope it honestly up front.
Compliance depends on the channel, consent evidence, purpose, jurisdiction, provider, and message. Cognautic's built-in email outreach uses suppression records; text, voice, quiet-hour, disclosure, and cross-channel opt-out behavior must be implemented and tested in the selected provider workflow. We keep unsupported audiences out and review the proposed flow, but no generic configuration can guarantee legal compliance for every use case.
Free consult
Get the keys — and the receipts.
Bring your toughest trust question to a free consult. We’ll show you exactly how the oversight, ownership, and guardrails work for your business — before you commit to anything.
- Exactly which accounts and data stay in your name
- Which actions you'll always approve by hand
- How the audit log and rollback actually look
- Any compliance needs, scoped honestly up front