Frequently asked questions

Skill Chamber is a platform that helps trainers and academies turn their teaching methodology into AI-powered role-play practice apps. Learners complete short, conversational AI sessions based on what they're learning in class or training. Teachers set the learning objectives and stay in control of the experience. Learners personalise each conversation to their own context and interests. Skill Chamber is neither an LMS nor a standalone AI tutor; it's a practice layer that sits alongside your existing teaching.

The process has four steps. First, the trainer sets a learning goal: what the learner needs to practise (negotiating a deal, ordering in a restaurant, handling a phishing email). Second, the learner personalises the scenario to their own situation (their actual client, their next trip, their role in the company). Third, the learner practises in a real-time conversation with AI that plays a role and adapts to their ability. Fourth, the trainer reviews the learner's performance, individual sessions and group trends, to guide the next class or training session.

An LMS distributes content and tracks completion. A standalone AI tutor gives every learner the same generic conversation. Skill Chamber is neither. It's a practice platform. The trainer designs the practice around their methodology and feedback criteria. The learner practises real conversations, not content review. The AI is configured by the trainer, not generic. The trainer reviews transcripts and progress data to guide future lessons. You keep control. The AI amplifies your teaching, it doesn't replace it.

No. Skill Chamber integrates with your existing setup. If you run an LMS, Skill Chamber can sit inside your course flow as a practice module. If you teach live classes, learners access Skill Chamber between sessions. If you run a blended programme, Skill Chamber fills the practice gap. You don't need to restructure anything. It's a layer you add on top of what you're already doing.

A learner logs in and sees a scenario the trainer has created (for example: you're calling a client to negotiate a contract). The learner personalises it (I'm negotiating with XYZ Corp about pricing). The AI plays the other role and adapts to the learner's level. The learner speaks or writes. The AI responds in real time. At the end of the session, the learner gets immediate feedback (what went well, what to work on). The whole session takes 15-30 minutes. The learner can repeat it as many times as they want.

The trainer defines the learning objective (what the learner should be able to do). The trainer describes the scenario and the AI's role (for example, play a customer who is angry about a late delivery). The trainer sets feedback criteria (what success looks like). Skill Chamber's AI assistant helps structure the scenario and generates natural conversational variations. You don't need to write dialogue. You don't need technical skills. The system generates the practice.

Yes. Within the framework the trainer has set, learners choose details that matter to them. A language learner might choose whether they're booking a hotel in Barcelona or Paris. A negotiation trainer might let them choose the deal they're actually working on. A social engineering trainer might let them choose their company role. Personalisation drives engagement and makes practice feel relevant. Studies show learners come back more often when the practice matches their real situation.

Learners get immediate, actionable feedback after each session. The feedback uses a structured approach: first, recognition of what they did well; second, observation of what needs improvement; third, a specific suggestion for how to improve. The feedback is generated in real time by AI that's been configured with the trainer's criteria. Trainers can also review transcripts, see what each learner struggled with, and provide written feedback for the next class.

Skill Chamber is for trainers and academies that want to scale guided practice. That includes language academy directors, corporate L&D teams, professional trainers in negotiation or sales, security awareness trainers, and communication specialists. If you teach something that requires real conversation practice (languages, soft skills, communication, social engineering awareness), and you want learners to practise between sessions or on their own time, Skill Chamber is for you. It's built for institutions that want to keep their methodology and their brand, not replace them.

Currently, Skill Chamber is validated for language learning (Spanish and English) and professional communication skills (negotiation, sales, pitch and storytelling, confident communication). In development are social engineering awareness training and AI literacy training. The platform architecture supports any skill that requires real-time conversational practice. If you're a trainer in a niche area and want a custom app, we can build it on the same platform.

Language learners use Skill Chamber to practise speaking and listening between classes. The trainer sets the learning objective based on what the class is studying (ordering food, negotiating a price, explaining a process). Learners personalise the conversation to something real (their actual trip, their actual client). They practise with AI that plays a natural role and gives immediate feedback on pronunciation, vocabulary, and grammar. Teachers see which students are practising, how long they spend, and where they struggle. Early pilots show a 76% return rate over five weeks, with average session times of 23-33 minutes.

Soft-skills trainers use Skill Chamber to turn workshop theory into real-world practice. A negotiation trainer might create scenarios around deal types their learners actually work on. A sales trainer might have learners practise pitches with different client profiles. A communication trainer might set conversations around difficult conversations the learner anticipates. Learners practise on demand, get feedback, and can repeat until they feel ready for the real conversation. Trainers see group trends and can follow up in the next session with the specific skills that need work.

Yes. A meta-synthesis of 57 studies found that AI-driven dialogue tools consistently improved learners' speaking ability and motivation compared with traditional study methods. The reason is simple: speaking reinforces learning. When you use a skill in conversation, you remember it. When you get immediate feedback on your attempt, you learn faster. When you can repeat the same conversation without embarrassment, you build confidence. Research also shows that AI role-play reduces speaking anxiety: learners are more willing to take risks when there's no audience and no real consequences.

Skill Chamber is built on three principles. First: practice as homework. Learning doesn't happen from reading or watching; it happens from doing. Second: personalisation drives engagement. When practice feels relevant to the learner's real life, they come back. Third: teachers stay in control. The AI is configured with the trainer's methodology, feedback criteria, and learning objectives. The trainer reviews progress and guides next steps. The AI amplifies the trainer's reach; it doesn't replace the trainer's judgment.

Teachers define the learning goal. Teachers describe the scenario and the AI's behaviour. Teachers set the feedback criteria (what success looks like). Teachers can review transcripts of every conversation. Teachers see individual and group progress data. Teachers can adjust the scenario or feedback criteria if needed. The AI executes the conversation, but the teacher designs the practice and reviews the results. You are not handing your learners off to a machine; you are extending your reach.

Skill Chamber tracks progress at five levels. First: per-message evaluation, where the AI assesses each turn of the conversation in real time. Second: per-session evaluation, where a complete evaluation is generated at the end of each session. Third: per-scenario synthesis, where results across multiple sessions of the same scenario are consolidated. Fourth: learner profile, where all scenarios build a longitudinal profile of each learner's progress. Fifth: group synthesis, where individual profiles are aggregated to show the trainer a cohort view. This data is anchored to what learners actually did (objectives met, mistakes made, time spent), not AI interpretation of what happened.

Early data from our pilot shows measurable improvement within five weeks. Learners who spend 20 minutes per day show faster gains in fluency and confidence than learners who practise sporadically. The key is repetition and immediate feedback. Each time a learner repeats a conversation, they improve. The forgetting curve is steep (you forget 50% within an hour and 90% within six days), but active practice interrupts that forgetting. That's why daily 20-minute sessions are more effective than occasional longer sessions.

Yes. Every client operates in their own isolated environment. Your scenarios, your learners, your conversations, and your evaluations are all separated by design and never visible to another client. Skill Chamber is built on multi-tenant architecture where isolation happens at the database layer, not just the application layer. Data is encrypted in transit. Your data does not leave your environment except when it's transmitted to the foundation model API for inference (the AI needs to read your conversation to respond to it). The transmission is encrypted and the API provider does not retain or train on your data.

No. Learner conversations power the AI's responses inside your sessions. They are not used to train our systems or the foundation models behind them. Skill Chamber does not use your data to improve its own AI. The foundation model providers (OpenAI, Anthropic) contractually commit to not training on API inputs. Your data does the work for you. Then it stops.

Skill Chamber is built to GDPR principles. Under GDPR, you are the data controller of your learners' data. Skill Chamber is the processor. We process your data on your instructions, only for the purpose you've engaged us for, and only as long as needed. The relationship is governed by a Data Processing Agreement. Every client's data is kept in an isolated environment, separate from other clients' data. You have the right to access your data, delete it on request, and understand how it's processed. For specific questions about your data, get in touch.

A standalone AI tutor gives every learner the same generic conversation. You don't control the pedagogy, the feedback style, or the brand. Skill Chamber is different. You design the practice. You set the feedback criteria. The AI is configured to teach the way you teach, not the way a generic AI vendor thinks teaching should work. Your learners see your brand, not another company's logo. You stay in control. The AI amplifies your methodology, it doesn't replace it.

The market is real. AI role-play practice reached USD 500 million in 2025 and is projected to grow at 25% annually through 2031. Digital language learning is already outpacing traditional delivery models. Professional soft-skills training is shifting to blended delivery (classroom plus digital). The core insight is that digital delivery has traditionally sacrificed one-to-one guidance for scale. Recent advances in generative AI have made it possible to scale practice without sacrificing outcomes. That's why the market is growing. Skill Chamber helps trainers move with that shift while keeping control of their methodology.

Platforms like Yoodli, Second Nature, and Virti are in the market, but they target large corporate L&D teams with internal training programmes. Skill Chamber targets trainers and academies: language schools, professional communication coaches, niche specialists, and SME training providers. We're built for the trainer who wants to keep their methodology and their brand. Those other platforms are more one-size-fits-all, designed for scale in large organisations. We're designed for expert trainers who want to amplify their reach without losing control.

Onboarding is a one-time setup. You define your scenarios, your learning objectives, and your feedback criteria. Skill Chamber's AI assistant helps you structure the scenarios; you don't need to write dialogue or code. Teacher training covers how to create exercises, review learner progress, and use the data to guide your next class. Most trainers are ready to launch with their first cohort within 2-4 weeks. Language academies can launch faster because the scenarios are more standardised. Corporate trainers often take longer because they're customising around their specific business context. The exact timeline depends on how much customisation you need.

We offer pilots. A typical pilot is 4-6 weeks with a small group of your learners (20-50 people). You get to test the scenarios, see how your learners engage, review the progress data, and decide whether it's working for your context. Most pilots show strong engagement (our pilot data: 76% return rate, 23-33 minute average sessions). You get to see the evidence before you commit to a full rollout. If you want to talk through whether a pilot makes sense for your context, get in touch.

A blueprint is the core unit of practice in Skill Chamber. It captures your pedagogical intent: what the learner should practise, the scenario context, the role the AI plays, and the measurable goals you want to track. You create one by opening a conversation with the AI assistant, describing your learning objective in plain language, and refining the goals collaboratively until they feel right. The assistant proposes 2 to 4 measurable goals (for example, use conditional forms at least 3 times, or maintain formal register throughout), explains the reasoning behind each, and iterates with you. No forms to fill, no dialogue to write, no technical knowledge required. Once saved, every practice session a learner runs is anchored to that blueprint.

The platform is designed around a cap of 3 goals per session, each requiring no more than 3 instances. This is not arbitrary; it comes from practice design research. With an optimal session of 12 to 20 conversational turns, a learner produces roughly 6 to 10 productive turns. Three goals at 3 instances each requires 9 instances total, which fits naturally inside that window without forcing the conversation or crowding out natural dialogue. Goal counts that were too high (5 to 8 instances per goal) were found in early pilots to produce unnatural, strained conversations where learners were performing grammar rather than communicating. The current design keeps the conversation feeling real while still making progress measurable.

Yes. You can attach a Template Scenario to any blueprint. When you do, students who arrive via a practice link bypass the personalisation wizard entirely and land directly on the practice screen, ready to start. The template scenario includes pre-written roles, context, and reference facts, giving students the same quality of grounded briefing as if they had personalised it themselves. This is particularly useful for standardised courses where you want every student practising the same scenario, or for sharing practice links where students should start immediately without any setup steps.

Yes. This is called Practice Mode, and it is set per blueprint. There are three options. Conversation mode is speaking only: the text input disappears, students see a microphone button and must speak, and the AI's text is hidden until they choose to reveal it, forcing real listening practice rather than reading and typing. Read and Write mode is text only: standard keyboard interaction with all AI text visible immediately, and the AI uses richer vocabulary and longer descriptions suited to reading comprehension. Student Choice lets the learner decide each session. When you lock a mode, students see a lock badge and cannot override it. The AI automatically adapts its response style to match the mode, with shorter dialogue-focused responses for speaking and richer text for reading, and no extra configuration.

After clicking a practice link or logging in, the learner sees a briefing card describing the scenario: who they are, who the AI plays, and the situation context. They then enter the conversation. The AI opens the scenario and adapts in real time to the learner's level and responses. Sessions are designed for 12 to 20 turns (roughly 10 to 15 minutes), which research identifies as the sweet spot for flow state and learning. The first AI response is critical; it sets the tone and engagement for the whole session. At the end, the learner receives structured feedback: what went well, what is improving, what to work on, and a recommended next practice. They can repeat the same scenario as many times as they like, and repetition with feedback is one of the strongest drivers of skill retention.

Frequency beats intensity. Five 10-minute sessions per week produce dramatically better retention than one 50-minute session. Research on spaced repetition shows daily practice retains 85% of vocabulary after one year versus 22% with weekly practice (University of Reading). The platform is designed around a target of 15 minutes per day rather than long occasional sessions. The forgetting curve is steep, with roughly 50% of new learning gone within an hour without reinforcement. Skill Chamber interrupts that curve through regular short practice. Internally, data shows that learners who hit a 7-day practice streak are 3.6 times more likely to remain engaged long term.

The platform detects this automatically. A goal is flagged as stuck when it is not yet completed, the learner has practised it across 2 or more sessions, and there has been no improvement (zero or negative delta). When this happens, the learner's dashboard surfaces it clearly with a prompt that their coach can help, and the teacher's group dashboard surfaces it as a learning gap, showing which students are stuck on which goals and for how many sessions. This follows the same logic as the Response to Intervention (RTI) framework used in educational practice: learners who do not respond to standard instruction are identified for targeted support. The platform uses both performance level and growth rate to identify the flag, not just whether a goal is unmet.

The current Phase 1 teacher dashboard is 100% deterministic, with zero AI. Every number is computed directly from practice records, not interpreted or summarised by a language model. It shows four sections in a deliberate sequence: context stats (total students, conversations, and scenarios in the group); group learning gaps (goals where students are stuck, ranked by most affected, with per-student detail); student activity (a sortable table of sessions, scenarios, last active date, and stuck goal count per student); and group strengths (goals completed by 70% or more of students). A teacher can verify any number by clicking through to the individual student dashboard. Phase 2 will add AI-generated recommendations, but only when new practice data exists, and they will always cite the specific goals and session counts behind them, not produce abstract scores.

This is a deliberate, research-grounded inversion of the student dashboard. The student dashboard leads with strengths: research on learner motivation (Dornyei, 2001; Bandura, 1997) shows that acknowledging effort first creates psychological safety for engaging with growth areas. The teacher dashboard inverts this because teachers are not in a motivational context, they are in a decision-making context, and they need the signal, not the encouragement. The design follows Black and Wiliam's foundational formative assessment research: assessment only becomes formative when evidence is used to adapt teaching to meet student needs. Surfacing gaps first is what makes the dashboard instructionally useful, not just reportable.

The student dashboard follows a Data First, AI Second philosophy. It leads with strengths, the goals the learner has completed, to build motivation before surfacing growth areas. Every number shown is extracted from real practice data: sessions completed, goals achieved versus target, and progress trajectory across sessions. The AI generates a narrative summary, but only when new practice data has arrived since the last view (a hash comparison detects changes, avoiding unnecessary AI calls on repeat visits). If a goal is stuck, the dashboard tells the learner clearly and suggests asking their coach. The tone is designed to be motivating, not evaluative, framing practice as ongoing progress rather than pass or fail.

No. Teachers generate a practice link from any blueprint in one click and share it anywhere: Moodle, WhatsApp, email, Google Classroom. When a new student clicks the link and enters their name and email, an account is created automatically. They are placed in the right group, enrolled in the right course, and taken directly to the practice screen. The whole process goes from 5 steps (create account, log in, navigate, find blueprint, click personalise) to 2 steps (click link, enter name and email). Links are valid for 6 months, work for an unlimited number of students, and use cryptographically secure tokens that cannot be guessed or forged. Students who already have accounts are recognised automatically, with no duplicates created.

Practice links work with any platform that can hold a hyperlink: Moodle, Sensei, Google Classroom, WhatsApp, email. You paste the link where your students already are. A deeper LTI 1.3 integration (which would embed Skill Chamber inside Moodle and enable grade passback to the gradebook) is under consideration. The key tradeoffs: LTI takes 3 to 5 weeks to build, requires Moodle admin access to configure, and may block microphone access for voice practice inside an iframe. Practice links deliver the core value in days, work across all platforms including those without LTI support, and keep voice practice fully functional. The right choice depends on whether grade passback is a requirement for your institution.

Yes. The platform is currently validated for English and Spanish learning contexts. The AI can operate in either the target language or the learner's native language for briefings and instructions. Blueprint content, scenario context, and reference data can be written in any language. The platform architecture is language-agnostic; the AI adapts to whatever language the scenario is configured in. Current active pilots include learners from the UK, France, Nigeria, Ghana, and the United States.

CEFR alignment is actively being built. The Common European Framework of Reference for Languages (A1 to C2) is the international standard used by virtually every language school, university, and certification body worldwide. The current platform captures the teacher's learning objectives and tracks goals against them; the next step is mapping blueprints and feedback to CEFR's exact descriptors. Three high-priority features are in development: CEFR level tagging on blueprints (teachers label each scenario A1 to C2); a CEFR-based evaluation rubric replacing generic feedback with the six official dimensions (vocabulary range, grammatical accuracy, fluency, coherence, phonological control, and sociolinguistic appropriateness); and Can-Do statement objectives drawn from the CEFR descriptor banks.

The platform design is grounded in several specific findings. Flow state requires at least 10 turns; below this, students disengage before deep practice begins, and the optimal range is 12 to 20 turns per session. The first AI response is critical: the first 3 turns determine whether flow establishes, and a fumbled opening produces a disengaged student for the rest of the session. Response latency must stay under 3 seconds; above this students disengage, though a typing indicator from 500ms eliminates perceived latency up to 3 to 5 seconds. Frequency beats intensity, with 85% vocabulary retention after one year for daily practice versus 22% for weekly practice (University of Reading). And goal counts must fit the session, because too many measurable goals force unnatural performance rather than real communication.