Skip to content

CUSTOMERS

Regulated operations, run by agents under human approval.

We work with a small number of enterprises in close co-build engagements. Here is the shape of the work, described honestly, without borrowed logos or invented numbers.

CASE · IN PRODUCTION

A LATAM financial institution

Agent-assisted operations on governed screens.

A regulated financial institution runs day-to-day operations on T6X screens generated from their own data. Alongside their team, AI agents work those same screens, proposing and drafting the operational writes their staff would otherwise do by hand.

Every write an agent proposes is approval-gated: it suspends for a plain-language explanation a person reviews before anything commits. Reads flow freely; writes wait for a human. And every action, human or agent, lands in a tamper-evident activity record.

The outcome that matters to them: their team moves faster on routine operations without giving up the control and the paper trail their regulator expects.

ACTIVITY RECORD · GOVERNED WRITE
ai_session · approval-gated
  • agent approval needed

    proposed write · update operation record

  • analyst approved

    reviewed explanation · Approve

  • system recorded

    write committed · row patched in place

HASH-CHAINED · SIGNED RECEIPT

CASE · IN PRODUCTION

A bank’s security operations

Every alert triaged, with a written verdict.

A bank’s monitoring stack fires far more alerts in a day than a human team can read. On T6X, agents pick up each fire as it lands, pull the surrounding context, and write a plain-language verdict an analyst can confirm or overturn in seconds.

The agents also watch for the opposite failure: detection rules that quietly stop firing. Silent alerts get surfaced and flagged for review instead of going unnoticed for months.

The outcome that matters to them: analysts spend their attention on the fires that deserve it, and nothing, noisy or silent, goes unread.

ALERT TRIAGE · AGENT VERDICT
ai_session · analyst-confirmed
  • monitor queued

    alert fired · anomalous login pattern

  • agent verdict

    wrote verdict · benign, matches change window

  • analyst confirmed

    reviewed verdict · Confirm

HASH-CHAINED · SIGNED RECEIPT

CASE · CO-BUILD

A consumer operations team in LATAM

The follow-up queue works itself.

Follow-up calls are the work that always slips: pending payments, unconfirmed orders, quotes going cold. Here, outbound voice agents work the day’s queue as flow steps, and every call ends in a transcript plus a structured result the flow can branch on.

The governance is the same as everywhere else on T6X: reads flow, writes wait. Anything the call implies (a rescheduled date, an escalation) is drafted by the agent and committed only under the team’s rules.

The outcome they are after: their people handle the exceptions and the conversations that need a human, not the dialing.

VOICE FOLLOW-UP · CALL RESULT
flow_run · voice step
  • agent completed

    call placed · follow-up on a pending payment

  • flow recorded

    extraction recorded · commitment for Friday

  • system queued

    next step scheduled · reminder before the date

HASH-CHAINED · SIGNED RECEIPT

EARLY ACCESS

We build alongside a small number of enterprises.

T6X is delivered today through co-build engagements: we work directly with your team to stand up your first governed, agent-native operations. If that fits how you want to adopt AI, let’s talk.

HASH-CHAINED · SIGNED RECEIPT