DeepBlocker / Products / 02 · Shield
BLUE TEAM · PRODUCT 02

Shield

An AI agent stands on your most valuable phone lines and checks every caller in real time. Deny by default. Allow only the verified.

Request a pilotSee pricing
In one sentence
A guard for your payment desk and executive lines that knows the voices of the people allowed to give instructions, and challenges everyone else, on every call.
What it is
Softwarean AI agent on your line
Who buys it
Regulated firmsafter Assess shows the risk
Direction
Inboundcallers reaching your desk
Status
Pilot Q3 2026early deployments

The Verified Caller List

This is the heart of Shield, and the part that does not go out of date. A payment desk hears instructions from a small, known set of people, perhaps 30 to 200 voices. Enroll them once. From then on, the agent knows who is real.

01

Enroll the real voices, once

Each authorised person reads a few short phrases on the phone. We compute a mathematical fingerprint of their voice. New voice-AI models coming out every week do not change what your real people sound like, so this never expires.

02

Every call runs three checks

Is the voice live and human, not a clone? Is it one of the enrolled people? Does the request make sense against policy? Three independent signals, so a clone that beats one is caught by the next.

03

Green, amber, red

Verified and in-policy passes through, logged. Anything unusual is held, escalated to a human, with the full context of why. Deny by default, allow the verified.

04

Every call sealed

Whatever the verdict, a tamper-proof record is written: time, model version, scores, decision. Your proof that the line was protected.

Honest limit: a perfect clone of a real person can pass voice-matching, because it sounds like them. That is why Shield never relies on one check. The liveness check catches the clone, the match check catches the human impostor, the context check catches the social-engineering script.

Pricing

Priced per protected line, so you spend money only where the loss would be catastrophic: the payment desk, the CFO's direct number, the dealing line.

Per protected line
£250-500
per line, per month
  • All three real-time checks on every call
  • Verified Caller List enrollment included
  • Green / amber / red verdict with escalation
  • Sealed evidence record for every call
Typical deployment
3-8 lines
the lines where money moves
  • Payments / treasury desk
  • Executive and partner direct lines
  • Client dealing lines
  • Volume pricing above 10 lines
Why per-line works here and mass detection does not: on a low-volume, high-value line a false alarm costs a real caller twenty seconds with an AI receptionist, not a bank an escalation storm. The economics only make sense on the lines that matter, which is exactly where Shield lives.
Protect what moves money

Put a guard on the line a clone would target.

Most firms start with Assess to find the weak line, then Shield to close it.

Request a pilotStart with an assessment
PRODUCT 01 · RED TEAM
Assess
Find out which line and people are exposed.
PRODUCT 03 · CARRIER
Network
For phone companies screening outbound calls.
PRODUCT 04 · API
Detect
The raw sensor for developers and platforms.