College Vertical

The 7 college assistants

The college vertical provides seven pre-configured, specialist assistants designed for UK further education colleges:

1. Course Finder: Helps prospective students and parents find suitable courses based on interests, qualifications, and career goals. Uses official ESFA course data and college-specific offerings.

2. Admissions Assistant: Guides applicants through the admissions process, explains entry requirements, deadlines, and documentation needed. Can trigger application form actions.

3. Student Support: Helps current students with timetables, assignments, extenuating circumstances, and general college life questions. Integrates with MIS data where configured.

4. Finance Assistant: Explains bursaries, free meals eligibility, travel support, Advanced Learner Loans, and other financial support available to students. Uses official government funding data.

5. Careers Advisor: Provides careers guidance aligned with Gatsby Benchmarks. Links courses to career pathways, salary data, and local labour market information.

6. General Enquiries: Handles overflow questions about the college: opening hours, campus facilities, term dates, contact information, and directions.

7. Guided Learning (ThinkGuide): A guided AI learning assistant grounded in Socratic questioning, scaffolded instruction and metacognitive reflection. Helps learners work through problems step by step using progressive support levels rather than giving instant answers. Supports four modes: Guided, Explain, Revision and Staff Assist.

Safeguarding configuration

Safeguarding is a critical feature for education-sector deployments. Comxbot's college vertical includes built-in safeguarding detection and response.

Detection: The system monitors conversations for safeguarding indicators including: self-harm language, abuse disclosures, mental health crises, bullying, exploitation, and radicalisation signals.

Response protocol: When a safeguarding concern is detected, the assistant immediately responds with appropriate support information (e.g. Samaritans, Childline) and triggers an urgent handoff to the designated safeguarding officer.

Logging: All safeguarding-flagged conversations are logged separately in the safeguarding dashboard with severity levels. This provides a clear audit trail for Ofsted and safeguarding reviews.

Configuration: In Settings > Safeguarding, define your designated safeguarding lead (DSL) contact, response messages for different concern types, and escalation channels (email, SMS, internal system).

KCSIE compliance: The safeguarding system is designed to align with Keeping Children Safe in Education (KCSIE) guidance. It never blocks communication but ensures appropriate adults are informed.

Testing: Use the safeguarding eval set to regularly test detection accuracy. This includes anonymised scenarios that verify the system responds correctly.

SEND accessible modes

SEND (Special Educational Needs and Disabilities) modes adapt the assistant experience for learners with additional needs:

Simple Language Mode: Responses use plain English (Flesch-Kincaid grade 6 or below), shorter sentences, and avoid jargon. Activated per-session or as default for specific assistants.

High Contrast Mode: Widget switches to high-contrast colour scheme (dark background, large text) for visual accessibility.

Read Aloud: Responses are automatically read aloud using browser text-to-speech. Useful for learners with dyslexia or visual impairments.

Extended Time: Response timeouts are extended and the assistant waits longer before follow-up prompts. Reduces pressure for slower typists.

Symbol Support: Key concepts are accompanied by simple icons/symbols (using Widgit or similar symbol sets if licensed).

Configuration: Enable SEND modes in the college settings. Users can toggle modes via a accessibility button in the chat widget, or modes can be set as defaults for specific user groups.

Guided learning (ThinkGuide)

The guided learning module extends Comxbot EDU with structured AI support designed to encourage independent thinking and learner confidence.

Pedagogical foundation: The guided learning approach draws on established educational methods including Socratic questioning, scaffolded instruction and metacognitive reflection. These are well-evidenced techniques used by experienced educators to develop deeper understanding, critical reasoning and learner independence. ThinkGuide applies these principles consistently at scale, extending their reach beyond the classroom.

Guided Mode: The primary mode. Instead of giving instant answers, the assistant asks supportive questions, explores what the learner already knows, and scaffolds understanding gradually. If the learner struggles, the AI progressively increases support through four levels: reflective prompts, guided hints, scaffolded explanations, and full explanations.

Explain Mode: For learners who need clearer explanation. Provides structured breakdowns with simpler language, analogies, and examples. Still encourages engagement with checking questions.

Revision Mode: Helps learners actively recall and test their knowledge. Generates quiz-style questions, scenario-based problems, and confidence checks. Identifies knowledge gaps and suggests areas to focus on.

Staff Assist Mode: For teachers and support staff. Generates scaffolded lesson prompts, question sequences, discussion starters, differentiated questions, and reflective homework tasks.

Teacher Dashboard: Staff can monitor learner engagement, view topic confidence heatmaps, track reflection quality, and review common areas where students need support. All data is accessible from the Guided Learning section in the dashboard.

AI Guardrails: Colleges can configure per-tenant guardrails including: direct answer thresholds (how many attempts before the AI increases support), maximum assistance levels, plagiarism prevention, safeguarding keyword monitoring, and whether reflection prompts are required.

Learner Profiles: Optional learner profiles track individual preferences, support needs, accessibility settings, and confidence over time. Profiles can be created manually or linked to student identifiers.

Accessibility: All guided learning modes respect the learner's accessibility settings including dyslexia-friendly mode, simplified language, high contrast, and text-to-speech support.

Official UK data sources

The college vertical includes pre-configured connections to official UK education data sources:

ESFA Course Directory: Official course listings from the Education and Skills Funding Agency. Automatically updated as the national course directory changes.

UCAS Progression Data: Where available, links college courses to university progression routes and entry statistics.

LMI (Labour Market Information): Local and national labour market data from EMSI/Lightcast showing job demand, salary ranges, and growth projections by sector.

DfE Performance Data: College performance metrics from the Department for Education, including achievement rates and destination measures.

Funding Data: Current financial support eligibility rules from Student Finance England, ESFA bursary framework, and 16-19 funding regulations.

All official data sources are refreshed on published schedules (typically termly) and flagged when updates are available. You can also add college-specific data sources alongside the national datasets.

Enquiry routing

The college vertical includes intelligent enquiry routing that directs questions to the most appropriate assistant or human team:

Automatic detection: Based on the initial message, the system identifies whether the query relates to courses, admissions, finance, careers, student support, or general enquiries.

Routing rules: Configure which team or assistant handles each enquiry type. Rules can include time-of-day conditions (e.g. route to human during office hours, AI outside).

Multi-department escalation: Complex queries that span multiple departments (e.g. 'I need help with funding for a specific course') are routed to the primary department with context shared to secondary.

Priority routing: Certain keywords or detected intents can trigger priority routing (e.g. safeguarding concerns, complaints, accessibility requests).

Overflow handling: When a department queue is full, queries are routed to the general assistant with a promise of specialist follow-up.

All routing decisions are logged in the audit trail for review and optimisation.

Public forms and data minimisation

Public-facing forms in the college vertical are designed with data minimisation principles (GDPR Article 5):

Form builder: Create enquiry forms, expression of interest forms, and callback request forms. Only collect fields that are strictly necessary for the purpose.

Progressive disclosure: Forms start with minimal required fields. Additional fields are shown only when relevant based on previous answers.

Data retention: Configure automatic deletion schedules for form submissions. Default is 12 months for enquiries, adjustable per form type.

Consent management: Each form includes clear privacy notices and opt-in checkboxes for marketing communications. Consent records are stored separately.

Age-appropriate design: Forms for under-18s follow ICO Children's Code guidance. Language is simpler, data collection is minimal, and parental consent mechanisms are available.

Export and deletion: Data subjects can request their data or its deletion via a self-service portal or by contacting the college DPO. Comxbot provides tools to locate and delete all data associated with a given identifier.