How it works in practice
- 1. Upload an EHCP plan - The system removes personal details first, then reads the plan and pulls out each goal into a table for you to review and approve.
- 2. Upload daily notes - Staff reports are checked for personal information, then the AI reads each note, links it to the right goals, and flags any vague entries that need more detail.
- 3. Track progress automatically - Every note updates a traffic-light status (green, amber, red) showing how each goal is progressing. Alerts appear when progress stalls.
- 4. Generate reports - At the end of each period, the system produces a report summarizing evidence, highlighting gaps, and suggesting next steps to improve progress.
0 PII to AI
Mask names and identifiers before any model call; keep a proof log.
6 phases
Move in steps: anonymise > extract goals > tag notes > monitor > report > secure.
Multi-format
Upload PDF, Word, Excel, or paste text for plans and daily notes.
What you get
Proof of anonymisation
Every upload is redacted first. We keep the clean copy and a log of what was masked.
Goal-aware notes
Notes are linked to the right objectives automatically and flagged if they are vague.
Clear progress
Traffic-light status with evidence and next steps so managers see gaps early.
Phased delivery plan
Start with anonymisation and goal extraction. Then add note tagging, monitoring, and reporting in quick iterations.
flowchart TB
A[Phase 1: Anonymise uploads] --> B[Phase 2: Extract goals]
B --> C[Phase 3: Tag daily notes]
C --> D[Phase 4: Monitor progress]
D --> E[Phase 5: Report and share]
E --> F[Phase 6: Secure and scale]
subgraph mvp["🎯 MVP - Start Here"]
A
B
end
subgraph expansion["🚀 Expansion - Build Forward"]
C
D
E
F
end
classDef accent stroke:#0ea5e9,stroke-width:2px,fill:#0a1d2f,color:#e5e7eb;
class A,B,C,D,E,F accent;
Phase 1: redaction and audit log
Phase 2: goal table for user edits
Phase 3: note tagging plus quality checks
Phase 4: traffic-light thresholds and alerts
Phase 5: reports with evidence
Phase 6: hosting, encryption, GDPR
How the data flows
Anonymisation sits in front. Redacted files and logs stay separate from goals, notes, and reports.
flowchart TD
U[📤 Upload EHCP plan or notes] --> PII{Contains personal info?}
PII -->|Yes| RED[🔒 Anonymise and mask]
PII -->|No| CLEAN[✅ Use as-is]
RED --> CLEAN
CLEAN --> SPLIT{What type?}
SPLIT -->|EHCP Plan| G[📋 Extract goals]
SPLIT -->|Daily Note| N[📝 Intake daily notes]
G --> TABLE[📊 Goal table for review]
TABLE --> VAULT[✅ Approved goals]
N --> CHECK{Specific and measurable?}
CHECK -->|Needs detail| PROMPT[⚠️ Ask for specifics]
PROMPT --> N
CHECK -->|Clear| TAG[🏷️ Auto-tag to goals]
TAG --> LINKS[🔗 Link to one or many goals]
LINKS --> EVIDENCE[💾 Store evidence]
EVIDENCE --> STATUS[🚦 Traffic-light status]
STATUS --> ALERTS[🔔 Alerts and nudges]
STATUS --> REPORTS[📄 Periodic reports]
classDef primary stroke:#0ea5e9,stroke-width:2px,fill:#0a1d2f,color:#e5e7eb;
classDef success stroke:#22c55e,stroke-width:2px,fill:#0f291a,color:#e5e7eb;
classDef warn stroke:#f59e0b,stroke-width:2px,fill:#2a210c,color:#e5e7eb;
class RED,PROMPT warn;
class TAG,LINKS primary;
class STATUS,VAULT success;
Daily notes: simple path
Notes must be specific. If they are vague, the system asks for more detail before they count toward evidence.
flowchart TD
start[👤 Staff uploads note] --> pii{Contains names or personal details?}
pii -->|Yes| mask[🔒 Mask PII automatically]
pii -->|No| clean[✅ Clean text]
mask --> clean
clean --> quality{Specific and measurable?}
quality -->|No| prompt[⚠️ Ask staff for missing detail]
prompt --> start
quality -->|Yes| tag[🏷️ Auto-tag to goal or goals]
tag --> store[💾 Store evidence with confidence score]
store --> status[🚦 Update traffic-light status]
status --> check{Status changed?}
check -->|Yes| alerts[🔔 Send alert if needed]
check -->|No| done[✅ Complete]
alerts --> done
classDef good stroke:#22c55e,stroke-width:2px,fill:#0f291a,color:#e5e7eb;
classDef warn stroke:#f59e0b,stroke-width:2px,fill:#2a210c,color:#e5e7eb;
classDef flow stroke:#0ea5e9,stroke-width:2px,fill:#0a1d2f,color:#e5e7eb;
class start,clean,tag,store,status,done flow;
class mask,prompt warn;
class alerts good;
- Quality gate blocks vague notes until specifics are provided.
- One note can support several goals when progress overlaps.
- Confidence scores ride with each tag to tune alerts and reports.
What happens first
MVP (weeks 1-3)
- Standalone anonymiser with audit log and redacted download.
- Goal extraction from EHCP plan into an editable table.
- Basic note intake and auto-tagging using approved goals.
Phase 2 (weeks 4-6)
- Quality scoring for notes; prompts for missing detail.
- Traffic-light dashboard with adjustable thresholds.
- Early alerts for stalled goals; next-step guidance.
Phase 3 (weeks 7+)
- PDF/HTML narrative reports with evidence per goal.
- Role-based access, encryption at rest and in transit.
- Decision on local vs private cloud AI endpoints.