Max Budget

£1,000,000

Over 2 Year Contract

Efficiency Target

£2.5m+

Projected Savings

Evaluation Split

70 / 30

Quality / Cost Weighting

Implementation

4 Months

Induction to Go-Live

A Speech Transcription

Difficulty: Medium
Technical Complexity 3/5

The risk is Annex 2a: Speaker Identification in multi-party meetings (diarization). Standard APIs like Azure Whisper/Cognitive work well, but fine-tuning for "diverse regional accents" and "voice regardless of age" (children/seniors) is a labor-intensive testing task.

Challenge: Handling background noise in field notes vs. high-fidelity corporate meetings.

B Doc Gen & Report Writing

Difficulty: Medium
Technical Complexity 3/5

The risk is Legal Hallucination. Generating EHCPs (Special Needs plans) is high-stakes. The tech is standard template-mapping, but the Prompt Governance to ensure compliance with WCC's specific legal terminology is a high-risk QA burden.

Challenge: Multi-source ingestion (SharePoint, CSV, DBs) into a unified template engine.

C Interpretation & Translation

Difficulty: Low
Technical Complexity 2/5

Most accessible via Azure Translator / DeepL API. The primary risk is latency in real-time interpretation on mobile devices over poor network conditions (e.g., social workers in the field).

Challenge: "Contextual accuracy suitable for official use" – requires a manual overrides layer.

D Digital Assistants

Difficulty: High
Technical Complexity 5/5

The biggest risk. WCC wants "Digital Assistants that can authenticate users and log service requests directly into line-of-business systems." Integrating with legacy council ERPs/CRMs via secure APIs is a massive engineering hurdle for a small team.

Challenge: Secure authentication (Entra ID) and bidirectional data flows with back-office systems.