Finding Allies¶

How to Identify Potential Coalition Members¶

"Your allies are already there. They're the ones sighing in meetings, rolling their eyes in corridors, and wondering if they're the only one who sees it. They're waiting to be found."

The Ally Discovery Mindset¶

They're Already Out There¶

In any organization facing AI decisions, there are already people who: - Share your concerns about responsible AI - Have been burned before by technology projects - Care about citizen outcomes, not just efficiency - See the risks others ignore - Feel alone in their skepticism

These are your potential allies. Your job is to find them.

The Search Problem¶

The challenge is that potential allies often: - Don't express concerns publicly (unsafe) - Don't know who else shares their view (isolation) - Assume they're alone (learned helplessness) - Wait for someone else to speak first (collective action problem) - Are scattered across the organization (no natural connection)

Coalition building is solving this search problem.

Where to Look¶

The Usual Places¶

Risk and Compliance Functions: - They see what can go wrong - Their job is to anticipate problems - They're often overruled and frustrated - They have documentation of concerns

Legal Team: - They see the liability - They ask uncomfortable questions - They've seen similar failures before - They have formal power to slow things down

Ethics Advisors (if they exist): - Literally their job to raise concerns - Often marginalized but persistent - Have frameworks you can use - Connected to external ethics networks

Internal Audit: - See across the organization - Independent mandate - Not captured by project enthusiasm - Can create formal findings

Frontline Staff: - Know what actually works - See citizen impact directly - Have concerns that don't reach leadership - Often most affected by AI decisions

The Less Obvious Places¶

The Recently Burned: - People who were on failed projects - People who raised concerns and were ignored - People who saw what happens when it goes wrong - They know, even if they're quiet

The Old Hands: - Long-tenured staff who've seen cycles before - Institutional memory of past failures - Credibility from years of service - Often overlooked by change programs

The About-to-Leave: - People with exit options (retirement, other jobs) - Less to lose from speaking out - Might be willing to say what others won't - Can become external allies

The Methodologists: - Data scientists who care about rigor - Statisticians who see flawed analysis - Researchers who know what "evidence" actually means - Often frustrated by "AI theater"

The Ethicists-in-Hiding: - People who care about ethics but don't have that title - Often in policy, service design, or communications - Express concerns indirectly - Waiting for permission to raise issues

The Newcomers: - Fresh eyes that see what others have normalized - Not yet captured by culture - May have different experience from elsewhere - Window closes as they're socialized

The External Places¶

Oversight Bodies: - Ombudsman - Auditor-General - Human Rights Commission - Privacy Commissioner - They have independent mandates and need information

Academic Researchers: - Study government AI - Need case studies - Can amplify concerns - Provide external validation

Advocacy Organizations: - Represent affected citizens - Monitor government AI - Have public platforms - Can apply external pressure

Journalists: - Investigate government technology - Need sources - Create accountability - Last resort, but real

How to Identify Allies¶

Signal Detection¶

Look for people who:

Verbal Signals: - Ask "difficult" questions in meetings - Use qualifying language ("I'm concerned that...", "Have we considered...") - Reference past failures - Mention affected citizens unprompted - Push back (even gently) on optimistic claims

Non-Verbal Signals: - Eye rolls, sighs, or skeptical expressions (watch, don't assume) - Disengagement from enthusiastic discussions - Private conversations after meetings - Careful, hedged participation - Body language that suggests disagreement

Behavioral Signals: - Documentation that covers concerns - Requests for additional review - Escalation of issues (even if unsuccessful) - Connection with external bodies - Unusual interest in risk/ethics topics

Historical Signals: - Past concerns raised (check records) - Association with projects that tried to do it right - History of pushing back - Reputation for integrity (even if politically costly)

The Corridor Test¶

The most reliable signal is what people say when it's safe: - In corridors, not meetings - After hours, not during - In private conversations, not group settings - Over coffee, not over conference tables

If someone says something different in private than in public, pay attention. The private version is often closer to their real view.

The "One Question" Test¶

In casual conversation, try: - "What do you really think about [the AI project]?" - "Off the record, are you worried about anything?" - "If you could change one thing about how we're doing this..." - "What would you tell a friend who asked about this?"

Watch for: - Relief that someone asked - Hesitation that suggests they want to say more - Checking who might overhear - A different answer than the official line

The Ally Assessment Framework¶

For each potential ally, assess:

1. Alignment¶

Do they share your core concerns?

2. Capability¶

What can they contribute?

3. Risk Profile¶

How much can they safely do?

4. Reliability¶

Can you count on them?

The Ally Classification System¶

Tier 1: Inner Circle¶

High trust, high alignment, willing to take risks

• You share everything with them
• They would take personal risk for the cause
• You coordinate closely
• Mutual protection is guaranteed
• Small number: 2-5 people

Tier 2: Active Allies¶

Aligned, willing to act, but not all-in

• They support you openly when it's reasonable
• They share information
• They won't lead, but they'll follow
• Limited risk appetite
• Moderate number: 5-15 people

Tier 3: Sympathizers¶

Agree with you, but won't act

• They wish you well
• They won't oppose you
• They might share information passively
• They won't take any risk
• Larger number: many

Tier 4: Potential Allies¶

Don't know them well enough yet

• Could be any tier once you know them
• Need more conversation
• Worth cultivating
• Unknown reliability

Building the Initial List¶

Step 1: Brain Dump¶

Write down everyone you can think of who might share your concerns: - People you've heard express skepticism - People in functions that tend toward caution - People burned by past failures - People known for integrity - People who've asked good questions

Step 2: Network Expansion¶

For each person on your list: - Who do they know? - Who do they trust? - Who might they recommend? - Who works with them?

Step 3: Prioritization¶

Rank potential allies by: - Likelihood of alignment - Value they could add - Risk of approaching them - Effort required to build relationship

Step 4: Approach Strategy¶

For each high-priority potential ally, plan: - How will you create a conversation opportunity? - What will you say to test their views? - How will you protect both of you if they're not aligned? - What's the next step if they are?

The Approach¶

Creating Opportunity¶

Natural conversation opportunities: - Coffee after a meeting where concerns arose - Lunch with someone who asked a good question - Hallway chat about something you noticed - Shared attendance at a training or event - Working together on an adjacent issue

Manufactured (but still natural) opportunities: - Asking their advice on something in their expertise - Sharing an article relevant to your concerns - Inviting them to something where topics arise naturally - Volunteering for the same working group

The Initial Conversation¶

Start indirect: - "I've been thinking about [topic]..." - "I noticed you asked about [issue]..." - "What's your take on [concern]?" - "Do you ever worry about [risk]?"

Read the response: - Enthusiasm: lean in - Caution: go slowly - Deflection: back off - Checking surroundings: good sign, but nervous - Agreement: explore further

If aligned, gently probe: - "Are you the only one who sees this?" - "Have you talked to anyone else about it?" - "What do you think would happen if we raised this?" - "What would need to be true to change things?"

If not aligned, exit gracefully: - "Interesting perspective, thanks for sharing" - Don't argue, don't reveal your full position - Remain friendly - They're not an opponent unless they choose to be

Building From There¶

If the initial conversation goes well: - Find another opportunity to talk - Share something that demonstrates your values - Introduce them to another potential ally - Slowly increase trust and disclosure - Move toward coordination

Protecting Yourself During Search¶

The Risks¶

• Approaching someone who reports back
• Expressing views that become ammunition
• Creating a paper trail of dissent
• Being labeled a troublemaker prematurely
• Burning bridges before coalition is strong

The Protections¶

Plausible deniability: - Phrase concerns as questions, not positions - "I wonder if..." rather than "I think..." - Discussing hypotheticals rather than specifics - Gathering perspectives rather than advocating

Testing before trusting: - Share something small first - See if it gets repeated - Observe whether they check with others - Trust is earned incrementally

No paper trail (initially): - Verbal conversations first - Nothing in email that you wouldn't want forwarded - No meeting invites that look like conspiracy - Written coordination comes later, when trust is established

The Ally Discovery Template¶

For each potential ally:

Potential Ally Assessment Card¶

| Next Step | ________________________________ |

The Numbers Game¶

Not everyone you approach will become an ally: - Some won't share your concerns - Some won't trust you - Some won't be willing to act - Some will surprise you (good and bad)

Expect: - 1 in 3 potential allies become actual allies - Half of actual allies become active - A fraction of active allies join the inner circle

This is normal. Cast a wide net. The coalition you build will be smaller than the pool you explore, but it will be the right people.

"The question is not 'who agrees with me?' The question is 'who agrees with me and will still be standing next to me when it gets hard?'"