The AI Agent QA platform built around the outcomes you measure, not just the scores you generate. Three contractual commitments. One average outcome our customers see documented in 120 days.
If you read nothing else, read this. Each beat is the answer to a question your procurement team will ask.
CMP Research maintains separate Prisms for Auto QA/QM and Customer Analytics because buyers shop them as distinct decisions. Most vendors are in one. We compete in both, on the same architecture.
Of 21 vendors surveyed in May 2026, zero publish any of these three. Remediation if accuracy slips. Money-back if deployment misses. Exit without penalty.
We do not build AI agents. We score whichever platforms you run, plus your human agents, on one scorecard.
Etech founded 2003. Three countries, seven sites, zero mergers or acquisitions. We are the first customer of every release.
The Fortune-500 anchor's $5.3M (of $6.5M) came from Layers 2 to 6. We measure all six on the same interaction data.
Documented against your own KPIs by your finance, ops, and quality leadership. Not in the MSA; a measured outcome across deployments.
Customer experience leaders and quality leaders shop the category with different questions. Both kinds of pain live inside the same conversation.
CMP maintains separate Prisms for Auto QA and Customer Analytics because buyers shop them as distinct decisions. QEval® competes credibly in both.
Three commitments live in the master services agreement with remediation and exit terms. The fourth is what most customers see in practice, measured against their own KPIs.
Contractual floor across every scoring dimension. Remediation if it slips.
Production scoring inside 30 days. Money-back if we miss.
Leave with 60 days' notice. No penalty. Full data portability.
What most customers see on average. Measured against your KPIs, reviewed by your finance, ops, and quality leads.
Outcome ranges documented across QEval® production deployments. Measured against the customer's own KPIs.
The Fortune-500 anchor's documented $6.5M annual outcome, broken down by Six Layers attribution.
| Layer | Documented outcome | Value |
|---|---|---|
| L1 Quality + Compliance | 32,871 manual audits replaced; 11 FTEs retasked; 85% drop in compliance violations across 5 brands | $1.2M |
| L2 Customer Intelligence | +21.5pp CSAT recovery across 5 brands, 59% to 80% arc | $1.2M |
| L3 Revenue Intelligence | Resolution roughly 60% faster; retention recovery on at-risk accounts surfaced by sentiment trajectory | $0.3M |
| L4 Operational Intelligence | 6,000 transfers cut; ~45 FTE-equivalent capacity recovered | $3.0M |
| L5 Training Intelligence | +13pp QA score across 5 brands; +300% coaching frequency lift; impact-ranked skill-gap queue | $0.3M |
| L6 Strategic Intelligence | Cross-LOB executive dashboards; portfolio compliance risk reduction across 5 brands | $0.5M |
| Total realized | 5 brands, 1,200+ agents, 6 months | $6.5M |
Three blocks. Three columns. No vendor names. Cells read "Not published" because the rows below are factually defensible against publicly available documentation.
| Dimension | QEval® | Standalone QA Vendors | CCaaS-Native QA |
|---|---|---|---|
| Block A · Contractual commitments | |||
| Classification accuracy SLA | ✓94%+ contractually | Not published | Not published |
| Deployment guarantee | ✓30 days, money-back | Estimate only | Bundled with CCaaS rollout |
| Exit rights | ✓60 days, no penalty | 12 to 36 month lock-in | Bundled lock-in |
| Block B · Capability coverage | |||
| Foundation AI model | Proprietary closed-source MoE | Generic LLM wrapper | Vendor-bundled |
| Scores AI agents from other vendors | ✓Vendor-neutral | Conflict of interest | CCaaS-native only |
| Same scorecard, human + AI agents | ✓One scorecard | Varies | Separate flows |
| PHI / PII redaction sequencing | ✓Pre-LLM, at ingest | Post-redaction | Varies |
| Channels (voice, chat, email, SMS) | ✓All four, native | Voice + chat at best | CCaaS-native only |
| Six Layers of Intelligence attribution | ✓L1 through L6 | No equivalent | Layer 1 only |
| Languages, same scorecard | ✓35+ languages | English-first | Varies by tenant |
| Coaching loop on same data model | ✓HI Model lifecycle, native | Separate vendor or manual | Add-on SKU |
| Next Best Action with predicted impact | ✓Auto-recommended per gap, with confidence | Not published | Limited or absent |
| Gamification engine | ✓Native: leaderboards, peer recognition, challenges | Separate product | Bundled or absent |
| Goal setting for agents | ✓Individual plans tracked against scored skills | Not published | Not published |
| Agentic QA operators (dispatchable workflows) | ✓8 operators: reporting, audit, compliance, coaching, trend, dashboard, calibration, analytics | Fixed 3-bot trio at best | Not published |
| Block C · Architecture and proof | |||
| Built and validated inside | ✓A 4,000-agent live operation | Software lab | Cloud-platform engineering |
| Data sovereignty | ✓No third-party foundation model | Often unspecified | Per CCaaS T&Cs |
| Third-party recognition | ✓CMP both Prisms + ICMI 2025 | Mixed | Embedded in suite ratings |
Most QA platforms stop at the score. QEval® is built around the full loop: same data model, same scoring engine, every step on one architecture.
MoE expert sub-models score every interaction at 94%+ accuracy
Next Best Action surfaces with predicted impact and confidence
One-Click Coaching auto-fires with the right module + acknowledgment
Behavior change tracked against the scored baseline
Six Layers attribution rolls outcomes up to L1 to L6 business value
Each card is an anonymized deployment with a measured outcome.
Replaced a keyword QA engine. $5.3M (82%) from Layers 2-6. +21.5pp CSAT, 85% compliance violation drop, +13pp QA score.
Confusion 42% to 28%. Used trajectory data to redesign IVR routing.
Churn-risk signal surfaced before cancellation calls. Negative sentiment -38%.
RTAA surfacing the right knowledge article inside the conversation.
Four questions every AI vendor should be able to answer in writing. Including us.
It is the contractual SLA. Remediation obligation if QEval® falls below the floor, exit right if remediation does not restore it. The three contractual numbers (94%+ accuracy / 30-day deployment / 60-day exit) live in the master services agreement, with sample clauses shown above. The 120-day ROI window is a measured customer-average outcome, not a contractual provision.
Industry-average deployment for enterprise QA and CCaaS-bundled QA runs 90 to 180 days. The 30-day commitment is contractual, with money-back if QEval® misses the window. Of 21 vendors surveyed, none publish a comparable timeline.
CCaaS-native QA cannot score interactions from competitor CCaaS platforms, cannot apply one scorecard to human and AI agents from other vendors, and does not produce the Six Layers of intelligence beyond Layer 1. CCaaS QA is also bundled as a feature, which is why none of them publish accuracy SLAs.
Yes. The relevant clauses (the three commitments shown above) are shared with your procurement team on request before any NDA. Full MSA shared under mutual NDA, red-line markup welcomed.
30 minutes. Your scorecard, your AI agents, your CCaaS. We will score the call and walk through what QEval® saw, layer by layer.