# Lewis Crook

> Practitioner-grade analysis of enterprise voice AI and conversational AI for contact centres — from someone who has been the buyer, builder, and in the room.

## Independence and affiliation

I work as Lead Solutions Engineer, UK & Ireland at Parloa. This site is my own — Parloa does not sponsor, fund, commission, or review anything published here.
The frameworks and evaluation criteria are vendor-neutral by design. No vendor (including Parloa) is named, ranked, recommended, or criticised in any guide, glossary entry, framework, or note. Views are my own.
If you spot anything that reads as vendor-favourable or vendor-hostile, that is a content bug — please flag it via the corrections page.

## Pillar guides

- [Voice AI vs legacy IVR: the honest unit economics](/guides/voice-ai-vs-ivr-unit-economics): Honest unit economics for voice AI vs legacy IVR — what call-centre automation actually costs, and where the savings really come from.
- [How to evaluate enterprise voice AI platforms: a vendor-neutral framework](/guides/how-to-evaluate-voice-ai-platforms): A vendor-neutral framework for comparing enterprise voice AI platforms across nine dimensions: integration, latency, evaluation, security.
- [Voice AI containment rate: what's real vs what vendors claim](/guides/voice-ai-containment-rate-reality): How voice AI containment rate is measured, where vendor claims diverge from production, and what a defensible enterprise number looks like.
- [Why enterprise voice AI pilots fail to reach production](/guides/why-enterprise-voice-ai-pilots-fail): The five repeating reasons enterprise voice AI pilots stall before production — integration, operating model, measurement, intent, procurement.
- [Who maintains a voice AI after go-live? The operating-model question](/guides/voice-ai-operating-model): Voice AI is not deploy-and-forget. The operating model required to keep an enterprise voice AI useful, accurate, and economic after go-live.
- [Enterprise voice AI integration depth: a real evaluation checklist](/guides/voice-ai-integration-depth-checklist): A practical voice AI integration checklist — the read/write, authentication, and edge-case capabilities that decide enterprise success.
- [Call deflection with AI: where it works and where it backfires](/guides/call-deflection-ai): A vendor-neutral look at AI call deflection — which intents resolve, which generate re-contact, and how to measure the difference honestly.
- [Customer service automation: an honest guide for enterprise CX leaders](/guides/customer-service-automation): Vendor-neutral analysis of enterprise customer-service automation — what to automate, what to leave alone, and how to measure it honestly.
- [Conversational AI vs voice AI: what's the actual difference?](/guides/conversational-ai-vs-voice-ai): Conversational AI is the broader category; voice AI is the spoken-telephony subset. A vendor-neutral comparison of the two for CX leaders.
- [Voice AI security and compliance: the enterprise buyer's checklist](/guides/voice-ai-security-compliance): A vendor-neutral checklist for enterprise voice AI security and compliance: PCI, HIPAA, GDPR, data residency, recording, and vendor risk.
- [Conversational IVR / IVR replacement: the phased migration playbook](/guides/ivr-replacement-playbook): Vendor-neutral playbook for IVR replacement and conversational IVR migration — intent triage, parallel running, fallback design, and the cutover gates that protect CX.
- [Voice AI pricing models: per-minute, per-resolution, and platform compared](/guides/voice-ai-pricing-models): A vendor-neutral comparison of voice AI pricing models — per-minute, per-resolution, and platform — and which one transfers containment risk to the vendor.
- [Call deflection benchmarks: realistic 2026 numbers by intent and channel](/guides/call-deflection-benchmarks): Realistic 2026 call deflection benchmarks by intent and channel — gross vs net rates, the gaps that flip the business case, and how to baseline honestly.
- [Voice AI RFP template: what to actually ask, and how to score the answers](/guides/voice-ai-rfp-template): A vendor-neutral voice AI RFP template — the 42 questions worth asking, the answers that disqualify, and a weighted scoring sheet procurement can defend.
- [Voice AI security questionnaire: the questions IT-Sec actually needs answered](/guides/voice-ai-security-questionnaire): A vendor-neutral security questionnaire for voice AI procurement — data flow, sub-processors, model training, DPA, incident response, and exit. The answers IT-Sec needs in writing.
- [Voice AI first 90 days: a week-by-week post-launch operating plan](/guides/voice-ai-first-90-days): A week-by-week post-launch plan for the first 90 days of a voice AI deployment — what to baseline, what to ship, and what the board pack looks like at day 90.
- [Voice AI QA rubric: a call-review template the operating model can actually run](/guides/voice-ai-qa-rubric): A vendor-neutral QA rubric for voice AI call review — eight dimensions, scoring guidance, sampling strategy, and the rubric that ships with the operating model.
- [Voice AI to live-agent handoff: the patterns that survive production](/guides/voice-ai-live-agent-handoff-patterns): Five handoff patterns from voice AI to live agent — when each is appropriate, what context to carry, and how to measure handoff quality in production.
- [Voice AI latency budget: where the milliseconds actually go](/guides/voice-ai-latency-budget): An end-to-end latency budget for production voice AI — ASR, retrieval, LLM, TTS, and telephony — with p95 targets and the diagnostic questions for each step.
- [Voice AI kill criteria: when to stop a pilot, in writing, before it starts](/guides/voice-ai-kill-criteria): Pre-committed kill criteria for a voice AI pilot — five binary gates, named decision-makers, and the cost of letting a bad pilot persist.
- [Voice AI board pack: the one-page template for the steering committee](/guides/voice-ai-board-pack): A one-page board-pack template for a voice AI programme — five lines that hold attention longer than a thirty-tile dashboard, with the questions sponsors ask.
- [Voice AI for FCA-regulated contact centres: a Consumer Duty compliance checklist](/guides/voice-ai-fca-consumer-duty-compliance): UK-specific compliance checklist for deploying voice AI in FCA-regulated contact centres — Consumer Duty, SYSC 10A call recording, ICO/UK GDPR, SM&CR accountability.
- [Agentic voice AI in the enterprise: what's real in 2026](/guides/agentic-voice-ai-enterprise): What 'agentic' voice AI actually means in enterprise contact centres, where it works in 2026, where it doesn't, and how to evaluate it without falling for the demo.
- [Enterprise voice AI vendor comparison: 2026 buyer's guide](/guides/voice-ai-vendor-comparison-2026): How to compare enterprise voice AI vendors in 2026 — the nine VERA dimensions, the categories that matter, and the questions to ask before shortlisting.
- [Legacy IVR replacement: migrating off Nuance-era platforms to modern voice AI](/guides/nuance-ivr-replacement-migration): A phased migration playbook for replacing legacy Nuance-era IVR platforms with modern voice AI — what to lift, what to rebuild, and how to de-risk the cutover.
- [Voice AI DPIA template: a working data protection impact assessment](/guides/voice-ai-dpia-template): A working DPIA template for enterprise voice AI under UK GDPR and EU GDPR — the sections regulators actually read and the answers procurement needs evidenced.
- [EU AI Act voice AI classification: limited, high-risk, or out of scope?](/guides/eu-ai-act-voice-ai-classification): How the EU AI Act classifies voice AI deployments — limited-risk transparency duties, high-risk Annex III triggers, and the documentation each tier requires.
- [PCI DSS v4.0 and voice AI: keeping cardholder data out of the model](/guides/pci-dss-v4-voice-ai): How to deploy voice AI under PCI DSS v4.0 — pause-and-resume DTMF capture, scope reduction, sub-processor obligations, and the controls QSAs actually test.
- [Voice AI RACI: programme governance that survives quarter two](/guides/voice-ai-raci-programme-governance): A working RACI for enterprise voice AI programmes — decision rights across CX, IT, security, legal, and the vendor, designed to survive the predictable second-quarter pressure.
- [2026 enterprise voice AI benchmark report: framework with illustrative numbers](/guides/2026-enterprise-voice-ai-benchmark-report): Vendor-neutral 2026 benchmark framework for enterprise voice AI — containment, latency, unit economics, integration depth — with illustrative worked numbers and a defensible measurement protocol.
- [Conversational IVR: defined, compared, and where it fits in 2026](/guides/conversational-ivr-defined-and-compared): Vendor-neutral definition of conversational IVR, the four-rung automation ladder, the realistic containment ceiling, and how to evaluate it in procurement.
- [Voice AI platform pricing models in 2026: the enterprise buyer's guide](/guides/voice-ai-platform-pricing-models-2026): The five pricing models you will see in 2026 voice AI RFPs, the hidden cost stack, an honest TCO worked example, and the contract terms that move the number.
- [AI call centre software in 2026: a vendor-neutral buyer's guide](/guides/ai-call-center-software-buyers-guide-2026): A vendor-neutral buyer's guide to AI call centre software in 2026: capability tiers, integration depth, scoring rubric, and the questions that separate marketing from architecture.
- [Voicebots in the enterprise: where they fit, what they cost, and how they fail](/guides/voicebot-enterprise-guide): What a voicebot actually is in an enterprise context — use cases that deploy past pilot, realistic cost ranges, the failure modes that show up in month three, and the rubric for choosing one.

## Frameworks

- [Voice AI evaluation matrix](/frameworks/voice-ai-evaluation-matrix): nine-dimension scoring matrix with weightings and over-claim catches.
- [When the cheapest AI voice vendor answered zero patient calls: the VERA framework](/frameworks/vera-vendor-evaluation): A healthcare network picked the cheapest AI voice vendor and answered zero patient calls. The eight-domain VERA framework (Vendor Evaluation & Risk Assessment) that came out of that failure.

## Industries

- [Financial services — voice AI](/industries/financial-services-voice-ai): What enterprise voice AI looks like in financial services: regulatory constraints, realistic containment, integration depth, and the controls auditors actually ask about.
- [Healthcare — voice AI](/industries/healthcare-voice-ai): Where voice AI realistically deploys in healthcare — patient access, scheduling, refills, RCM — under HIPAA, with the controls and integration depth payers and providers actually require.
- [Insurance — voice AI](/industries/insurance-voice-ai): Where voice AI deploys in insurance — FNOL intake, policy servicing, payments — and the regulatory and operational boundaries with licensed adjusters and underwriters.
- [Telecommunications — voice AI](/industries/telco-voice-ai): Honest assessment of voice AI in telecommunications — consumer mobile and broadband, B2B account management, field-service scheduling — and the integration realities that decide containment.
- [Utilities — voice AI](/industries/utilities-voice-ai): Where voice AI deploys in utilities — meter reads, billing queries, hardship routing, outage notifications — under Ofgem, Ofwat, and EU equivalents.

## Regions

- [United Kingdom — voice AI](/regions/uk-voice-ai): What enterprise voice AI looks like in the UK in 2026: ICO, FCA Consumer Duty, vulnerable-customer obligations, and the procurement patterns that actually clear.
- [European Union — voice AI](/regions/eu-voice-ai): What enterprise voice AI looks like across the EU in 2026: AI Act classification, GDPR DPIAs, residency expectations, and country-level variation that catches RFPs out.
- [North America — voice AI](/regions/north-america-voice-ai): Enterprise voice AI in North America in 2026: state-level variation (CCPA, CIPA), disclosure obligations, HIPAA for healthcare, and the procurement patterns that ship.
- [APAC — voice AI](/regions/apac-voice-ai): Enterprise voice AI across APAC in 2026: Australia, Singapore, Japan, India residency and consent regimes, multilingual realities, and procurement patterns that travel.
- [Australia — voice AI](/regions/australia): Enterprise voice AI in Australia in 2026: Privacy Act and APP reform, APRA CPS 230, ACMA outbound rules, state recording laws, and procurement reality.
- [India — voice AI](/regions/india): Enterprise voice AI in India in 2026: DPDP Act, RBI data localisation, TRAI DLT registration for outbound, multilingual and Hinglish reality, procurement.
- [Canada — voice AI](/regions/canada): Enterprise voice AI in Canada in 2026: PIPEDA, Quebec Law 25, OSFI B-13, AIDA readiness, CRTC outbound rules, and the bilingual service obligation.

## Benchmarks

- [Voice AI containment by intent — 2026 enterprise benchmark](/benchmarks/voice-ai-containment-by-intent): Realistic 2026 voice AI containment by intent type — gross vs net of 7-day re-contact — aggregated across enterprise deployments and case studies.
- [Voice AI latency by stack configuration — 2026 benchmark](/benchmarks/voice-ai-latency-by-stack): Realistic 2026 voice AI end-to-end latency by stack configuration — STT, LLM, TTS, and integration call latency components measured in production conditions.
- [Deflection gross vs net — 2026 enterprise benchmark](/benchmarks/deflection-gross-vs-net): Realistic 2026 gap between gross and net deflection rates by channel — what finance accepts, what marketing reports, and why the difference matters.

## Use cases

- [FNOL & structured intake](/use-cases/fnol-intake): How voice AI handles first notice of loss and other structured-intake workflows: intent boundaries, integration depth, KPIs, and the cases that must reach a human.
- [Appointment & field-service scheduling](/use-cases/appointment-scheduling): Voice AI for booking, rescheduling, and field-service scheduling — integration realities, containment bands, and the slot-search trap most deployments fall into.
- [Balance & account status](/use-cases/balance-and-account-status): Voice AI for balance, status, and account-state enquiries — the highest-containment use case across most industries, and the one where measurement honesty matters most.
- [Billing & payments](/use-cases/billing-and-payments): Voice AI for billing enquiries, payments, and payment plans — PCI architecture, hardship routing, and the line between self-service and Consumer Duty.
- [Authentication & identity](/use-cases/authentication-and-identity): Voice AI for authentication, identity verification, and step-up flows — what regulators expect, how voice biometrics fits, and the design that survives a SIM-swap.
- [Outbound & proactive notifications](/use-cases/outbound-proactive-notifications): Voice AI for proactive outbound calls — outage, billing, claims status, appointment confirmations — and the consent, opt-out, and frequency-cap design that decides regulator risk.

## Use cases by industry

- [Financial services: Balance & account status](/use-cases/balance-and-account-status/financial-services): How retail banks and card issuers deploy voice AI on balance, status, and last-transaction enquiries — authentication tiers, audit logging, and the metrics that actually matter.
- [Financial services: Billing & payments](/use-cases/billing-and-payments/financial-services): Voice AI for card payments, bill pay, and transfers in retail banking — PCI architecture, PSD2 SCA, and dispute initiation that survives downstream investigation.
- [Financial services: Authentication & identity](/use-cases/authentication-and-identity/financial-services): Authentication tiers for voice AI in retail banking — voice biometrics, KBA, push-to-mobile, and the SIM-swap design that survives a fraud post-mortem.
- [Financial services: Outbound & proactive notifications](/use-cases/outbound-proactive-notifications/financial-services): Voice AI for proactive outbound in banking — fraud verification, collections, payment reminders — with TCPA / PECR consent and the Consumer Duty layer on collections.
- [Healthcare: Appointment scheduling](/use-cases/appointment-scheduling/healthcare): Voice AI for scheduling, rescheduling, and cancellation in healthcare — EHR and practice-management integration, no-show reduction, and the clinical-triage boundary.
- [Healthcare: Account & coverage status](/use-cases/balance-and-account-status/healthcare): Voice AI for eligibility, coverage, balance, and authorisation-status enquiries in healthcare — clearinghouse integration and the routing rules for benefit ambiguity.
- [Healthcare: Billing & payments](/use-cases/billing-and-payments/healthcare): Voice AI for patient statements, payments, and payment plans in healthcare — PCI architecture under HIPAA, financial-assistance routing, and good-faith-estimate handling.
- [Healthcare: Outbound & proactive notifications](/use-cases/outbound-proactive-notifications/healthcare): Voice AI for appointment reminders, prep instructions, medication adherence, and recall — under HIPAA, with the consent and accessibility design that holds up.
- [Insurance: FNOL & claims intake](/use-cases/fnol-intake/insurance): Voice AI for first notice of loss across P&C, auto, and home — structured intake, fraud signalling, and the adjuster-handoff design that wins on handle time.
- [Insurance: Billing & payments](/use-cases/billing-and-payments/insurance): Voice AI for premium statements, payments, and payment plans across P&C and life — PCI architecture, lapse-prevention design, and the hardship boundary.
- [Insurance: Authentication & identity](/use-cases/authentication-and-identity/insurance): Authentication tiers for voice AI in insurance — policyholder vs claimant vs producer, voice biometrics inside step-up, and the proxy-access design for households.
- [Insurance: Outbound & proactive notifications](/use-cases/outbound-proactive-notifications/insurance): Voice AI for proactive outbound in insurance — claims status, renewals, premium reminders, catastrophe response — with TCPA, NAIC, and the catastrophe-cadence design.
- [Telecommunications: Field-service scheduling](/use-cases/appointment-scheduling/telco): Voice AI for engineer dispatch and field-service scheduling in telco — real availability, drive-time and skill match, and the slot-search trap that kills NPS.
- [Telecommunications: Balance & service status](/use-cases/balance-and-account-status/telco): Voice AI for balance, plan, and service-status enquiries in telco — BSS / OSS integration, prepaid balance handling, and the line into account-access security.
- [Telecommunications: Billing & payments](/use-cases/billing-and-payments/telco): Voice AI for telco billing, top-ups, and payments — PCI architecture, bill explain at scale, and the collections boundary with TCPA / PECR.
- [Telecommunications: Outbound & proactive notifications](/use-cases/outbound-proactive-notifications/telco): Voice AI for proactive outbound in telco — outage notifications, planned-works comms, billing reminders — with PECR / TCPA, frequency caps, and complaint-rate management.
- [Utilities: Field & meter-visit scheduling](/use-cases/appointment-scheduling/utilities): Voice AI for engineer dispatch, meter visits, and connection appointments in utilities — real availability, vulnerability routing, and the slot-search trap that kills SLAs.
- [Utilities: Account, balance & meter-read status](/use-cases/balance-and-account-status/utilities): Voice AI for balance, tariff, meter-read submission, and account-status enquiries in utilities — CIS and MDM integration, smart-meter edge cases, and vulnerability routing.
- [Utilities: Billing, payments & payment plans](/use-cases/billing-and-payments/utilities): Voice AI for utility billing, payments, and payment plans — Consumer Duty and EU AI Act hooks on hardship, plus PCI architecture and PSR integration on disclosure.
- [Utilities: Outbound outage & service notifications](/use-cases/outbound-proactive-notifications/utilities): Voice AI for proactive outbound in utilities — outage notification, planned works, price-cap changes — under PECR / TCPA and Ofgem / Ofwat scrutiny.

## Comparisons

- [Voice AI platform categories](/comparisons/voice-ai-platform-categories): vendor-neutral shortlist by platform category.
- [Build vs buy voice AI](/comparisons/build-vs-buy-voice-ai): honest decision matrix and the three conditions that force a build.

## Glossary

- [Voice AI](/glossary/voice-ai): Voice AI is software that answers the phone, understands what the caller wants, and takes action — not just a smarter IVR.
- [Containment rate](/glossary/containment-rate): Containment rate is the percentage of calls the automation finished on its own.
- [Agentic voice](/glossary/agentic-voice): Agentic voice is voice AI that can plan and act, not just answer.
- [Autonomous resolution rate](/glossary/autonomous-resolution-rate): Autonomous resolution rate is containment rate that survives re-contact.
- [Voice AI latency](/glossary/voice-ai-latency): Voice AI latency is the gap before the system starts talking back.
- [IVR replacement](/glossary/ivr-replacement): IVR replacement swaps menus and keypad input for natural conversation and actual resolution.
- [Barge-in](/glossary/barge-in): Barge-in lets the caller interrupt the bot without breaking the conversation.
- [Turn-taking latency](/glossary/turn-taking-latency): Turn-taking latency is the awkward pause before the bot starts talking back.
- [Intent recognition](/glossary/intent-recognition): Intent recognition is figuring out what the caller actually wants.
- [DTMF fallback](/glossary/dtmf-fallback): DTMF fallback uses the keypad to capture digits the model is not allowed to hear.
- [Voice biometrics](/glossary/voice-biometrics): Voice biometrics confirms who the caller is by how they speak.
- [Real-time transcription](/glossary/real-time-transcription): Real-time transcription is streaming speech-to-text fast enough to act on mid-call.
- [Deflection rate](/glossary/deflection-rate): Deflection rate is the share of contacts the live queue did not have to handle — gross or net of re-contact.
- [First call resolution](/glossary/first-call-resolution): First call resolution is whether the customer had to call back.
- [Average handle time](/glossary/average-handle-time): Average handle time is how long each call costs you.
- [After-call work](/glossary/after-call-work): After-call work is everything the agent does after the caller hangs up.
- [Escalation rate](/glossary/escalation-rate): Escalation rate is the share of automated calls the AI gave up on.
- [End-of-turn detection](/glossary/end-of-turn-detection): End-of-turn detection is the bot deciding it is its turn to speak.
- [Hallucination rate](/glossary/hallucination-rate): Hallucination rate is how often the AI says something confidently wrong.
- [Voice AI orchestration](/glossary/voice-ai-orchestration): Voice AI orchestration is the conductor for everything happening during a call.
- [SIP trunking](/glossary/sip-trunking): SIP trunking is the IP pipe that carries voice into and out of the contact centre.
- [Voice cloning](/glossary/voice-cloning): Voice cloning is making the AI sound like a specific voice rather than a stock one.
- [Prompt injection (voice)](/glossary/prompt-injection-voice): Prompt injection in voice AI is a caller trying to talk the agent out of its rules.
- [Conversational design](/glossary/conversational-design): Conversational design is writing the dialogue rules the AI lives inside.
- [LLM guardrails](/glossary/llm-guardrails): LLM guardrails are the things the AI is not allowed to do.
- [Automated-system disclosure](/glossary/automated-system-disclosure): Automated-system disclosure is telling the caller they are talking to a machine.
- [Voice AI evaluation](/glossary/voice-ai-evaluation): Voice AI evaluation is comparing platforms on what survives production, not what wins demos.
- [Voice AI ROI](/glossary/voice-ai-roi): Voice AI ROI is the saving per resolved call after the operating-model team is paid.

## Notes

- [Why containment rate is the wrong KPI to put on a dashboard](/notes/containment-rate-wrong-kpi): Containment rate is easy to measure, easy to game, and a poor proxy for what actually matters: whether the customer's problem was solved. A note on the metric voice AI programmes should use instead.
- [The integration tax nobody prices in](/notes/integration-tax-nobody-prices-in): Voice AI business cases routinely treat integration as a phase-two implementation detail. In production, integration depth is the single biggest predictor of whether the deployment can actually resolve calls.
- [Latency budgets and why 800 ms breaks the illusion](/notes/latency-budget-and-why-800ms-breaks-the-illusion): Conversational realism collapses above roughly 800 ms of turn-taking latency. A note on where the milliseconds go and which optimisations actually move the dial.

## About

- [About the author](/about)
- [Methodology](/about/methodology)
- [Editorial policy](/editorial-policy)
- [Corrections](/corrections)

## Machine-readable endpoints

- [Full corpus as Markdown](/llms-full.txt)
- [RSS feed](/rss.xml)
- [Per-page Markdown](/llm/): /llm/guides/<slug>, /llm/notes/<slug>, /llm/glossary/<slug>, /llm/industries/<slug>, /llm/benchmarks/<slug>, /llm/use-cases/<slug>, /llm/use-cases/<slug>/<industry>