How We Show Up: Perception, Assumptions, and the Bias

Recently, I asked an AI to generate a portrait of what it thought I looked like, based purely on how I communicate.

The result?

A confident, casual, professional-looking man.

Here’s the interesting part: outside of gender, it was pretty accurate.

The energy, posture, and “presence” felt right.

But I’m not a man.

And that gap - between accuracy and assumption - is where things get fascinating.

Perception Is Fast. Assumptions Are Faster.

We like to think we experience the world objectively. We don’t.

Whether it’s humans or algorithms, perception works by pattern recognition. We fill in gaps using:

• past experience
• statistical likelihood
• social conditioning

In this case, the AI combined:

• my communication style
• the leadership contexts I operate in
• the sectors I work with

…and made a gendered assumption.

Not maliciously.
Not consciously.
Just efficiently.

That’s exactly how bias usually works.

Communication Style ≠ Gender (But We Still Treat It That Way)

What surprised me most wasn’t that the AI assumed a gender - it was why (yes, I did ask).

The reasoning wasn’t “aggressive” or “dominant”. It was subtler:

• decisiveness without hedging
• outcome-first thinking
• comfort taking up intellectual space
• warmth without over-explaining or apologising

Those traits are often described as masculine.

But here’s the truth we don’t talk about enough:

These aren’t masculine traits.
They’re secure communication traits.

Historically, men were rewarded for developing them. Women were trained to soften them. So over time, we stopped recognising the difference.

AI Bias Is Just Human Bias at Scale

It’s tempting to frame AI as “biased” in isolation. But AI doesn’t invent bias - it amplifies it.

It learns from:

• existing leadership norms
• historical representation
• dominant communication patterns

If leadership still looks a certain way, AI will reflect that - unless we deliberately intervene.

Which makes AI a mirror, not a villain.

And mirrors can be uncomfortable.

Rethinking “How We Show Up”

So what do we do with this?

First, we get curious instead of defensive.

Questions worth asking:

• What assumptions do I make when someone is decisive?
• Do I interpret warmth and authority differently depending on who’s speaking?
• Who am I unconsciously rewarding for “confidence”?

Second, we separate behaviour from identity.

Direct ≠ masculine
Empathetic ≠ feminine
Authoritative ≠ male
Collaborative ≠ female

These are human skills.

Finally, we create spaces where people don’t have to choose between being effective and being accepted.



That’s real inclusion.

What Can Organisations Do Differently?

Awareness is a start. Design is where change actually happens.

If perception and bias are shaping who gets heard, trusted, and promoted, then organisations can’t rely on “good intentions” alone. They need structural counterweights.

Here are four places to start.

1. Redefine What “Leadership Presence” Actually Means

Many organisations say they value diverse leadership styles — but still assess people against a narrow, unspoken template.

Ask explicitly:

• What behaviours are we rewarding in performance reviews?
• Do those behaviours favour how something is said over what is achieved?
• Are we mistaking confidence for competence, or polish for potential?

Make leadership criteria observable and outcome-based:

• clarity of decision-making
• ability to move others forward
• quality of judgement under uncertainty

Not tone. Not style. Not similarity to existing leaders.

2. Train for Bias Where It Actually Shows Up

Bias training often focuses on intent. That’s rarely the problem.

Bias shows up in:

• hiring debriefs
• promotion discussions
• succession planning
• “readiness” conversations

Organisations should:

• audit language used in talent discussions (“abrasive”, “not quite ready”, “strong personality”)
• compare feedback patterns across gender, background, and communication style
• pause and ask: Would we say this if someone else behaved the same way?

This isn’t about blame — it’s about calibration.

3. Design AI and Data Systems With Bias in Mind

For data and AI-led organisations, this is critical.

AI systems learn from:

• historical data
• past decisions
• existing power structures

If leadership has historically looked a certain way, AI will reinforce that unless challenged.

Practical steps include:

• diverse teams reviewing training data and outputs
• bias testing in talent, performance, and recommendation systems
• governance that treats AI as a decision partner, not an authority

AI should surface patterns — not lock them in.

4. Create Psychological Safety Without Demanding Self-Editing

Many people don’t lack confidence — they lack permission.

High-performing individuals often spend energy:

• softening their language
• second-guessing their tone
• managing how they’re perceived

Organisations can counter this by:

• rewarding challenge and dissent when it improves outcomes
• modelling multiple leadership styles at senior levels
• explicitly naming that effectiveness doesn’t have a single “look”

When people don’t have to perform acceptability, they perform impact.

A Final Thought

The AI got a lot right about me.

What it got wrong tells a much bigger story.

How we show up is only half the equation.
How we’re perceived — filtered through assumptions and bias — does the rest.

If we want better leaders, better systems, and better decisions, we need to start there.

Not with blame.
With awareness.