The technology behind a voice 94% of people couldn’t tell was AI.

Human Staff Mirroring is the category of AI receptionists Futuro pioneered, in which the agent carries out the work of a role rather than only answering the phone. The distinction is operational: a Futuro agent books the appointment, updates the CRM, takes the payment and sends the follow-up inside the same call, across 150+ integrated tools. An AI receptionist answers and transfers. The line is whether the call ends with the job done or with a message taken.

VoiceAlive is Futuro's voice engine. It deliberately engineers in the imperfections of human speech — audible breaths, strategic micro-pauses, controlled disfluencies, formality adaptation and regional accents — on the finding that removing those signals is what makes synthetic speech detectable. The first build tested at 41% indistinguishability; the current version tested at 94% in a 1,000-participant double-blind study.

MasterMind, the knowledge system, holds a per-business knowledge graph built from that company's own handbooks, SOPs, catalogs and live systems, and answers only from within it. Constrained dynamic knowledge boundaries stop the agent inventing an answer it does not have; when it does not know, it says so and escalates. Predictive retrieval stages the answer before the caller finishes speaking, so grounding does not cost an audible lookup pause.

Because it is reading. When a number that has called before dials in, the agent loads the last ten conversation summaries during the ring window — what was said, what was decided, what was left open — so it is prepared before it speaks rather than searching while the caller waits. The three rings are spent loading, not waiting.

Each call is transcribed, stored against the agent and the caller's number, and reduced to a summary. Weekly, ten of those summaries are compressed into a single rolling memory. On a later call the agent retrieves recent summaries in full detail alongside the compressed rolling memories, so a caller on their hundredth call still carries the context of their third without the agent having to read a hundred transcripts.

1,000 participants took part in real customer-service interactions, having been told they were rating a customer-service experience — no AI context was given. Afterwards they answered one question: whether there was any possibility the representative had been artificial intelligence rather than a human. 94% said no. It was a double-blind design: neither the callers nor the administrators knew which calls were AI. The full methodology is published in the study breakdown.

Not yet, and we do not claim it is. The study was conducted by Futuro and reviewed by one outside company, which is not the same thing as independent verification. A peer-reviewed study with Florida Atlantic University, with video recording and faculty involvement in the process, is planned for approximately three months' time. Until that publishes, treat 94% as our own measurement rather than as an externally validated one.

Three years. Futuro was founded on 9 October 2021 and spent until 2024 on research and development — conversational neuroscience study, then VoiceAlive, then MasterMind, then the double-blind study — without selling or advertising a product. Development did not stop in 2024; what stopped was doing it with nothing to sell.