Street Support Network AI Policy

Introduction

This policy sets out how Street Support Network uses artificial intelligence tools responsibly and effectively. It's designed for our team, trustees, partners, and anyone who wants to understand our approach to AI.

We've been using AI across our organisation for over a year - from meeting transcription to our virtual assistant that connects people to local support. This policy captures what we've learned about where AI helps and where human expertise is absolutely essential.

The goal is simple: give our team confidence to use AI safely and effectively, while protecting the people we serve and maintaining the trust we've built.

Quick Reference

Our Approach in Five Points

Traffic Light System

Executive Summary

We use AI to support our people, not replace them. This policy sets out how we use AI in line with our values of compassion, collaboration and practical support. It's here to give our team confidence, so we can experiment and innovate safely, knowing where the boundaries are.

What this policy does:

• Sets clear principles for ethical AI use
• Defines where human input is essential
• Encourages safe experimentation
• Protects people's personal information
• Aligns with our strategy to build shared, digital-first infrastructure

Who We Are and Why This Matters

We're Street Support Network. We connect people seeking support to help, and help organisations work better together.

AI helps us do that more effectively, but only when it reflects our values, centres human connection, and protects people's dignity. This policy exists so our small team can use AI confidently, knowing what's okay, what's not, and why.

Our Core Principles

What We Never Use AI For

Clear boundaries protect everyone:

• Processing personal data in consumer platforms
• Making final decisions about individuals without human input
• Replacing human conversations or direct support
• AI must never be used to assess, triage, or respond to safeguarding concerns — any concerns about someone's safety must always go directly to a human
• Negotiating partnerships or service agreements
• Managing emergencies or crisis situations
• Anything involving vulnerable groups without proper human review

Where Human Expertise is Non-Negotiable

These areas always require people, not machines:

• Direct support conversations with people seeking support
• Safeguarding decisions and risk assessments
• Complex case work requiring understanding of individual circumstances
• Crisis response and emergency situations
• Partnership-building and relationship management
• Strategic service decisions and board-level choices
• Final approval of all public communications
• Quality assurance for our virtual assistant responses

Where AI Helps (With Oversight)

AI adds value in these areas, but always with a human in the loop:

Admin & Operations

• Calendar scheduling and meeting coordination
• Meeting transcription and summaries
• Data entry and basic reporting
• Email and template drafts

Content Creation

• Drafting policies and internal documents
• Social media and blog content (reviewed against our tone of voice)
• Funding bid first drafts (extensively edited by humans)
• Newsletter and communication templates

Analysis & Strategy

• Meeting analysis and action point extraction
• Research and information summarisation
• Trend spotting from aggregated data
• Strategic planning support and thinking partnerships

Virtual Assistant: Real-Time Help with Safeguards

Given the higher-stakes nature of real-time advice, our virtual assistant follows different protocols:

• Pulls from verified, regularly updated service databases
• Doesn't collect or store personal information about users
• Always includes recommendations to local organisations that provide broader human advice beyond Housing Options
• Where such organisations don't exist locally, we work to establish them or temper AI expectations accordingly
• Quarterly assessment of recommendation quality and human support pathways
• Immediate incident reporting for concerning responses

Detailed governance framework for our Virtual Assistant is outlined in our AI Governance Plan, which follows IBM's five pillars of responsible AI

Data, Email & Cloud Safeguards

We've made deliberate choices about AI access to our systems:

Google Workspace Decision

We've disabled Google Workspace AI features (Smart Compose, Gemini) because they can access sensitive information across our email and file systems. This protects people seeking support from having their information processed by AI without their knowledge.

Email and Cloud Connectors

Consumer tools like ChatGPT, Claude or Gemini are great for drafting and ideas — but they're not safe places for anything involving real people's data.

Enterprise AI tools might be appropriate, but only after:

• Auditing our existing data permissions
• Understanding exactly what the tool can access
• Securing proper data processing agreements
• Getting Managing Director approval

GDPR Compliance

Our no-personal-data approach automatically ensures GDPR compliance (see our Data Protection Policy for full details). If personal information accidentally gets into an AI tool, report immediately to the Managing Director and follow our existing data breach procedures.

All AI use must comply with GDPR requirements, particularly given our work with vulnerable populations. Personal data of people seeking support requires the highest protection standards. Full GDPR compliance procedures are detailed in our Data Protection Policy.

Oversight & Quality Control

Human Review Process

• Everything AI produces gets reviewed before external use and follows our Information, Communication and Social Media Policy approval processes
• We fact-check claims, align with our tone of voice, and check for bias
• We review AI-generated content for warmth, clarity and sensitivity, because the way we speak can make a real difference to how safe and supported someone feels
• Content must sound authentically like Street Support Network
• All outputs must reflect our values and brand guidelines

Virtual Assistant Monitoring

• Monthly review of user interactions and response quality
• Quarterly assessment of service recommendations accuracy
• Regular testing with known scenarios
• Clear escalation pathways for concerning responses

Governance Structure

• Managing Director oversees all AI use and policy implementation
• Report any concerns directly to Managing Director
• Monthly team check-ins about AI effectiveness and challenges
• Annual policy review incorporating new learning and tools

How We Talk About AI

We use a layered approach to transparency, based on research showing blanket "created with AI" disclosures can damage trust:

Accessibility and Inclusion

We commit to reviewing all AI-generated content for:

• Plain language and cognitive accessibility
• Inclusion and sensitivity of tone
• Freedom from harmful assumptions or bias
• Alignment with our values of dignity and respect

Our communications should always be easy to understand and welcoming to everyone.

Incident Response

If personal information gets put into an AI tool, or if there are concerns about AI use:

1. Report immediately to the Managing Director
2. Document what happened — which tool, what information, when
3. Take action to limit potential impact
4. Follow our Data Protection Policy breach procedures if personal data is involved
5. Learn and improve — update guidance based on what happened
6. Support the person involved — no blame, we're all learning

Training and Support

For New Team Members

• Clear AI guidance during induction
• Examples of appropriate and inappropriate use
• Connection to our tone of voice and brand guidelines
• Ongoing support and mentoring

Ongoing Development

• Monthly team discussions about emerging AI tools
• Sharing what works (and what doesn't)
• Learning from other organisations' approaches
• Regular updates as technology evolves

Measuring Success

We'll know this policy is working when:

• Team feel confident using AI safely within clear boundaries
• Routine tasks get done faster, freeing time for relationships
• Our communications still sound authentically human
• People seeking support receive better, more timely help
• No incidents involving personal data
• Our AI use consistently reflects our values

Implementation Plan

Immediately:

• Share with all team members and include in team meeting discussion
• Add to induction process and contractor briefings
• Present summary to trustees for approval

Ongoing:

• Monthly team reviews of AI use and effectiveness
• Quarterly virtual assistant assessments
• Annual policy review incorporating new learning
• Regular incident response practice and team check-ins

Questions and Support

If you're unsure about anything, ask. We'd rather have too many questions than one preventable mistake.

This policy is a living document that grows with our experience. It's here to protect people and enable innovation — not block it.

This policy reflects our values of compassion, collaboration, and practical solutions. It aligns with our mission to connect people to help and help organisations work better together.