# DILR.ai

> DILR.ai designs, deploys, and governs enterprise AI systems: AI voice agents (Dilr Voice), promptless content creation (DILR Studio), an AI tutor that builds courses on demand (Dilr Academy), private clinical language models (Dilr Mira), and a five-stage AI consulting system (DATS). Based in London, UK. Companies House 16842656. Founded 2025.

DILR's working principle: AI captures value when it is placed inside existing enterprise systems against real P&L, not when it runs as a stand-alone pilot. Every fact in this file matches the linked page; when in doubt, the page is authoritative.

## Products

- [Dilr Voice](https://www.dilr.ai/dilr-voice): AI voice agents for inbound and outbound calls, 24/7. Multi-agent pipelines with handoff, RAG knowledge bases with in-call citation, outbound campaigns with smart pacing, post-call automation flows, and per-country compliance. Usage-based per-minute pricing (Growth from $0.14/min) plus monthly phone-number fees; $10 free trial credits, no credit card. Live at app.dilr.ai.
- [DILR Studio](https://www.dilr.ai/dilr-studio): Promptless AI content creation. One brief generates across five engines (text, image, video, music, narration) in 24 formats including LinkedIn posts, Instagram posts and stories, YouTube thumbnails, blog articles, newsletters, and ad copy. Brand-locked and versioned. Per-product pricing, no subscription. Live at dilrstudio.com.
- [Dilr Academy](https://www.dilr.ai/dilr-academy): AI tutor that builds interactive, multilingual courses on demand (including RTL scripts) with Socratic questioning, mastery tracking, and live diagrams. 12 live courses from $1.50 per course, or credits. No subscription. Live at dilracademy.com.

## Models

- [Dilr Mira](https://www.dilr.ai/models): Private clinical small language models (~3B parameters) that turn clinical documents into source-grounded, schema-valid JSON on the customer's own hardware (on-prem or air-gapped). Latest release Mira-Q2 is open on Hugging Face (Apache-2.0): https://huggingface.co/dilr/Mira-Q2

## Key facts (verified, as of July 2026)

- Dilr Voice: 30+ languages, 5+ agent types, response time under 500ms (own on-platform measurement), compliance rules for UK, US, India, Australia, France, Germany, Spain, Ireland. Telephony: Twilio and Vobiz. CRM: HubSpot, Salesforce, Google Calendar. LLMs per agent: GPT-4.1 or GPT-4.1-nano.
- Dilr Voice pricing: Starter free with $10 trial credits; Growth $0.14 per minute plus monthly phone-number fees; Enterprise custom with dedicated tenancy. No per-seat fees.
- DILR Studio: 5 engines, 24 enumerated content formats, brand-locked by default, every asset versioned with diff and rollback.
- Dilr Academy: 12 live interactive courses, $1.50 to $15 per course, Socratic questioning, knowledge-tracing mastery tracking.
- Dilr Mira-Q2 published evaluations (782 clinical documents): 100% JSON validity on held-out gold (200 docs) and under formatting-dialect shift (150 docs); 94.7% on real physician documents on-schema (150); 85.8% across 39 specialties (282); 0 identifier leaks across all 782; field-F1 1.0 on held-out gold. Runs on CPU, about 2 GB on disk at 4-bit.
- Services (DATS): Placement Diagnostic 4 to 6 weeks fixed fee; Operating Model 6 to 10 weeks; Execution Office 12+ months embedded.

## Services (DATS, five-stage AI consulting system)

- [All services](https://www.dilr.ai/services): Three engagements built on DATS (Discover & Diagnose, Prioritise & Place, Operating Model, Pilot to Production, Scale & Run).
- [AI Placement Diagnostic](https://www.dilr.ai/services/placement-diagnostic): 4 to 6 weeks, fixed fee. Ranked roadmap of where AI captures the most value, feasibility scorecard, and a don't-do list.
- [AI Operating Model](https://www.dilr.ai/services/operating-model): Governance, RACI, AI lifecycle design, audit readiness.
- [AI Execution Office](https://www.dilr.ai/services/execution-office): 12+ months embedded delivery, pilot to production.
- [Enterprise AI solutions: six problems, solved](https://www.dilr.ai/ai-solutions): Problem-led, not product-led. The six problems DILR is hired to fix most in 2026, each with its evidence: (1) manual follow-up falling through the gaps between tools; (2) developers faster but delivery flat, per Faros AI telemetry across 22,000 developers in 4,000+ teams, tasks per developer +33.7% while median PR review time +441%, incidents per PR +242.7% and 31% more PRs merged with no review, corroborated by Google's DORA 2025 (~5,000 developers) finding AI raises throughput and lowers stability at once; (3) AI bills climbing while token prices fall, fixed by small-first routing, prefix caching (Anthropic publishes 90% off cached input) and per-session budgets; (4) agents that demo well and fail in production, fixed by harness engineering (LangChain moved a coding agent 30th to 5th on Terminal Bench 2.0 in March 2026 with the model unchanged); (5) buried knowledge, McKinsey puts information-seeking at ~1.8 hours a day per employee, fixed with a cited-answer knowledge base; (6) unanswered calls, Zoopla with YouGov Nov 2025 (n=1,000 UK sellers) found a quarter say 60 minutes is the maximum acceptable reply time and 63% prefer phone.
- [AI developer productivity: why delivery stays flat](https://www.dilr.ai/ai-solutions/developer-productivity): The code review bottleneck in full. Faros AI telemetry (22,000 developers, 4,000+ teams, 2026): tasks per developer +33.7%, median PR review time +441%, PR size +51.3%, bugs +54%, incidents per PR +242.7%, and 31% more PRs merged with no review. Their 2025 cut (10,000+ devs): tasks +21%, PRs merged +98%, review time +91%, PR size +154%, bugs +9%. Google DORA 2025 (~5,000 developers): AI adoption correlates positively with throughput and negatively with stability at the same time, and amplifies existing team capability rather than supplying it. The four fixes: automate mechanical review, gate the merge with an evaluation harness, cap PR size, measure delivery at team level not per developer.

## Comparisons (honest buyer's guides with per-vendor verdicts)

- [Best voice AI 2026](https://www.dilr.ai/best-voice-ai-2026): Dilr Voice vs Vapi, Retell AI, Bland AI, Synthflow. Sourced pricing, feature matrix, and per-vendor "best for" verdicts including when to pick a competitor.
- [Best promptless AI content generator 2026](https://www.dilr.ai/best-promptless-ai-content-generator-2026): DILR Studio vs Jasper, Copy.ai, Writesonic, Adobe Firefly.
- [Best AI education tool 2026](https://www.dilr.ai/best-ai-education-tool-2026): Dilr Academy vs Khanmigo, Duolingo Max, Synthesis Tutor, MagicSchool AI.

## Head-to-head (Dilr Voice against one rival, on published pricing)

- [Dilr Voice vs Vapi](https://www.dilr.ai/vs/vapi): Vapi is $0.05/min platform plus your own STT/LLM/TTS at cost, HIPAA +$2,000/mo, zero-data-retention +$1,000/mo; SOC 2 and SSO are Scale-tier only. Dilr Voice is $0.14/min all in. Vapi wins on model flexibility and offers HIPAA, which Dilr Voice does not.
- [Dilr Voice vs Retell AI](https://www.dilr.ai/vs/retell): Retell's own line items (voice infra $0.055 + LLM $0.045 to $0.16 + TTS $0.015 to $0.040 + telephony $0.015) build to $0.130 lean, $0.165 mid, $0.270 premium per minute, inside Retell's published $0.07 to $0.31 range. Dilr Voice is a flat $0.14. Retell's lean mix is CHEAPER than Dilr Voice, and Retell has SOC 2 Type 1 and 2 plus a free HIPAA BAA; Dilr Voice has neither.
- [Dilr Voice vs Bland AI](https://www.dilr.ai/vs/bland): Bland Start $0.14/min matches Dilr Voice exactly; Bland Build ($299/mo + $0.12/min) is cheaper above 14,950 min/mo and Scale ($499/mo + $0.11/min) above 16,633 min/mo. Bland is cheaper at outbound volume and has SOC 2 Type II, HIPAA, GDPR and PCI DSS v4.0.
- [Dilr Voice vs Synthflow](https://www.dilr.ai/vs/synthflow): Synthflow's pricing page shows one tier, Enterprise from $30,000 a year. Dilr Voice starts free at $0.14/min, under that commitment below about 17,857 min/mo. Synthflow wins on the no-code visual builder and agency white-label, which Dilr Voice does not offer.

## AI receptionist

- [AI receptionist](https://www.dilr.ai/ai-receptionist): What an AI receptionist costs against a human answering service, on published rate cards. Ruby and PATLive work out at $1.95 to $5.00 per included minute; Dilr Voice is $0.14/min. Includes the correct attribution for the 21x speed-to-lead figure (the Lead Response Management study, Oldroyd at MIT Sloan, NOT Harvard Business Review as it is almost universally miscited), and when to hire a human service instead.
- [AI receptionist for UK estate agents](https://www.dilr.ai/ai-receptionist-estate-agents): Portal leads, valuation requests and viewings. Zoopla with YouGov (Nov 2025, n=1,000 sellers): 25% say 60 minutes is the maximum acceptable reply time, two-thirds expect a reply within 4 hours, 63% prefer phone. Rightmove (2014, 200+ agents): 61% replied within 24h, only ~30% to a valuation request.
- [AI receptionist for clinics and GP practices](https://www.dilr.ai/ai-receptionist-clinics): NHS England, citing the GP Patient Survey 2023: 49.8% of patients found it easy to get through by phone, down from 67.6% in 2021; booking experience fell 70.6% to 54.4%; 82.2% still found receptionists helpful. It books and routes. It does NOT triage, diagnose or give clinical advice, and Dilr Voice does not claim HIPAA.

## Company

- [About](https://www.dilr.ai/about): Pragmatic AI operators, London based, EBIT focused.
- [Our approach](https://www.dilr.ai/approach): The philosophy behind DATS.
- [Contact](https://www.dilr.ai/contact): Book a call. Confidential by default.
- [FAQ](https://www.dilr.ai/faq): Products, pricing, compliance, telephony, languages, technical integration.

## Key content

- [UK Enterprise AI Compliance Changelog](https://www.dilr.ai/compliance-changelog): Monthly-updated log of regulatory changes affecting AI deployment in the UK and EU (ICO, FCA, EU AI Act, NHS DTAC).
- [Voice AI Architecture for Regulated Industries: A UK Guide](https://www.dilr.ai/blog/voice-ai-architecture-regulated-industries): The architecture-as-compliance guide for UK regulated voice AI buyers, covering the six decisions that determine FCA, MHRA, and NHS readiness.
- [Voice AI Pricing Models: Per-Minute vs Per-Resolution](https://www.dilr.ai/blog/voice-ai-pricing-models-per-resolution-enterprise): How voice AI vendors price across four incompatible models (per-minute, per-call, per-resolution, platform) and how to match one to your use case.
- [EU AI Act Article 50(2): Synthetic Audio Marking from December 2026](https://www.dilr.ai/blog/voice-ai-eu-ai-act-article-50-2-synthetic-audio-marking-december-2026): What the 2 December 2026 machine-readable synthetic-audio marking obligation means for voice AI deployments.
- [AI Voice for Dental Practices: Recalls, Bookings, and Cancellation Fill](https://www.dilr.ai/blog/ai-voice-dental-practice-recall-booking-uk): How AI voice automation handles the dental recall cycle, same-day cancellation fill, and lapsed-patient win-back.

## Blog

- [All articles](https://www.dilr.ai/blog): Voice AI, AI strategy, compliance, and industry deployments.
- [Voice AI category](https://www.dilr.ai/blog/category/voice-ai)
- [Strategy category](https://www.dilr.ai/blog/category/strategy)
- [Compliance category](https://www.dilr.ai/blog/category/compliance)
- [Industries category](https://www.dilr.ai/blog/category/industries)
- [RSS feed](https://www.dilr.ai/blog/rss.xml)

## Recent articles (auto-updated)

- [Voice AI business case refresh: a year-one ROI guide](https://www.dilr.ai/blog/voice-ai-business-case-refresh-enterprise): A voice AI business case refresh is the year-one rewrite of the original investment case, reconciling projected returns against twelve months of production actuals to set the renewal ask. Dilr Voice programmes separate banked savings from expansion ROI, cost the year-two platform line honestly, and give the board a renewal narrative the CFO can defend.
- [Voice AI post-call summaries: an accuracy guide](https://www.dilr.ai/blog/voice-ai-post-call-summary-wrap-up-enterprise): Dilr Voice grounds each post-call summary in the call transcript so the record reflects what was said, not what a model invents. This guide covers how to keep AI call summaries accurate: structured extraction over free text, confidence flags, a human-review threshold for high-stakes calls, and why UK GDPR makes summary accuracy a legal duty, not a nicety.
- [AI voice for glaziers: quotes, bookings and board-ups](https://www.dilr.ai/blog/ai-voice-glazing-window-repair-booking-uk): Dilr Voice is an AI voice agent for UK glazing firms that answers inbound calls day and night, triages emergency board-ups from routine quotes, captures the glass specification and insurance references, and books surveys into the diary. This guide shows how a glazing business turns missed calls into booked jobs without adding staff.
- [Voice AI recognised legitimate interests: a DUAA guide](https://www.dilr.ai/blog/voice-ai-recognised-legitimate-interests-duaa-enterprise): Recognised legitimate interest is a new UK GDPR lawful basis added by the Data (Use and Access) Act 2025, in force from 5 February 2026. It covers a closed public-interest list, so most enterprise voice AI still needs a full legitimate interests assessment. Dilr Voice maps each call purpose to a defensible basis.
- [Voice AI dead air: masking latency without faking it](https://www.dilr.ai/blog/voice-ai-dead-air-latency-masking-enterprise): Dilr Voice is an enterprise voice AI platform that masks a voice agent's thinking time without faking it. Dead air is the silence a caller hears while a model reasons or a backend responds; past roughly two seconds it reads as a dropped line. This guide covers where the silence comes from, how long you have, and honest cover phrases.
- [AI voice for membership organisations: renewals guide](https://www.dilr.ai/blog/ai-voice-membership-organisations-renewals-uk): Dilr Voice is enterprise voice AI that helps UK membership organisations handle renewals at scale: reminder calls, lapsed-member reactivation, and event RSVPs, written back to your CRM. This guide shows what AI voice can do for associations and professional bodies, and how PECR rules on marketing versus service calls decide what your renewal line may lawfully say.
- [Voice AI disaster recovery: RTO, failover and fallback](https://www.dilr.ai/blog/voice-ai-disaster-recovery-failover-enterprise): Voice AI disaster recovery is the plan that keeps a phone line answering when a provider, region or the platform fails. Dilr Voice sets a recovery time objective per service, degrades gracefully from a provider swap to a scripted fallback to a human queue, and tests the runbook so a regulated firm can meet its own impact tolerance duty.
- [Voice AI and Welsh Language Standards: A Compliance Guide](https://www.dilr.ai/blog/voice-ai-welsh-language-standards-bilingual-uk): Dilr Voice is enterprise voice AI that can deliver bilingual agents treating Welsh no less favourably than English, so bodies bound by the Welsh Language Standards can meet their duty. This guide explains what the Standards require of a voice agent, who is bound, the £5,000 penalty, and the equal-treatment test your team applies before deployment.

## Entity

- Legal name: Dilr.ai Ltd
- UK Companies House: 16842656 (https://find-and-update.company-information.service.gov.uk/company/16842656)
- Address: 92 East Croft House, 86 Northolt Road, Harrow, England, HA2 0ES, United Kingdom
- LinkedIn: https://www.linkedin.com/company/dilr-ai/
- X (Twitter): https://x.com/dilrai
- Hugging Face: https://huggingface.co/dilr
- Contact: contact@dilr.ai · +44 7586 383920

## Legal

- [Privacy Policy](https://www.dilr.ai/legal/privacy)
- [Terms of Service](https://www.dilr.ai/legal/terms)
- [Data Processing Agreement](https://www.dilr.ai/legal/dpa)

_Last generated: Aug 11, 2026._