Instant explanations
Re-explains any concept in simpler words, step by step, as many times as the learner needs.
AI human agent
Live session
Smarter learning
with ANAS.
We teach and train AI models at Lxera. Pro Anas is ours — he speaks your students' language, answers from your own curriculum, and never goes offline.
Efficiency improvement
Save working hours
Curriculum reliability
Instant 24/7 support
Listening
4 weeksFrom day zero to a live expert agent
200+Questions in the graduation exam
92%Resolved without a human once live
1.8sMedian time to first answer
2Languages — Arabic & English, auto-detected
What we do
We teach and train AI models.
We teach and train AI models.
Pro Anas is ours.
Training models is the work we do at Lxera, and Pro Anas is the one we built for education. He is not installed — he is raised. Every academy gets its own Pro Anas: a newborn inside your platform, fed only your material, corrected by your instructors, and promoted level by level until he answers like the person who wrote the course.
Capability on your curriculum
Day 0 → livePick a level to see what happens in it — every promotion is a decision your team makes, not a switch we flip.
The training loop
Four moves, repeated
Four moves, repeated
until he stops being wrong
The loop runs daily during training and never fully stops after launch.
Feed
Courses, documents and policies are connected and indexed, chapter by chapter.
Your content onlyAnswer
He replies to real questions in a sandbox, always naming the lesson he used.
Sandbox, no learnersCorrect
Instructors grade each reply. Rewrites become the preferred answer permanently.
Human in the loopRe-test
The same questions come back later. Only a stable score moves him up a level.
Regression check
Course lessons
Slides & PDFs
Academy policies
FAQ articles
Past support tickets
Nothing from the open internet
The loop runs on its own — hover, tap or focus a step to hold it there.
Graduation
He does not meet a learner
He does not meet a learner
until he passes
Before launch, your team runs a fixed exam of real questions pulled from past intakes. Every answer is scored on accuracy, sourcing, tone and refusal behaviour. The thresholds below are the defaults — you can raise them.
Exam report · sample cohort
PassedCurriculum accuracy96%Minimum to launch · 90%
Answers with a valid source99%Minimum to launch · 98%
Correct refusal, out of scope94%Minimum to launch · 90%
Arabic dialect handling91%Minimum to launch · 85%
Escalated to the right person97%Minimum to launch · 95%
Once he is live
What a raised agent does all day
Personalised guidance
Reads progress, attempts and pace to suggest the next lesson instead of a generic answer.
Practice on demand
Generates quizzes and model answers from the lesson the learner just finished.
Gap reporting
Shows which lessons trigger the most confusion, so instructors fix the cause, not the symptom.
Works inside your stack
Drops into the Lxera LMS and EMS, or any system you already run, through a documented API.
Safe by default
Refuses what he cannot source, filters unsafe content, and logs every conversation for review.
Model spec
The details your IT team will ask for
Everything below is configurable per academy — scope, tone, thresholds and escalation rules are yours to set.
LMS chatWeb widgetWhatsAppEmail
- Languages
- Arabic (Modern Standard + Gulf dialect) and English, auto-detected per message
- Knowledge sources
- Courses, lesson text, slides, PDFs, policies and FAQ articles you connect
- Grounding
- Retrieval-based — every answer names the lesson or document behind it
- Training period
- Four weeks of supervised training, four promotion gates, your sign-off at each
- Latency
- 1.8s median to first answer, streamed as he writes
- Escalation
- Rule-based handoff to an instructor or support agent with the full transcript
- Analytics
- Top questions, unanswered log, per-course confusion score, weekly gap report
- Access control
- Role-based; answers scoped to the courses a learner is enrolled in
- Integration
- Native in the Lxera LMS and EMS; REST API and webhooks for everything else
- Deployment
- Cloud, with a Saudi data region option
- Training on your data
- Never. Your content teaches your agent only, and is not used to train external models
Who he serves
For learners
For instructors
For administrators
One agent, three jobs
Never stuck at 2am
- Explain this again, more simply
- Walk me through the solution step by step
- Quiz me before tomorrow's exam
- What is due this week, and where do I start?
The repetition stops
- Repeat questions answered automatically
- See exactly which lesson confuses the cohort
- Correct an answer once and it stays corrected
- Keep the conversations that actually need you
Support that scales
- Round-the-clock coverage without new headcount
- Enrolment, policy and schedule questions handled
- Consistent answers across every intake
- Engagement and response data in one dashboard
Off the shelf / raised
Why the upbringing
Why the upbringing
is the whole product
A generic chatbot
- Answers from the internet, not your course
- Confidently wrong, with no source to check
- Corrections disappear on the next release
- Goes live the day it is bought
- No record of where learners struggle
Pro Anas
- Raised on your approved material only
- Every answer names its lesson or document
- Instructor corrections are kept for good
- Goes live only after passing your exam
- Weekly gap report on what to fix next
Governance
Built to pass a procurement review
AI in education only works if the institution keeps control of the content, the data and the record.
Data residency
Content and conversation logs can be kept in a Saudi region.
Privacy aligned
Designed around PDPL expectations, with retention you configure.
Full audit trail
Every question, answer, source, correction and escalation is logged and exportable.
Human in the loop
Instructors review flagged answers and correct the knowledge base directly.
Questions
Asked by every academy
Who does the training — us or you?
Both. We run the pipeline and the sandbox; your instructors grade the answers. The grading is what makes the agent yours, so it cannot be outsourced to us.
How much instructor time does it take?
Plan for a few hours a week over the training month — mostly reviewing sandbox answers in batches. After launch it drops to reviewing flagged replies.
Does he only answer from our own material?
Yes. Answers are retrieved from the content you connect and approve. When a question falls outside that scope he says so and routes it to a human instead of guessing.
What happens when our curriculum changes?
Connect the new version and he re-reads it. Affected answers are re-tested against the exam before the update goes live.
Which languages does he handle?
Arabic — Modern Standard and Gulf dialect — and English. The language is detected per message, so a learner can switch mid-conversation.
Is our content used to train AI models?
No. Your material teaches your agent only, and is not sent for external model training.
What happens when he is wrong after launch?
Instructors flag the answer from the dashboard. The correction updates the knowledge base immediately, and the flagged exchange stays in the audit log.
Start raising
your own AI Agentic
Tell us about the idea or the AI agent you need. A Lxera builder will get back to you to scope it with you.