# Call1 Features | On-Prem Contact Center QA Software > Detailed Call1 capabilities for contact center RFPs: ingest, transcription, hybrid sentiment, QA scoring, review workflows, security, and model governance. Source: https://call1.cc/features --- [CALL 1](https://call1.cc/) Coverage Evidence Review queue Deployment RFP reference [Pricing](https://call1.cc/contact) [Talk to us](https://call1.cc/contact) Features # Contact center quality assurance, in the detail *your evaluation needs.* Call1 scores every eligible conversation, links each verdict to evidence, and keeps call data inside your trust boundary. Your reviewers keep the decisions that matter; Call1 transcribes and scores the rest. Six capabilities below, each with a technical article behind it. [Talk to us](https://call1.cc/contact) Jump to the RFP capability reference 01 **Total coverage** 02 **Hybrid sentiment** 03 **Evidence-linked scoring** 04 **Smart review queue** 05 **Private deployment** 06 **Versioned intelligence** 01 / Total coverage ## Score every eligible call. Call1 runs your scorecard across every eligible call, within the capacity you bought. The disclosure failure that shows up once in five hundred calls stops depending on sampling luck, and coaching patterns surface from the whole floor, not from whoever got sampled. Your recorder writes post-call audio straight to an S3-compatible endpoint on the box. Call1 processes the recording locally, builds the transcript and review artifacts, then routes the audio to the archive you designate. Your archive stays the system of record for recordings. [Read the technical article](https://call1.cc/features/total-call-coverage) One controlled path 1. **Recorder writes in** S3-compatible, on the box 2. **Local processing** Transcript, signals, scores 3. **Your archive** Audio continues where you choose 02 / Hybrid sentiment ## Call1 scores the words and the voice separately, then fuses them. Every eligible, speaker-aligned chunk can produce two independent affect signals: textual sentiment from what was said and tonal emotion from how it was said. A policy-versioned fusion step turns them into one useful hybrid result. Both source signals stay attached to the result. If the language reads neutral while the voice carries strain, your reviewer sees the disagreement. And when audio quality is too poor for reliable tonal inference, the tone signal is marked unavailable. [Read the technical article](https://call1.cc/features/hybrid-sentiment) Chunk 0042 / customer TEXT **Neutral** “No, that's fine.” TONE **Strained** Elevated tension detected FUSED PER CHUNK HYBRID RESULT **Frustration underneath agreement** Source disagreement is attached as evidence. 03 / Evidence-linked scoring ## Every verdict carries the evidence behind it. Each QA verdict points back to the moment that produced it. Your reviewer sees the transcript excerpt, timestamp, rationale, confidence, rubric version, and model lineage together. Machine output is immutable. When a human disagrees, Call1 records an adjudication beside the verdict. The audit trail shows what the system decided, what the reviewer decided, and why. [Read the technical article](https://call1.cc/features/evidence-linked-scoring) REVIEW SIGNAL 03:14 **82/100** ### Required disclosure likely missing. > “I'll pull up your plan and get that lowered right now…” 04 / Smart review queue ## The queue turns total coverage into focused human work. Scoring every call should not create a larger pile for reviewers. Call1 ranks the conversations and moments that deserve attention: compliance failures, low-confidence decisions, emotional escalations, rare events, and unusual changes in performance. Reviewers open a prepared case with audio, transcript, evidence, and the relevant scorecard question already in context. They confirm or override with a reason code and move on. Reviewers pull the next call; nobody cherry-picks an easy one. Under-buy your capacity block and the queue clears slower. Nothing goes unscored, but the wait grows until you add a block. [Read the technical article](https://call1.cc/features/smart-review-queue) Today / needs attention 1. **Disclosure failure** CALL-4821 · confidence 94% 03:14 2. **Escalation pattern** CALL-4798 · hybrid sentiment 08:42 3. **Low-confidence verdict** CALL-4830 · human check requested 01:57 05 / Private deployment ## Keep recordings inside your trust boundary. Call1 ships as an appliance that racks inside your network, under your SSO, your firewall rules, and your keys. The application, the models, the database, and the object store live together on the box. Call1 the company holds no standing access to call content. When the system needs attention, support access is a grant you approve: specific, time-boxed, and audited. - No cloud AI dependency - No standing vendor content access - Operational telemetry without call content [Read the technical article](https://call1.cc/features/private-deployment) Your trust boundary **CALL [1]** App · models · scores · audit trail YOUR AUDIO YOUR ARCHIVE CALL1 COMPANY **0 standing access** 06 / Versioned intelligence ## Change models without rewriting history. Speech, text, tone, embedding, and scoring models will improve. Call1 treats each one as a replaceable, versioned component behind a stable per-chunk data contract, so the product can adopt better models without rebuilding the review experience. New candidates can run beside the current model and compete on the same calls. Champion and challenger outputs keep their own confidence, version, and provenance. A swap changes future processing and leaves the historical record as it was. [Read the technical article](https://call1.cc/features/versioned-intelligence) Model registry / live comparison CHAMPION **Tonal model v4.2** Production CHALLENGER A **Tonal model v5.0-rc** Comparing CHALLENGER B **Multimodal fusion v2** Shadow Historical results remain pinned to their original versions. Technical capability reference ## Language your team can carry into an RFP. The middle column describes Call1's architectural response. The final column names the details that should become explicit, testable requirements in a proposal, statement of work, and acceptance plan. Formal responses should mark each requirement as included, configurable, optional, or roadmap, and attach the deployment-specific evidence. | Capability area | Call1 architectural response | Confirm in the proposal | | --- | --- | --- | | **Coverage and ingest** | An S3-compatible endpoint on the appliance accepts post-call recordings directly from your recorder, inside the deployment boundary. Audio can enter scoring without metadata; filename parsing, JSON/CSV sidecars, or scoped platform adapters enrich the canonical session asynchronously. Call ID correlation and idempotent re-drops prevent duplicate processing. | Source platform, audio formats, channel layout, daily call-hours, burst profile, metadata fields, and network path. | | **Transcription and audio processing** | Transcoding, speaker alignment, transcription, word timestamps, PII/PAN detection, transcript tokenization, and configured post-call redaction run locally. Processed audio can be encrypted and routed to customer-controlled object storage, NAS, or local storage; Call1 does not require a vendor-hosted inference service. | Languages, codecs, accent mix, transcription acceptance criteria, redaction categories, target latency, and archive destination. | | **QA scoring and rubrics** | Every rubric question is routed independently through a derived deterministic, adapter, or larger-model tier. Each result carries verdict, confidence, rationale, evidence span, scoring tier, rubric version, and model version. Conditional questions can resolve not-applicable; sections, weights, and automatic-fail rules support customer scorecards. | Rubric inventory, required evidence, thresholds, validation sample, section rollups, auto-fail policy, and legal/compliance ownership. | | **Sentiment, emotion, and audio events** | Eligible speaker-aligned chunks retain separate textual sentiment and vocal-tone outputs plus a policy-versioned hybrid result. Audio usability, confidence, model lineage, and text/tone disagreement remain attached. The same chunk contract can carry silence, interruption, crosstalk, hold music, static, DTMF, or other configured acoustic events. | Model set, label taxonomy, enabled speakers, chunking policy, audio-quality thresholds, event taxonomy, and human-review triggers. | | **Human review workbench** | A ranked queue prioritizes compliance failures, low-confidence verdicts, emotional escalation, off-script behavior, and customer-defined risk. Reviewers work from synchronized audio, waveform, transcript, evidence, and question-level verdicts; confirm and override actions create append-only adjudications with reason codes. | Reviewer volumes, assignment model, saved queues, escalation flow, reason codes, calibration process, and desktop/browser standards. | | **Identity and authorization** | The workbench separates reviewer, QA lead, and administrator responsibilities inside the customer's identity boundary. Remote access runs over a path the customer operates (their VPN, or a Cloudflare tunnel in their own account); Call1 does not require a vendor-hosted user account or standing employee account in the workbench. | OIDC/SAML requirement, identity provider, MFA policy, group-to-role mapping, session controls, deprovisioning, and privileged-access workflow. | | **Reporting, audit, and insights** | Coverage, question performance, agent scorecards, adjudication history, and compliance exports resolve back to the calls and evidence that produced them. Machine results, rubric versions, human decisions, and audit events remain distinguishable. Optional aggregate analysis supports drivers, tracked patterns, sentiment trajectories, and evidence-linked trend detection. | Required reports, filters, reporting periods, coaching hierarchy, audit packet format, insight modules, and downstream BI consumers. | | **Data lifecycle and export** | Call1 is not the long-term recording system of record. Customer policy controls working-audio retention and routing; durable transcripts, embeddings, scores, evidence references, and adjudications remain within the customer boundary. A stable canonical model supports config-driven map-in and map-out to files, SQL views, and outbound webhooks without exporting audio by default. | Retention windows, legal holds, deletion requirements, backup/restore, customer KMS, target schemas, permitted transcript fields, and export cadence. | | **Security and vendor access** | Processing and storage remain inside the customer-defined trust boundary. Call1 has zero standing access and requires zero inbound public ports; keys for routed audio stay with the customer, operational telemetry excludes content, and break-glass support is customer-approved, time-boxed, and audited. | Network zones, firewall rules, key custody, log retention, vulnerability-management requirements, break-glass approval, security questionnaire, and required attestations. | | **Deployment and operations** | One validated appliance configuration, specced and shipped by Call1, benchmarked to a stated call-hours/day. Offline images, locally validated licenses, and signed update bundles support restricted or air-gapped environments. Running the same deployment on customer-owned infrastructure is on the roadmap, not sold today. | Host OS, container runtime, GPU/VRAM, CPU/RAM/storage, high availability, recovery objectives, patch cadence, support hours, and air-gap procedure. | | **Model governance** | Models and fusion policies are registered by version. Champion and challenger models can evaluate the same calls in production and shadow modes without overwriting historical output. Promotion changes future processing only; previous results remain pinned to the original model, policy, and rubric versions. | Acceptance metrics, approval authority, benchmark corpus, drift checks, retraining schedule, rollback procedure, and residency rules for customer-trained adapters. | | **Capacity and commercial boundary** | The subscription ties to validated deployment capacity, not per-token or per-analysis consumption. The customer keeps control of long-term storage, so recording retention doesn't become a variable Call1 infrastructure charge. | Validated call-hours/day, concurrency, enabled analysis modules, number of rubric questions, support tier, appliance option, and contract term. | Acceptance language ### Six things to pin down before you sign. 1. **Performance benchmark** Define eligible call-hours/day, peak arrival rate, maximum processing lag, concurrency, test corpus, and the hardware configuration used to prove them. 2. **Accuracy and human-review policy** Define the labeled acceptance set per rubric question, precision/recall or error tolerance, confidence thresholds, mandatory-review cases, and the owner of final approval. 3. **Security evidence** Require a deployment diagram, data-flow diagram, port list, identity flow, key-custody statement, telemetry inventory, break-glass procedure, and sample audit record. 4. **Data ownership and exit** Name the system of record, retention and deletion rules, backup owner, export schemas, model/adapter residency, and the artifacts returned or destroyed at termination. 5. **Operational responsibility** Assign monitoring, patching, backups, key rotation, incident response, disaster recovery, support access, and hardware replacement between Call1 and the customer. 6. **Commercial boundary** State the validated capacity band, included analysis modules, rubric-question limits, support tier, hardware line items, renewal treatment, and any condition that can create overage. ## Bring us your requirements. Share your call volume, environment, rubric, security questionnaire, and acceptance criteria. We'll map the deployment and say what is included, what is configurable, and what needs to be scoped. [Talk to us](https://call1.cc/contact) [CALL 1](https://call1.cc/) On-prem contact center quality assurance. [Pricing](https://call1.cc/contact) [hello@call1.cc](mailto:hello@call1.cc)