Conversation agents, voice and text, built for the long conversation | yaab
yaab
AI Agents · Conversation

Agents that can hold a real conversation.
A long, structured one.

Most conversational AI demos run a handful of turns. We build voice and text agents that hold long, structured conversations, stay on mission the whole time, adapt to the person in front of them, and end with a decision-grade deliverable backed by verbatim evidence. In production, at scale, on infrastructure we designed for exactly this.

Why this is hard

In a long conversation, everything that "just works" in a demo breaks.

The agent forgets its mission

Without explicit state, topics repeat, topics get skipped, and the agent cannot say what is covered or missing.

Loops appear

The most reported failure in long AI conversations: the same question asked again and again. Real users count out loud and abandon.

Latency and cost compound

A long conversation re-sends its history every turn. Without caching architecture, cost grows until the conversation dies.

Voice adds its own failure modes

Turn-taking (talking over people), recognition noise, dropped calls that must resume without losing the conversation, and a strict latency budget.

The output is the product

In high-stakes use, the conversation exists to produce a decision document that cites what the person actually said, verbatim, and never contains anything they did not say.

We solved each of these in production.

One brain, two channels

Voice or text, the reasoning stack is identical. Voice adds a real-time layer (turn-taking, latency budgets, resumed calls) that we tune separately. You choose the channel per use case; the agent's judgment does not change.

Architecture and scale

A conversation is a system, not a prompt.

Every piece exists because a long conversation breaks without it.

01
Separation of voice and reasoning

The voice layer (recognition, synthesis, turn-taking) runs on a dedicated real-time platform; reasoning runs on a frontier LLM behind a server we control. Each tunes independently; either swaps without rebuilding the product.

02
Caching architecture

The agent's expertise is compiled and cached in layers, so latency and marginal cost stay near-flat across a full-length session. This is what makes the unit economics viable at scale.

03
A real state machine, not vibes

Every session carries explicit state: current phase and budget, what has been covered and to what depth, attempts per topic, topics closed by refusal, pending angles, records extracted. The model executes turns; the mission lives in deterministic state.

04
An orchestration layer that owns the trajectory

Turn by turn it decides where the conversation goes next: which topic is in focus, when a topic is exhausted, when depth should match relevance, and when the conversation has earned the right to close.

05
Per-turn analysis in parallel

Specialized workers analyze each exchange as it happens and steer the next turn: coverage enforcement, claims without evidence, rehearsed answers, contradictions, vague numbers. Small, cheap, independently testable.

06
Gated closing

The agent cannot decide it is done. A coverage audit reviews the conversation against the mission checklist and either lets it wrap up or steers it back to what is missing.

Built to scale

The same architecture runs one conversation or thousands: model tiering (the frontier model holds the conversation; small fast models do the auxiliary analysis), caching that keeps marginal cost near-flat, lifecycle tooling (invitations, batch monitoring, live session state, re-invite flows), and safe iteration in production (knowledge updates live without restarts, code behind flags, staged rollouts). Scaling a conversation agent is an economics and operations problem; this architecture solves both.

Conducting quality

Conducting skills, encoded as hard rules.

A long conversation is a professional skill. We encode it as rules, each born from a real failure and each protected by tests.

Never loop

If the person already answered or declined, the agent may rephrase once, then closes the topic. The state machine counts re-asks, so the "third time you ask me" failure is impossible by construction.

Never fabricate a recap

The agent only summarizes what the person actually said, in their words. Cut-off turns trigger a "please continue", never a guessed completion.

Neutral acknowledgment, zero flattery

No cheerleading, no leading the witness. Praise contaminates evidence.

No rescue

When an answer is weak, the agent does not complete it charitably. It probes: reformulate, ask for a concrete case, escalate for hard numbers, then note and move on.

Adaptive depth

Recent and relevant material explored deeply, older material summarized, and integrity data always collected, no exceptions.

Sensitive topics with framing

Personal questions carry a normalizing preface, are optional, and have a hard cap. Protected topics are prohibited outright.

Handling the difficult person

Encoded moves for the rambler, the evasive, the stonewaller, and the collective answerer (redirect "we did" to "what did YOU do").

Multi-session continuity

A dropped call resumed later hydrates state and history; the agent references prior statements correctly and confronts cross-session contradictions politely, with verbatim quotes.

The know-how compounds. Every engagement generates domain-specific know-how: the question banks, follow-up patterns, red-flag catalogs and conducting rules for YOUR conversation, encoded, versioned and protected by tests. That know-how is what the Neocortex is made of: for interviews we built a full methodology this way, and new conversation domains start from it instead of from zero.

Voice and the deliverable

Voice is hard. The deliverable is the point.

The voice layer
  • Turn-taking, diagnosed and tuned. We diagnosed and tuned the exact failures that make agents talk over people, and pair configuration fixes with prompt-level mitigations so a cut-off turn degrades gracefully.
  • Latency budgets. Streaming everywhere, cache-friendly prompt layout, auxiliary analysis off the critical path.
  • Resumed sessions, whole. Dropped calls resume without losing what came before, and the final record is reconstructed complete.
  • Transcript hygiene. Real transcripts arrive polluted; we sanitize before anything downstream evaluates them.
  • Voice-specific writing. Short sentences, spoken rhythm, no lists read aloud, numbers spoken naturally.
The deliverable
  • Structured extraction into your domain schema: whatever the decision downstream needs.
  • Verbatim citation discipline. Every score and claim carries quotes validated automatically against the transcript. Fabricated evidence filters itself out.
  • Deterministic decisions. The final classification is recomputed from hard extracted data, so the same evidence always yields the same verdict. The engine only lowers, never raises.
  • Reports built for human reviewers. The exact question and answer, confidence that reflects evidence quality, flags for review, and embedded audio to hear the exact moment behind any score.
  • Honest uncertainty. Truncated conversations produce explicit "insufficient evidence" outcomes instead of invented scores.
Quality that does not regress

Every conducting rule and scoring criterion is anchored by automated tests that run on every change: rules cannot be silently deleted, frozen verdicts cannot silently move, and a cast of synthetic personas (the fabricator, the concealer, the rambler, the honest-but-slow control) exercises the agent in full simulated conversations before releases. Expert feedback is verified against transcripts before it becomes a rule. This is how the agent that worked last month still works after every change.

Use cases

One architecture. Any conversation that matters.

What changes per domain is the methodology we encode, and encoding it is part of what we deliver.

Production flagship
Interviews & talent evaluation

Screening and deep interviews with methodology-grade rigor, in production for a high-volume hiring operation: real interviews by voice at volume, multi-stage, with evidence-cited scorecards. We built the interviewing methodology itself, encoded and protected by tests.

Clinical & intake interviews

Structured histories where completeness and verbatim fidelity are mandatory: every topic covered to protocol depth, sensitive topics framed correctly, a record that quotes the patient.

Discovery & requirements calls

Extracts a complete, structured brief from a rambling stakeholder conversation: adaptive depth, the collective "we" redirected to specifics, a deliverable your team can build against.

Compliance interviews & audits

Consistent scripts across every interviewee, no leading questions, evidence trails with verbatim citations, honest recording of refusals. The same conversation, every time.

Insurance claims & financial onboarding

Long fact-gathering with integrity checks: contradiction detection within and across sessions, escalation for hard numbers, structured records that feed the decision system downstream.

Coaching & assessment

Behavior-anchored evaluation with evidence, not vibes: proportional depth, no rescue of weak answers, reports the person can actually learn from.

Complex support

Multi-topic, multi-session support conversations that keep state: what was tried, what was promised, what remains open, with continuity when the conversation resumes days later.

Why us

We are not assembling a chatbot from a template. We run this architecture in production, we have the infrastructure to build and scale it, and every engagement leaves you with the encoded methodology of your own conversation, versioned and protected by tests.

Proof

Long, structured voice interviews with real candidates, in production, for a client's high-volume hiring operation, in Spanish with English variants.

Every conversation lands as an evidence-cited, auditable deliverable.

The conducting rules and scoring are protected by an automated regression suite, so the agent that worked last month still works after every change.

Born with a brain

Every conversation agent runs on its own mini brain, the live context and encoded methodology it needs to conduct well. Complete on its own, and already a module of your Company Brain.

Learn about the Brain

Powered by the Neocortex. The intelligence layer that ships with every agent we build. It arrives knowing the job and keeps getting smarter with every deployment.

Meet the Neocortex

Bring us a conversation worth automating.

In one working session we map the conversation, the deliverable it must produce, and what it takes to run it at your volume.