Course-grounded retrieval
RAG over your courseware: lessons, slides, transcripts, FAQs, assessment items. Citations point back to the specific lesson the answer came from. Refusal when the answer isn't in your material.
Course-Aware · 100+ Languages · 24/7
AI Tutor · ATCwise
A tutor who knows your courses, in their language, on their schedule.
ATCwise is a course-aware AI assistant that learners can interrogate at 2 a.m. in Tamil, in Hindi, in Vietnamese — and get answers grounded in *your* curriculum, not generic web content. It also generates quizzes, summarises lectures, and produces revision guides on demand.
400K+
Active learners
4.9★
Learner rating
100+
Languages
including 22 scheduled Indian languages
~40%
Fewer L&D support tickets
first-line questions deflected
The problem
Why learner support is the hidden cost of L&D
01
Learners stuck wait days for instructor responses
A finance trainee gets confused at 9 p.m. on Sunday. They can't reach the instructor. Monday morning, they're behind, demotivated, and likely to drop off.
02
Generic AI chatbots hallucinate on your specific curriculum
Off-the-shelf ChatGPT doesn't know your safety protocol or your compliance manual. Plugging it into your LMS without grounding produces plausible-sounding wrong answers — worse than nothing for compliance training.
03
Translating courses into 22 Indian languages costs ₹50L per language
Real localisation requires native-speaking SMEs, not Google Translate. For most L&D budgets, multi-lingual delivery is impossible.
Capabilities
Course-aware AI grounded in YOUR curriculum, available in 100+ languages.
ATCwise indexes every lesson, slide deck, transcript, and assessment in your courses. When a learner asks a question, it answers from your material — not the public internet. And it answers in their language without separate localisation pipelines.
100+ languages on-demand
Learners chat in their preferred language. Same source content, instant translation including the AI tutor responses, subtitles, and on-screen text. 22 scheduled Indian languages day-one.
AI exam & quiz generation
Generate practice quizzes from any module: MCQ, short-answer, scenario-based. Difficulty levels. Item analysis. Feeds directly into ATClearn's adaptive engine.
Lesson summarisation & revision guides
On-demand TL;DR for any lesson. Revision flashcards from the assessment items. Audio versions for commute-friendly review.
Voice in, voice out (optional)
Speak the question, hear the answer. Useful for blue-collar workforces, partially-sighted learners, and commute-friendly review. Works in all supported languages.
Confidence scoring + hallucination guardrails
When ATCwise isn't confident, it says so and surfaces the lesson the learner should consult. We make refusal a feature, not a bug — critical for compliance and safety training.
Use cases
How teams use ATCwise
Head of Learner Success
After-hours learner support for a 200K-student EdTech
Replace ₹3 Cr/year of human support staff with ATCwise on first-line questions. Escalate to humans only on ambiguity. NPS went up; cost per supported learner went down 70%.
L&D + Plant Heads
Compliance training in 8 Indian languages for a manufacturing major
Shop-floor safety training in Hindi, Marathi, Gujarati, Tamil, Telugu, Kannada, Bengali, Odia from a single English source. Workers ask questions in their language and get cited answers from the safety manual.
Dean of Academics
University-wide tutor for an engineering college's 12 subjects
ATCwise indexed all 4-year syllabi, lecture transcripts, and faculty notes. Students ask 3 a.m. questions and get cited answers. Teaching-assistant load dropped 50%; pass rates improved 15%.
Solo creator on the marketplace (2026)
Course chatbot for an instructor selling a Cybersecurity course
ATCwise auto-attaches to the creator's course. Students get instant answers grounded in the creator's lectures, not the public internet. Creator's support DMs drop 80%.
Outcomes
What customers see in production.
~40%
fewer L&D support tickets
70%
lower cost per learner-support interaction
<3s
median response time (vs. ~24 hours human)
15%+
higher pass rates with always-available tutor
Integrations
Works with what you already have
Existing courseware (SCORM, xAPI, PDF, video, slides)Webhook-driven ingestion for new contentLMS embed widgets (iframe / Web Component)Voice (Web Speech API + on-prem TTS/ASR optional)
Frequently asked
How is ATCwise different from ChatGPT in our LMS?
Two ways. First, grounding: ATCwise answers from *your* course content (RAG), not the public internet. So compliance and safety answers are correct for *your* policy, not "in general". Second, deployment: ATCwise can run entirely on your infrastructure with open-source models — no learner data leaves your network.
Which 100+ languages exactly?
All 22 scheduled Indian languages (Hindi, Bengali, Marathi, Telugu, Tamil, Gujarati, Urdu, Kannada, Odia, Malayalam, Punjabi, Assamese, Maithili, Sanskrit, Sindhi, Konkani, Kashmiri, Nepali, Manipuri/Meitei, Bodo, Dogri, Santali) + English; plus Spanish, Portuguese, French, German, Italian, Russian, Mandarin (Simplified + Traditional), Cantonese, Japanese, Korean, Vietnamese, Thai, Indonesian, Malay, Tagalog, Arabic, Persian, Turkish, Hebrew, Polish, Czech, Hungarian, Romanian, Greek, Bulgarian, Ukrainian, Dutch, Danish, Swedish, Norwegian, Finnish, Swahili, Yoruba, Zulu, Amharic — and a long tail of low-resource languages with reduced fluency.
How do you prevent hallucinations on safety/compliance topics?
Three layers. (1) Strict RAG: answers must cite a passage from your courseware; without a citation we refuse. (2) Confidence threshold: below the threshold, ATCwise says "I'm not sure — see lesson X" instead of guessing. (3) Topic-class guardrails: safety, compliance, and medical topics get tighter thresholds and explicit human-escalation paths.
Can it run entirely on-premise?
Yes, with open-source models (Llama, Mistral, Phi, Qwen, Gemma) running on your GPUs. No data leaves your network. Common configuration for BFS, healthcare, defence, and government deployments.
What does it cost per learner-question?
On-premise: marginal cost ≈ 0 once GPUs are amortised; you control the budget. Cloud foundation models: variable per-query — typical range ₹0.10–₹2 per question depending on the model. We help you model the trade-off during the architecture review.
Ready to see ATCwise on your data?
30-minute architecture review. We scope the deployment to your compliance, integrations, and language requirements.