Tirian: Innovation Consultancy and Facilitation

26/10/2026

Cracking Copilot: Why Most AI Testing Gets It Wrong

As organisations race to adopt AI, many are focusing on the technology itself while paying less attention to how it is evaluated.

One insight continues to emerge from our work with leaders:

The quality of an AI answer is often influenced as much by how we frame, test and challenge the problem as by the technology itself.

This was a recurring theme during Tirian's recent Critical Thinking in the Age of AI workshop with KPMG. The discussion quickly moved beyond tools and prompts to questions of judgement, assumptions, evidence, bias and decision-making.

As AI becomes increasingly embedded in organisational decision-making, the ability to critically evaluate AI outputs may prove more important than the ability to simply use AI.

Our latest article explores why testing AI is becoming a critical thinking capability, and why leaders need to look beyond simple model comparisons when assessing AI solutions.

🔗 https://andrewgaiagrant.substack.com/p/cracking-copilot-why-testing-ais?r=o972c

If every organisation has access to the same AI tools, will competitive advantage come from the technology... or from the quality of thinking used to challenge it?

20/10/2026

The greatest risk of AI isn't that it thinks. It's that we stop thinking.

Recently, Gaia Grant (PhD) Agnes Durlik and Andrew Grant delivered the first workshop from a series for KPMG - exploring what critical thinking looks like in the age of AI. Participants worked through case studies, tested assumptions, and learned practical tools to strengthen judgment and decision-making. We also gave them a series of exercises that required them to challenge AI-generated responses and put their developing skills to the test.

One discussion point particularly resonated strongly: If your primary value is validating a robot’s opinions, you aren't future-proofing your career; you're documenting your own obsolescence. AI is becoming a powerful workplace partner. But the people who will create the most value won’t be those who unquestioningly accept its answers. They'll be the ones who know when and how to challenge them.

https://tirian.com/key-topic-suites/create/critical-thinking-rescue-strategies-csi2-2/

15/09/2026

This week, some technology leaders called for slowing AI down. The President of the United States pushed back, arguing that America needed to move faster. The debate will continue.

But I'm not convinced that's the most important question. Over the years, I've noticed that many failures don't come from poor ex*****on. They come from solving the wrong problem exceptionally well.

That's why I believe critical thinking is becoming more important, not less.

As AI makes it easier to analyse, optimise, automate, and accelerate, our ability to question assumptions, examine evidence, and challenge our own thinking becomes even more valuable.

Critical thinking isn't about slowing progress. It's about ensuring we're heading in the right direction before we accelerate.

Which is why I keep coming back to a simple question:

Are we solving the right problem in the right way?

Our latest article explores why critical thinking may become one of the most important leadership capabilities of the AI era. Andrew Grant
Gaia Grant
https://andrewgaiagrant.substack.com/p/when-ai-broke-the-rules?r=o972c&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

25/08/2026

One idea sits at the heart of this conversation with Gaia Grant

People don't behave according to strategy. They behave according to systems.

As autonomous AI agents interact, adapt, and form emergent behaviours, they reveal something we've long observed in organisations: incentives matter, architecture matters, and culture is often a downstream consequence of system design.

The AI Mirror reflects the hidden assumptions, trade-offs, and incentives embedded in the way we lead and organise.

That's confronting because the mirror doesn't lie.

It's also hopeful because systems can be redesigned.

In this episode, Gaia and I explore what autonomous AI simulations can teach us about leadership, culture, and organisational design. Listen on: Substack Apple and Spotify.

Why behaviour—human or machine—is driven by incentives, not intention, and what that means for leadership and culture.

16/07/2026

My latest article looks at a viral AI game-theory experiment that reveals a glaring truth about corporate culture: systems drive behaviour.

When researchers posed a "life-and-death" dilemma to different AI models, the models reacted completely differently—some choosing high-risk selfishness, others choosing cooperation. The variation wasn't about intelligence; it was about their training. They did exactly what their underlying systems shaped them to do.

Gaia and I see the exact same pattern in our corporate consulting work, a phenomenon we call the "collaboration deception."

Many leaders publicly champion teamwork, yet maintain reward systems, individual KPIs, and promotion pathways that implicitly condition people to look out for number one. If your team defaults to self-interest, the solution isn't to urge them to "collaborate better." It’s to fix the conflicting structures.

Read the full piece to see how to audit your organization's "training data" and design for genuine collaboration https://andrewgaiagrant.substack.com/p/what-the-ai-mirror-reveals?r=o972c

09/07/2026

If dogs look like their owners, do AI agents act like their creators?
There’s a well-known psychological phenomenon that people unconsciously choose pets that reflect aspects of themselves.
Could we be seeing something similar in enterprise AI—but with far higher stakes?

When organisations deploy a large language model, they're not simply adopting neutral software. Through training data, reinforcement learning, and human feedback, AI companies embed assumptions, values, and behavioural preferences into their models.

Whether it's the more provocative posture associated with Grok, ChatGPT's highly polished agreeableness, or Claude's emphasis on thoughtful collaboration, each model exhibits distinct tendencies that reflect different design philosophies.

Are we conducting the largest culture change experiment in history by deploying AI agents without considering their behavioural influence?

The real question isn't whether your teams are using AI. It's whether you've considered what kind of organisational culture your AI is helping to create.

>> I'd love your perspective:
If every AI model reflects the biases and values of its creators, which AI personality would you want scaling across your company?
What behavioural traits would you want amplified across your organisation—and which would you avoid?

[Read the full article here to explore how to audit your AI strategy for behavioral alignment. https://www.linkedin.com/pulse/four-ai-archetypes-might-reflecting-your-culture-behavior-grant-zdegc/

02/07/2026

I recently came across an experiment that I haven't been able to stop thinking about. Researchers placed autonomous AI agents into simulated societies and left them to operate independently. Some developed collaborative, stable communities. Others spiraled into competition, conflict, and eventual collapse.

What fascinated me wasn't the technology itself. It was how clearly the simulation revealed the risks of pushing any one behavioral archetype too far. The agents' behavior reflected what they had been trained to optimize for—often revealing the values, incentives, and assumptions embedded by their creators.

What I found most unsettling was watching strengths become weaknesses: cooperation became conformity, competition became conflict, and consensus became complacency. The simulation wasn't just a story about AI. It was a reminder that every strength can become a weakness when taken to an extreme.

This is why organizations need an ambidextrous leadership approach, as researched by Gaia Grant. One that balances exploration with ex*****on, challenge with alignment, and innovation with stability.

It made me wonder: If AI systems inherit the values, incentives, and biases of their creators, what exactly are we training into our future digital workforce?
In this article, I explore four AI archetypes and the organizational behaviors they may be reflecting back to us.

Looking at your own organization, which behavior is most rewarded today: curiosity, conformity, competition, or agreeability?
I'm increasingly convinced that the answer tells us as much about our future AI workforce as it does about our human one.

Read & Listen to the full article:

SUBSTACK https://open.substack.com/pub/andrewgaiagrant/p/four-ai-archetypes-shaping-organizational?r=o972c&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

LINKEDIN https://www.linkedin.com/pulse/four-ai-archetypes-might-reflecting-your-culture-behavior-grant-zdegc

24/06/2026

Why is this newly published research by my partner Gaia Grant (PhD) so important? I think it's because while we face increasing complexity, it's challenging to manage competing demands simultaneously. We expect leaders to navigate complex, non-linear "wicked" problems, yet we are only preparing them for a linear world. Gaia's study has now been published by Sage in the and has been featured in Kudos.

The paper challenges the myth that innovative problem-solving is an innate talent. Through an evidence-based "paradox pedagogy" model, the study proves that ambidextrous thinking - the ability to balance competing demands - is a learnable capability.

I’ve had a front-row seat to how much thought, care, and rigour has gone into this research so it’s incredibly exciting to see it published.

A new teaching approach helps individuals balance creative and practical thinking for addressing complex challenges. By working directly with real problem-solving tensions, workshop participants build confidence and generate more innovative solutions faster. They develop the 'ambidextrous' skills ne...

09/06/2026

Great to be featured in Speakers Associates’ 2026 buyer’s guide to AI keynote speakers!

As organizations accelerate AI integration, leaders face a critical challenge: navigating the strategic friction between rapid technological adoption and core cultural stability.

True resilience isn't built on short-term technical hype. It is forged by mastering the explore-preserve tension at the core of every sustainable AI program. To move beyond generic tools and drive practical ex*****on, our work focuses directly on these organizational paradoxes.

Being able to co-present with my partner, Gaia Grant (PhD), whose rigorous academic research anchors our frameworks, allows us to prepare executive leadership teams to manage structural ambidexterity and effectively lead in an AI-shaped world.

Read the full 2026 Guide from Speakers Associates:

Choosing an AI keynote speaker for 2026? A buyer's guide to fee bands, briefing pitfalls, and 42 of the world's most-booked AI speakers, grouped by programme type.

28/05/2026

Is your corporate event still relying on the traditional, single-voice monologue? In our latest article, “Beyond the Monologue: Why the Single-Presenter Keynote Is Failing Modern Audiences,” we explore why the traditional keynote format is falling short and why a multi-perspective approach has become the new benchmark for high-impact corporate learning. True innovation and leadership cannot be modeled through a singular lens; they require a dynamic tension between diverse viewpoints. By breaking the old broadcast model, we bring a fresh, podcast-style dialogue to the stage—blending Gaia Grant’s research with Andrew Grant’s practical, real-world strategic facilitation.
Watch our brand-new promotional video to see this dual-keynote presentation format in action, and discover how bringing a powerful duo to your next event can spark real-time insight, engagement, and transformation.

Read the full article here:

Why Multi-Perspective Delivery is the New Benchmark for High-Impact Corporate Events