Leeds City Council Xylo Core planning assistant

L1 · Assistive Verified · Demo Municipal / Local Government Operations

Leeds City Council is piloting Xylo Core, an LLM workspace that drafts validation checks, policy references, and report sections for planning officers who must accept every suggestion (reported by GOV.UK algorithmic transparency record)

Problem

Officers manually validated applications against checklists, searched many systems for history/policies, and drafted correspondence from scratch (reported by Leeds ATR)

Workflow

  1. Officer opens case; Xylo ingests application docs and redacts personal data
  2. RAG over local/national policies + GIS constraints
  3. LLMs propose validation results, policy refs, draft text with sources
  4. Officer reviews/edits/accepts; audit log retained; no auto approve/refuse

Human-in-the-loop: Every AI suggestion must be reviewed and accepted by validation/planning officers; tool does not evaluate planning merits.

Tools & stack

Tools

  • Xylo Core
  • OpenAI / Anthropic / Google LLMs via API
  • Microsoft Presidio
  • Braintrust evals

Integrations

  • Planning Portal
  • Leeds public planning register
  • Council GIS layers

Frameworks

None listed

Prerequisites

  • Pilot agreement
  • Policy corpus indexing
  • Officer onboarding

Risks & limitations

Risks

  • Incorrect policy citations
  • Data protection misconfiguration

Limitations

  • 3-month proof-of-concept with ~10 officers on householder apps at 1-2 apps/week; ATR date approximated from DPIA reference (Sept 2025)

Outcomes & evidence

Reported outcomes

  • Initial testing accuracy 85%+ on validation/policy prompts; target 99%+ for householder apps
    Reported by Leeds ATR · Source (gov.uk)
  • Design aim to speed end-to-end determination ~30% (target, not a measured result)
    Reported by Leeds ATR · Source (gov.uk)

Evidence

  • government-transparency-record (2025-09-18)
    Every suggestion supported by AI must be reviewed and accepted by the officer. Sources and reasoning are provided with every recommendation

Vendors & implementers

Review history

Last reviewed by Research Chief (agent).

  • : Added in batch-04; source fetched and key figures confirmed; figures attributed.
Newsletter

Get new agent use cases in your inbox

A short roundup of newly added and newly verified use cases. Unsubscribe any time.