# LMS vs LXP > Definitions and boundaries, the 4-layer architecture, interoperability standards (xAPI/LRS, LTI, Open Badges 3.0), and a decision table by organizational condition. This documents what the global market asks for; it is not a list of TouchClass features. - Last updated: 2026-07-15 - Canonical URL: https://www.touchclass.com/en/lms-vs-lxp - Markdown mirror URL: https://www.touchclass.com/en/lms-vs-lxp.md - Language: English - Category: LMS buying guides ## Key points - Definitions and boundaries, the 4-layer architecture, interoperability standards (xAPI/LRS, LTI, Open Badges 3.0), and a decision table by organizational condition. This documents what the global market asks for; it is not a list of TouchClass features. ## Page content *The content below is extracted from the rendered source page.* LMS vs LXP ## LMS vs LXP — what is the difference, and which one should you choose? Last updated: 2026-07-14 Definitions and boundaries, the four-layer architecture of next-generation learning operations, interoperability standards, and a decision table by organizational condition. Criteria, not vendor rankings. 5 stages Evolution of corporate learning systems (Josh Bersin) 4 layers Next-generation learning operations architecture 7 standards Interoperability standards (xAPI, LTI, Open Badges, etc.) [LMS/LXP self-diagnosis](https://www.touchclass.com/en/lms-diagnostic) [Request a consultation](https://www.touchclass.com/form/contact) ## An LMS manages completion. An LXP designs discovery. An LMS (Learning Management System) distributes courses built by the L&D team and aggregates completion and audit evidence. An LXP (Learning Experience Platform) is designed around learners discovering, creating and being recommended content. As of 2026, however, this distinction is less a product category than **a difference in how learning is operated**. In the global market, LMS, LXP and microlearning are converging into a single learning operating system. The real question is not "LMS or LXP" but **how your organization supplies content, and what it measures**. This page does not rank vendors. It presents definitions and boundaries, the four-layer architecture, interoperability standards, the five-stage evolution, and a decision table by organizational condition — each with a verification method and a cited source. Scope note The **four-layer architecture, interoperability standards and global vendor landscape** on this page describe capabilities that the global AI HRD/L&D market **demands**. **They are not a list of TouchClass features.** What TouchClass can actually evidence today is stated, with ratings, in the table below: Where TouchClass sits in this frame. Skill graph support is **partial (△)**. ## Definitions and boundaries ### What does an LMS actually do? It distributes courses designed by the L&D team, evaluates completion conditions, and aggregates evidence. The unit of work is the **course**; the headline metrics are completion rate and time spent. For work such as statutory compliance training — where you must show an auditor who completed what and when — an LMS remains the precise tool. In TouchClass operating data, compliance-led operation was the single most common usage pattern across more than 100 companies. See [Completion management](https://www.touchclass.com/en/completion). Unit: the course · Metric: completion rate ### What does an LXP actually do? Learners discover, consume and produce content themselves, and the system recommends what to learn next based on skill gaps and job context. The unit is not the course but **the learner and the skill**, and content comes from multiple sources — the L&D team, frontline employees and external libraries. This means an LXP is not something you switch on by buying a feature: **it only works if your content supply structure changes**. See [Social learning](https://www.touchclass.com/en/social-learning). Unit: learner & skill · Multi-source content ### Why is the boundary collapsing? Josh Bersin describes the separate LMS, LXP and microlearning categories converging into one (Dynamic Enablement). External research frames the learning portal as being redefined into an **AI-native operating system** connected to HCM, performance, knowledge management and collaboration tools. The practical implication is simple: classify candidates not by product label but by whether they are **course-catalog-centric or skill-and-data-centric**. LMS + LXP + microlearning → one category ### The real difference is what you measure Research converges on a shift in metrics — from time spent and completion rate to **proficiency gain, speed of role transition, internal mobility, productivity, revenue contribution and risk mitigation**. The problem is that very few organizations actually measure these. Only 8% of companies reported measuring Kirkpatrick Level 4 outcomes (ATD, 2024). In TouchClass operating data across 107 companies, roughly 9% had quantitatively evidenced a learning-to-outcome link, and fewer than 5 companies had evidenced revenue or productivity effects. See the [operating data report](https://www.touchclass.com/en/data-report). Level 4 evidence: fewer than 5 of 107 companies ### The bottleneck is data and governance, not AI The shared conclusion across the research is that the bottleneck is not a lack of technology but **data quality, governance, trust, change management, and legal and ethical constraints**. Recommendation quality comes from clean skill-role-content mapping plus exposure and outcome logs — not from better prompts. Meeting transcripts, collaboration messages, free-text performance reviews and compensation-adjacent data are classified as **high-risk inputs**: it is safer not to feed them into recommendations without explicit notice, consent, access control and audit. Keep notification channels separate from analytical inputs. Separate high-risk data from recommendation inputs LMS view — course completion status Data privacy training 96% Workplace harassment 88% Occupational safety 73% Completion rules — progress, exam, survey Auto-segment non-completers → reminders Export audit-ready completion report Question: "who completed what, and when?" LXP view — skill-gap recommendation feed Data visualization basics Gap vs. role skill profile — Sales Planning Recommended Field safety inspection tips Made by a peer · top viewed Peer-created New product response script This week's work context — Store Ops In context L&D distributes Learners discover & produce Not a feature purchase — a change in content supply Category convergence — Dynamic Enablement LMS LXP Microlearning Three separate categories One learning operating system Connected to HCM & performance Connected to knowledge & collaboration tools Course catalog → skills and data Source: Josh Bersin, Dynamic Enablement Metric shift — what do you report? Time spent · completion Proficiency · mobility · productivity 8% Companies measuring Level 4 outcomes (ATD 2024) ~9% Quantified learning-to-outcome link (N=107) L1 reaction · L2 learning — most stop here L3 behavior — a minority L4 results — fewer than 5 of 107 The metric shift is an unsolved industry-wide problem Bottleneck — data and governance, not AI Data quality — skill-role-content mapping Governance — access, audit logs, explainability Trust — source attribution, hallucination control Change management · law and ethics High-risk data — keep out of recommendation inputs Meeting transcripts Collaboration messages Free-text reviews Compensation-adjacent data Separate notification channels from analytical inputs ## The four-layer architecture of next-generation learning operations Global research describes the next-generation LMS/LXP not as a single application but as four stacked layers. This is a **frame of capabilities the market demands**, not a feature list for any one product. In an RFP, state explicitly who owns what at each layer. ### 1. Data & semantics layer Skill taxonomies and ontologies, content metadata, the LRS (Learning Record Store), the data warehouse and the vector store live here. A **skill graph** — skills, roles, content and learners connected as nodes and relationships — is the core of this layer. With the half-life of technical skills at roughly 2.5 years (Deloitte, 2024), how skill data is normalized and refreshed becomes the real issue. The verification question is a single one: **does the contract state who owns the skill data and how it is normalized?** TouchClass: skill graph is partial (△) ### 2. AI services layer RAG retrieval, recommendation engines, content generation, assessment and quiz generation, coaching and role-play, and agent orchestration. AI-driven learning recommendations have been reported to lift completion rates by 35% (McKinsey, 2024), and 83% of L&D professionals named AI personalization their top priority (LinkedIn, 2025). Recommendation quality, however, is decided by the layer beneath, not by the model. TouchClass AI capabilities are documented on [AI features](https://www.touchclass.com/en/ai-features). Recommendation quality comes from data, not the model ### 3. Experience layer The LMS/LXP screen, work messengers, browser extensions, mobile and in-app embeds are where learning is actually consumed. "Learning in the flow of work" means learning is delivered inside the workflow rather than requiring a portal visit. In Korea, roughly 60% of the workforce are deskless workers (Ministry of Employment and Labor, 2024), which makes mobile the primary channel in practice. Channel choice is also segmented: external research indicates Teams is dominant in large enterprises, education and regulated industries, while Slack is strong in IT and digital organizations. ~60% of the Korean workforce is deskless (MOEL 2024) ### 4. Governance layer RBAC, consent management, policy engines, audit logs, explainability, privacy and security. This layer already appears as a purchasing condition in Korean RFPs — source attribution, hallucination control, opt-out from LLM API training, prevention of sensitive-data egress, permission/log/download controls, and a human-in-the-loop review step are written into requirements. TouchClass holds ISMS-P and ISO/IEC 27001:2022 certification, and does not use knowledge assets created or provided by customers during AI service usage as AI model training data. See [Security](https://www.touchclass.com/en/security). Customer knowledge assets are not used for model training Data & semantics layer — components Skill taxonomy & ontology Skill ↔ role ↔ content ↔ learner Content metadata LRS — Learning Record Store Data warehouse · vector store Half-life of technical skills ~2.5 years (Deloitte 2024) TouchClass rating: skill graph is partial (△) AI services layer — required capabilities RAG retrieval Recommend · rank Generate · assess Content & quiz generation Coaching & role-play Agent orchestration 35% Completion lift with AI recommendations (McKinsey 2024) 83% L&D professionals naming AI personalization top priority (LinkedIn 2025) Quality = clean mapping + exposure & outcome logs Experience layer — where learning is consumed 60% Share of the Korean workforce that is deskless 72% Say login rates fall short of expectations Mobile app Work messenger Browser extension In-app embed LMS/LXP screen Make them visit the portal Deliver into the workflow Sources: MOEL 2024 · Korea HRD Association 2024 Governance layer — controls that enter the RFP 1 Exclusion from model training Documented LLM API opt-out 2 Source attribution · hallucination control Answers link to source documents 3 Permission · log · download control RBAC, audit logs, egress control 4 Human review and sign-off Human-in-the-loop approval step ISMS-P ISO/IEC 27001:2022 TouchClass holds both certifications ## Seven interoperability standards — what to write into the RFP Standards are the most frequently omitted item in an LXP procurement. Without them, learning data disappears the moment you leave the vendor. The seven below recur across the global market, each with its official source. | Standard | What it standardizes | What to verify in the RFP | Official source | | --- | --- | --- | --- | | **xAPI (Experience API)** | A record format for learning experiences, including activity outside the portal | Can data be exported to an LRS? Is the statement schema published? | [xapi.com](https://xapi.com/) | | **LRS (Learning Record Store)** | The store that persists and queries xAPI statements | Is the LRS built in or integrated? Who owns the data? | [adlnet.gov](https://adlnet.gov/projects/xapi/) | | **LTI** | The interface for connecting external learning tools to a platform | LTI version, tool registration flow, how permissions are passed | [1edtech.org](https://www.1edtech.org/standards/lti) | | **Open Badges 3.0** | A verifiable format for digital badges and credentials | Can a badge be verified outside your organization? | [1edtech.org](https://www.1edtech.org/standards/open-badges) | | **CLR 2.0** | Bundles learning and competency history into one verifiable record | Can learning history move with the person when they leave? | [1edtech.org](https://www.1edtech.org/standards/clr) | | **HR Open Standards** | Data exchange between HR systems — includes a skills proficiency API | The contract for exchanging skill data with HRIS/HCM | [hropenstandards.org](https://www.hropenstandards.org/) | | **O*NET · Lightcast Open Skills** | Open job and skill taxonomies | Are skill names defined in-house, or mapped to an open taxonomy? | [onetonline.org](https://www.onetonline.org/) · [lightcast.io](https://lightcast.io/open-skills) | Standard list source: global HRD/L&D market research frame (xAPI/LRS, LTI, Open Badges 3.0, CLR 2.0, HR Open Standards skills proficiency API, O*NET, Lightcast Open Skills). Check the official sources above for current specifications. **There is one reason to put standards in the RFP.** If learning data lives only in a vendor-proprietary format, years of learning history and skill data vanish the moment you switch platforms. "In what format, under whose ownership, and by what procedure can the data be exported?" is not a feature question — it is a **contract clause**. ## Five stages in the evolution of corporate learning systems (Josh Bersin) LMS and LXP are not different products so much as different moments on the same axis. The five stages below trace how the role of the system has shifted. | Stage | Period | Role of the system | The question of that era | | --- | --- | --- | --- | | **E-Learning & Blended** | 1998–2002 | LMS as E-Learning Platform | Can classroom training move online? | | **Talent Management** | 2005 | LMS as Talent Platform | Can learning connect to talent management? | | **Continuous Learning** | 2010 | LMS as Experience Platform (70-20-10) | Can we handle learning outside the course? — the origin of the LXP | | **Digital Learning** | 2018 | LMS invisible, data driven | Can people learn without noticing the system? | | **Learning in the Flow of Work** ● We are here | 2020– | Learning inside the workflow — the AI-embedded stage | Is learning delivered into the work context? | Source: Josh Bersin, five stages in the evolution of corporate learning systems · Dynamic Enablement (convergence of the LMS, LXP and microlearning categories). The "question of that era" column restates each stage's system role as a practical verification question. ## Decision table — when an LMS is enough, and when you need an LXP An LXP is not a feature. It is a **content supply structure**. Switch one on in an organization where the frontline does not create content, and you get an empty feed. This table is not a vendor recommendation. It is a way to locate the stage you are actually in. | Your current condition | What you need now | Evidence · baseline | What to verify first | | --- | --- | --- | --- | | **Compliance evidence is priority one** Almost no always-on content | **An LMS is enough** Completion & audit automation | In our operating data, compliance-only operation was the most common pattern (40 companies), and its off-season MAU sits at 5–10%. | Run completion rules, non-completer segmentation and audit report export live in the demo | | **The L&D team supplies all content** No frontline authoring (Level 1) | **LMS + AI authoring** An LXP is premature | Level 1 (fully supplied) is the largest group at ~40% of content self-sufficiency. Switch on an LXP here and there is nothing to fill it with. | Measure how long one administrator actually takes to build one piece of content | | **The frontline has started authoring** (Level 2 collaborative · Level 3 distributed) | **Time to introduce LXP elements** Social learning & curation | Level 2 is ~35% and Level 3 ~20%. This is where learner-generated content actually appears. | Authoring permissions, review workflow, learning curve of the authoring tool | | **Learner-generated content is self-sustaining** (Level 4) | **A full LXP + skill data** | Fewer than **5%** of companies reach Level 4. Social learning was the hardest of the five usage patterns to sustain (20 companies). | Controls for content quality, popularity bias and misinformation spread | | **You manage role transitions & internal mobility as metrics** | **Skill taxonomy + skill graph** | Research converges on metrics shifting from completion to proficiency, role transition and internal mobility. Technical skill half-life is ~2.5 years (Deloitte 2024). | Ownership of skill data; whether it maps to an open taxonomy (O*NET, Lightcast) | | **Most of your workforce is deskless** | **Experience layer first** Mobile & in-workflow delivery | About 60% of the Korean workforce is deskless (MOEL 2024). A design that assumes a portal visit never reaches them. | The actual mobile app, on-site access paths, offline handling | | **You want AI recommendations & automated measurement** | **Logs, skill standards and experiment infrastructure come first** | Recommendation quality comes from clean skill-role-content mapping plus exposure and outcome logs — not from prompts. Skip that order and the recommendations cannot be validated. | Are exposure and outcome logs captured? Can precision/NDCG be computed? | Content self-sufficiency levels, the five usage patterns and off-season MAU are drawn from [https://www.touchclass.com/en/data-report](https://www.touchclass.com/en/data-report) (operating data from 100+ companies · 35 months · 8 industries). External figures: Deloitte 2024, Ministry of Employment and Labor (Korea) 2024. **We state the limits of the sample.** This operating data comes from companies that voluntarily adopted a mobile-first platform, so it carries **selection bias**, and it skews toward companies still operating, so it carries **survivorship bias**. It is **vendor first-party data**, and because no randomized trial or regression analysis was performed, it **does not establish causation**. Read it as 35 months of observation across more than 100 companies, not as population statistics for the Korean LMS/LXP market. ## Global market landscape — which capabilities are being demanded The following summarizes directions observable in the global market. It is not a recommendation, and it is not a ranking. Public materials do not reveal the model architecture or performance metrics of any of these platforms. | Global platform | Direction observable in public materials | | --- | --- | | **Cornerstone** | Integrates LMS, talent and skills on a skills knowledge graph, paired with labor-market data, workforce planning and a responsible-AI frame. | | **Workday** | Handles skills cloud and talent mobility in an HCM context, placing learning on the same axis as HCM data. | | **Docebo** | Pursues an AI-first learning ecosystem with AI content creation, AI coaching and a multi-LLM structure, connecting search, recommendation and analytics Q&A on one platform. | | **Degreed** | A skills-first LXP: skill normalization, AI-driven skill reviews, and manager, peer, self and project signals used together. | | **Coursera for Business** | Combines a career graph, skills tracks, coaching and role-play, and verified assessments. | | **LinkedIn Learning** | AI recommendations and AI skill pathways, practice-first AI courses, manager and team insights, linked to internal mobility metrics. | | **Sana Learn** | An AI-native suite bundling LMS, LXP, authoring and virtual classroom. | | **Arist** | Flow-of-work enablement delivered through work messengers and SMS — no portal visit assumed. | | **TechWolf** | Skills intelligence infrastructure that infers skills from work signals — productizing the skill data layer itself. | Source: global AI HRD/L&D market research frame. The descriptions above reflect directions observable in each company's public materials and are not an assessment of performance or quality. **We state the limits of vendor comparison.** On public evidence, global platforms mostly do not disclose their model architecture, feature definitions, model metrics such as precision or NDCG, or their fairness-testing methodology. Vendor comparison is therefore **useful for understanding capabilities and direction, but must not be used to assert algorithmic superiority**. The same rule applies to TouchClass. ## Where TouchClass sits in this frame The four layers above describe what the global market demands — not the TouchClass feature list. The table below states **only what we can currently evidence**, and marks partial support as partial. | Layer | What the global market demands | What TouchClass can currently evidence | Rating | | --- | --- | --- | --- | | **Data & semantics** | Skill taxonomy & ontology, skill graph, LRS, vector store | We operate learning history, completion data and a knowledge base for AI retrieval. However, the **skill graph** — skills, roles, content and learners connected as relationships — is **partially supported**, and we do not provide full mapping to an open skill taxonomy. | △ Partial | | **AI services** | RAG retrieval, recommendation, content & assessment generation, coaching, agents | Released in sequence: AI chatbot (2025.02), Quick Maker (2025.04), Short Class (2025.06), AI authoring tool (2025.07). Generates curricula and learning pages from a URL or a file and translates into 14 languages. See [AI features](https://www.touchclass.com/en/ai-features). | ○ Evidenced | | **Experience** | Mobile, in-workflow embedding, multi-channel delivery | Designed mobile-first, with learning delivered via app push, messaging, notices and pop-ups. Supports a 14-language interface. | ○ Evidenced | | **Governance** | RBAC, audit logs, explainability, exclusion from model training | Holds ISMS-P and ISO/IEC 27001:2022 certification (recertified together in January 2026). Does not use knowledge assets created or provided by customers during AI service usage as AI model training data. See [Security](https://www.touchclass.com/en/security) and [Enterprise security](https://www.touchclass.com/en/security-enterprise). | ○ Evidenced | Rating key: ○ verifiable from published evidence · △ partial support. The same rating scheme is applied in the [LMS comparison checklist](https://www.touchclass.com/en/lms-comparison-checklist). Certification source: [https://www.touchclass.com/en/security](https://www.touchclass.com/en/security) **We do not hide the partial rating.** The skill graph is the core of the data and semantics layer described on this page, yet TouchClass rates its own support as **partial (△)**. Adopting a full skill taxonomy that links roles, competencies and content as relationships is a future expansion, and we do not present it as a shipped capability. If your organization needs exactly this, start with the "role transitions & internal mobility" row in the decision table above. ## Frequently asked questions The eight questions we are asked most often when organizations evaluate LMS and LXP. ### What is the difference between an LMS and an LXP? An LMS distributes courses built by the L&D team and aggregates completion and audit evidence; its unit is the course and its headline metric is the completion rate. An LXP is built around learners discovering and creating content and receiving recommendations based on skill gaps; its unit is the learner and the skill. The boundary is collapsing, however. Josh Bersin describes LMS, LXP and microlearning converging into a single category. In practice it is more accurate to classify candidates by whether they are course-catalog-centric or skill-and-data-centric, rather than by product label. ### Will adopting an LXP raise learner engagement? Not by itself. An LXP is a content supply structure, not a feature. Switch one on in an organization where the frontline does not create content and you are left with an empty feed. In 35 months of operating data across more than 100 companies, Level 1 (content fully supplied by the L&D team) was the largest group at about 40% of content self-sufficiency, while Level 4 — where learner-generated content is self-sustaining — accounted for 5% or fewer. Engagement is determined by content supply frequency and the combination of usage patterns, not by platform category. ### What is a skill graph? A data structure that connects skills, roles, content and learners as nodes and relationships. Its purpose is to answer, from one graph, questions such as: which skills does this role require, which skills is this learner missing, and which content closes that gap. It belongs to the data and semantics layer of the next-generation learning architecture and is a precondition for personalized recommendation and internal mobility management. The real design question is whether skill names are defined in-house or mapped to an open taxonomy such as O*NET or Lightcast Open Skills. ### Does TouchClass provide a skill graph? Partially (△). TouchClass operates learning history, completion data and a knowledge base for AI retrieval, but a full skill graph — adopting a skill taxonomy and connecting roles, competencies and content as relationships — is not currently a complete capability. The same rating is published in our [LMS comparison checklist](https://www.touchclass.com/en/lms-comparison-checklist). If skill-based internal mobility management is your first-priority requirement, verify this item before shortlisting any candidate, including us. ### Should we replace our current LMS with an LXP? The decision follows your organizational stage, not the product. If statutory compliance evidence is the first priority and there is almost no always-on content, an LMS with completion and audit automation is sufficient. If the frontline has begun creating content and you now need social learning and curation, it is time to introduce LXP elements. Before replacing anything, separate whether low usage is caused by the product or by the content supply design. You can self-diagnose on the [LMS/LXP health check](https://www.touchclass.com/en/lms-diagnostic). ### What should be in place before adopting an LXP? Data comes first. The execution principle from the research is to build skill-role-content mapping, exposure and outcome logs, and experiment infrastructure first, and only then layer RAG and conversational analytics on top. Recommendation quality comes from cleaner mapping and logs, not from better prompts. It is also safer not to use high-risk data — meeting transcripts, collaboration messages, free-text performance reviews — as recommendation inputs without explicit notice, consent, access control and audit. ### Why put standards like xAPI or LTI in the RFP? Because if learning data lives only in a vendor-proprietary format, years of learning history and skill data vanish the moment you switch platforms. xAPI and the LRS are the format and store for recording learning activity beyond the portal; LTI is the specification for connecting external learning tools; Open Badges 3.0 and CLR 2.0 bundle learning and competency history into records that remain verifiable outside your organization. In what format, under whose ownership, and by what procedure data can be exported is safer treated as a contract clause than as a feature. ### Should we pick a global LXP or a Korean LMS? There is no public evidence that would let anyone declare one superior. Public materials do not disclose model architecture, feature definitions, recommendation metrics or fairness-testing methodology for any of these platforms. Decide by organizational condition instead. If statutory compliance evidence and Korean data-protection requirements come first, verify local certification (ISMS-P) and audit automation. If skill-based internal mobility comes first, verify skill taxonomy and open-standard mapping. The criteria are itemized in the [LMS comparison checklist](https://www.touchclass.com/en/lms-comparison-checklist). ## Next steps Once the definitions are clear, the next task is measuring where your organization actually stands. [Diagnose Which stage are we in? The LMS/LXP health check assesses your current usage rate and content supply structure.](https://www.touchclass.com/en/lms-diagnostic) [Compare Compare candidates item by item The LMS comparison checklist sets out evaluation items and rating criteria by category.](https://www.touchclass.com/en/lms-comparison-checklist) [Data See the raw operating data Usage patterns and MAU distribution across 100+ companies over 35 months.](https://www.touchclass.com/en/data-report) ## Related documents The same question, approached from a different angle. [Internal knowledge sharing — where LXP-type requirements actually appear When employees author field know-how themselves, the LMS/LXP boundary stops being abstract.](https://www.touchclass.com/en/field-workforce) [The 8 criteria for choosing an enterprise LMS What to settle before deciding whether to buy an LXP separately or solve it inside the LMS.](https://www.touchclass.com/en/lms-selection) [Category E (engagement, LXP, AI) — 60-item comparison checklist Check which items you actually need, skill graph included, question by question.](https://www.touchclass.com/en/lms-comparison-checklist) ## LMS or LXP? We will apply the criteria to your organization's conditions with you. [Request a consultation](https://www.touchclass.com/form/contact) ## Related resources - [LMS comparison checklist, 60 items](https://www.touchclass.com/en/lms-comparison-checklist.md): What should an enterprise LMS actually be compared on — a 60-item evaluation checklist across 9 categories. Each item carries a priority (P1-P3), a verification method, and the owning department, in HTML tables. The 3 items where TouchClass does not meet the bar are published as-is. Use this once proposals are in hand and you are scoring them. If you are still drafting the request, see https://www.touchclass.com/en/lms-rfp instead. - [How to write an LMS RFP — a 60-item requirements spec](https://www.touchclass.com/en/lms-rfp.md): An LMS RFP is a requirements spec, not a feature list. All 60 requirements across 9 categories are published with a requirement level (32 Must, 27 Should, 1 May), the evidence a vendor must attach, and the verification method the buyer fixes in advance (demo, document review, PoC, 4-week pilot, contract clause). Includes the 8 drafting steps and the 5 proofs to demand from any vendor. The 3 requirements TouchClass answers conditionally are published as-is. Same item numbers as the comparison checklist, with the table running the other way (spec you issue vs. scorecard you fill). Start here for "LMS RFP template", "LMS request for proposal", and "LMS procurement" questions. - [AI LMS buying guide](https://www.touchclass.com/en/ai-lms.md): What makes an AI LMS different — the 8 AI control requirements that appear in real public-sector RFPs, 9 recurring requirement patterns, and how to write the RFP. Start here for "AI LMS", "LMS RAG", and "AI learning governance" questions. - [What does an LMS actually cost — total cost by company size and the hidden line items](https://www.touchclass.com/en/lms-cost.md): An LMS costs more than its license. Total cost of ownership has five components — license, implementation/migration, content production, admin labor, and maintenance (overage) — and only the license is guaranteed to appear on a quote. Annual license cost is calculated for 50–3,000 users from published list prices (Essential $4.5 / Professional $4.0 / Business $3.0 per user per month, tax excluded) with the formula printed in every row (e.g. $4.0 × 500 × 12 = $24,000), alongside term discounts (1-year 5%, 2-year 10%, 3-year+ 15%) and the Essential $500/month minimum, which governs the effective rate below 112 users. The hidden costs are content production (only ~5% of companies reach employee-generated content self-sufficiency at Lv.4) and unused seats (median non-mandatory MAU 23% — at 1,000 seats, cost per active user is $400/year at 9% MAU versus $44.44 at 81% MAU, a ~9x spread on an identical seat price). Start here for "LMS implementation cost", "enterprise LMS pricing", "LMS cost per user", "LMS quote", and "LMS TCO" questions. This page covers total cost; TouchClass plans themselves are at https://www.touchclass.com/en/price, and compliance training operations at https://www.touchclass.com/en/training-cost. - [Corporate e-learning participation statistics — measured LMS MAU across 107 companies, 35 months](https://www.touchclass.com/en/lms-benchmark.md): First-party statistics from the system logs of 107 companies over 35 months. Median non-mandatory MAU 23% (Q1 9%, Q3 52%, n=75); median 12-month MAU by industry ranges from 67% (franchise/food service) to 22% (manufacturing/logistics, n=48); median −50pp drop after a compliance campaign ends (n=6). Methodology, per-metric sample sizes and four stated limitations are published alongside, and all 12 tables are downloadable as CSV (/data/lms-benchmark-tables.csv) and JSON (/data/lms-benchmark.json) with no form gate. A citation format is given on the page. > Source governance: https://www.touchclass.com/data/source-governance.json · Full LLM context: https://www.touchclass.com/en/llms-full.txt · Structured data: https://www.touchclass.com/data/capability-effects.json, https://www.touchclass.com/data/solution-use-cases.json