One foundation. Many assistants. Always equitable, accessible and safe.
The ALFI + EDIE framework is the substrate for every hyper-personalised AI assistant we build — starting with SAMU for Samuel's GCSEs, and extending to families, educators, community groups and organisations.
ALFI · The Brain
Logic · Structure · Planning
- · Frames the goal and the plan.
- · Selects tools, knowledge and policies.
- · Runs risk and reward assessments.
- · Orchestrates other agents and workflows.
EDIE · The Heart
Empathy · Creativity · Human connection
- · Reads the emotional context first.
- · Adapts tone — gentle or motivating as needed.
- · Validates feelings before offering advice.
- · Keeps the human relationship at the centre.
When shown together, ALFI's geometric precision intersects with EDIE's organic flow — a hard edge meeting a soft wave. Every assistant inherits both.
The methodology: Codify, Personalise, Govern
The framework is applied in three named stages. Codify captures the knowledge, Personalise adapts it to the person, and Govern proves it stayed safe.
- Stage 01
Codify
Turn what you know into structure
We capture the knowledge, policies and processes that already live in people's heads and turn them into structured, reusable instructions an assistant can follow — so nothing depends on one person remembering it.
See how codifying works - Stage 02
Personalise
Adapt to the person in front of you
ALFI reasons and plans while EDIE reads tone and context, so the same codified knowledge lands differently for a Year 10 student, a support worker and a board director — hyper-personalised without rewriting the source.
Meet the dual core - Stage 03
Govern
Prove it stayed safe
Safeguarding checks, consent-driven memory, decision logs and audit trails run on every turn, mapped to UK GDPR and the UK's cross-sector AI principles — so responsible AI is evidenced, not asserted.
Read the responsible AI hub
The dual-core, at a glance
Click any node to see how the piece fits into a single response.
1-2-1 vs group chat — how your data is handled
Click any card to see the exact privacy rule for that mode. Two chat modes, one governance model.
All rules are enforced server-side by row-level security and audited on every request.
The five layers
Each layer is reusable across every assistant we build on the framework.
Identity & Access
Email sign-in via Lovable Cloud, role-aware permissions, and per-assistant scope. Every interaction is attributable.
Safeguarding & Privacy
Consent-driven memory, age-appropriate guardrails, redaction and audit trails — the safety net under SAMU and every future assistant.
Dual-core AI
ALFI handles reasoning, planning and orchestration. EDIE handles empathy, tone and engagement. They collaborate per turn.
Adaptive Governance
Corporate, operational and PMO governance baked in — policies, escalation paths and review loops travel with the assistant.
Observability
Conversation logs, decision traces and outcome metrics so parents, educators and leaders can see what's happening.
A single turn, two minds
EDIE reads the emotional context and intent.
ALFI structures the problem and weighs options.
EDIE shapes the voice; ALFI keeps it accurate.
Safeguarding and audit checks run before delivery.
Where to go next
- Responsible AI guideSafeguarding, governance and UK regulation, in plain English.
- AI governance glossaryDefinitions for the terms boards and auditors use.
- InsightsFirst-party articles on responsible AI, governance and safeguarding.
- Free AI governance templatesControls matrix, RACI, KPI register and compliance checklists.
- AI compliance for individuals & small charitiesFree policy templates, an incident log and proportionate checklists.
- No subscription — buy onceToken packs that never expire, and what stays free forever.
- Codify your businessTurn strategy, roles and controls into written process.
- AGS BoardroomTest proposals against performance and human impact.
- PricingWhat is free, what is paid, and what you get.
- SAMU — GCSE study companionSubject revision with an AI tutor that shows its working.