Phil Birchenall Phil Birchenall

Critical Thinking, Curiosity and Clarity: The Real Skills We Need in an AI World

When someone on TV said maths was the key to thriving with AI, I nearly snorted coffee out of my nose. Here's why I think that’s way off….and what skills actually matter instead.

Saturday morning. I’ve a fresh coffee in one hand and I’m trying desperately not to think about AI. Because, y’know…weekend.

I look up and there’s an interview on the BBC’s Breakfast programme on what schools should be doing to adapt to world where AI is becoming increasingly widespread.

Bang. My ‘not thinking about AI at the weekend’ thing went right out of the window the moment a question was directed at the interviewee.

So what are the skills we need to prepare young people with, in order that they can thrive with AI.

Now, I’m not entirely sure who the educational expert was (I know, I’ve normally got a rigorous attention to detail but, y’know….weekend), but I couldn’t help thinking their response was misinformed.

We’re actually going to need more traditional skills like maths.”

[cue cappuccino exiting my mouth via my nostrils]

Maths? MATHS?

Maybe if you’re in the sub 1% of the UK involved in developing and shipping AI models. But what about every other member of the workforce that’ll be working with AI day-in, day-out?

Vast amounts of development costs haven’t been sunk into Large Language Models only to have them locked down for use by only the numerically gifted: the clue’s in the name.

Because, as the models advance and become ever more nuanced and capable, it strikes me that ‘technical’ proficiency becomes less and less important for the casual business user.

Forget Maths: Can You Frame the Problem?

In the first couple of months of 2025, we’ve seen ‘vibe coding’ explode: the notion that agentic coding platforms can muster up convincing web applications (and the like) from a couple of well-written prompts. And that feels like a natural continuum of the rate of change we’ve seen since ChatGPT-3.5 first dropped at the end of 2022. You start with the baby steps of HTML versions of Pong, but over time, the complexity of what’s possible ratchets up. Considerably.

So if it’s not maths that we need to be cultivating in the next generation, what skills will we need to thrive in a world powered by AI?

It's Not About Knowing Stuff. It's About Knowing What to Ask + Why You're Asking

At this point in the article, you’ll be almost as pleased as I am to find out there are “three c’s” (classic consultant chat) to consider. Uncanny.

C1: Critical Thinking

Critical thinking is going to be, erm, critical if we’re to grasp the opportunity of working with AI models. To me, this is about fully understanding the business challenges we’re trying to fix. How well do we truly understand the problem, what’s the context, what does a great outcome look like?

C2: Curiosity

Curiosity: what needs to happen to address the challenge - and what are the tools that might help us? Remember when there were four TV channels in the UK, and four ‘go to’ business tools (gee, thanks Microsoft) we used in the office? Things were simpler. But things were boring, too. AI brings not just a myriad platforms to bear, but a myriad ways we might interface with these tools. Could vision, video, or voice input help us solve our challenge in a completely new way?

C3: Clarity

Let me make this clear. Success won’t be found hiding behind corporate bull and waffle. Key word thinking is out, as are unstructured brain dumps (context dumps, on the other hand…). If you want good results out of your models, clarity is your super skill. How well can you articulate the problem you’ve thought critically about, and how well can you describe the creative solution you’ve envisaged? Simple, really, but clarity doesn’t come easy…

To me, if we can find ways to really develop those skills in the younger generation, we’ll truly capitalise on the opportunities that AI is opening up for all of us.

Of course, if you do need help with your maths, speak to my dog, the AI Maths Tutor

If you’re curious about AI, and need help exploring how to make it work for your business, why not talk to me about the training I provide through my programme, The AI Advantage, described as ‘the best workshop I’ve ever been on!’

I work with clients through whole-team workshops through to one-to-one c-suite coaching, designed to immerse you and your team in how to get the most of the tools, and how they can fit into your day-to-day operations.

Interested? Let’s have that chat, one-to-one.

Read More
Phil Birchenall Phil Birchenall

ChatGPT's Quiet Evolution: A productivity suite in the making...?

The AI landscape is evolving rapidly, with new models like DeepSeek and Grok3 emerging. But for most businesses, the real challenge isn’t just raw model power - it’s about how AI integrates into daily workflows. OpenAI understands this, and its latest updates to ChatGPT, like Canvas, Projects, Operator, and Deep Research, signal a shift towards a true productivity suite. These tools are making AI adoption in corporate environments more seamless, structured, and effective.

The AI landscape is moving at speed. While it's easy to get distracted by newer, shinier models like DeepSeek and Grok3, for most businesses, the real battle isn't simply about the sheer power of the models they're using: it's about how AI integrates into your regular daily workflows. And OpenAI knows it.

AI That Works The Way You Do

I've spent a lot of time immersed in tools such as ChatGPT, Claude and Gemini over the last couple of years. But you knew that already, right? And it strikes me that there's a lot of chatter around the advancing of the underlying LLMs that drive these tools, especially as the likes of DeepSeek and Grok3 have emerged; offering more powerful, efficient models trained for a fraction of the cost of their predecessors. All good, I guess.

(Mind you, it was literally just a month ago when you couldn't read an article about AI without being told we'd hit 'peak AI': the world had run out of data to train the models on and that was that. Those discussions didn't age well, did they?)

Yet as I've said before, the power of these emergent models is only as good as your application of - or interaction with - them.

ChatGPT is starting to feel like a nascent productivity suite rather than just a chatbot.

So, if you're building apps that connect with these models using APIs, it makes total sense to look at how DeepSeek et al might give you a competitive advantage for lower ongoing costs. But I sense that not many people are at that stage of adoption yet. And that's certainly true amongst the majority of companies that I work with.

Making AI Stickier Than Ever

Meanwhile, OpenAI has taken a two-pronged approach to its roadmap. Clearly, the company is heavily invested in the development of its core models: with o3 rolling out and GPT4.5 and GPT5 both being readied for launch 'soon'.

However, what's often a little less obvious is the investment OpenAI is making in the product itself, evolving ChatGPT's interface from the simple chatbot we know and love, into what I'd describe as something that's increasingly looking like a more general business productivity tool.

Think about it: for regular business users, infatuations with the 'latest, fastest' model will come and go, but if you build a platform that becomes intrinsic to their daily workflows, it's far harder to unpick when that shiny new LLM drops.

Beyond the Chatbox

Personally, that's where I see the updates at the end of 2024 and beginning of 2025 forming the basis of a more robust push into enterprise for OpenAI. In addition to the security assurances that a ChatGPT Team or Enterprise account brings, most of the new front-end functionality is designed to keep users increasingly 'in' the OpenAI ecosystem.

Build a platform that becomes intrinsic to people's daily workflows, and it's far harder to move when that shiny new LLM drops.

For most business users, the biggest friction point with using LLMs are around basic usability. With Canvas and Projects, OpenAI is making ChatGPT feel like a nascent productivity suite rather than just a chatbot: that's a crucial step towards AI adoption in corporate environments.

Hidden Powers

You may have used Canvas, and if you haven't, why the heck not?

Busting text out of the chat box and into fully formed documents eliminates one of the biggest complaints about originating work with these tools; the ability to edit, tweak, format, comment on a large chunk of text makes document co-creation feel like a genuine collaboration, without the constant back-and-forth of a chat-based discussion.

And where you once had a long stream of past GPeeTee chats, Projects brings structure amongst the chaos; folders for chats that can draw on shared documents and custom instructions set at a Project level. Prefer your 'proposals' project chat punchy, but you’d very much rather that your 'blogs' might be somewhat more verbose, no? It's a couple of lines of text away...although, I rarely let my AI friend near my precious articles. Oo-er.

Add in Operator and Deep Research, and you're starting to build something stickier than a treacle-dipped hedgehog in a velcro jumpsuit.

Even Tasks, which launched to little fanfare in January 2025 brings a degree of automation to help with your workflows...without the merest mention of an AI Agent. And yet, it's a fantastic way of having regular or time-based tasks schedule themselves around your agenda. Think 'send me my AI round-up news every Friday at 2pm', 'give me some SEO tips for my website every Monday morning', or 'let's pick up on this workshop planning next Tuesday.'

Add in the updates that have grabbed more attention of late, Operator and Deep Research, and you're starting to build something stickier than a treacle-dipped hedgehog in a velcro jumpsuit.

Productivity + Integration => Sheer Power

Operator is an autonomous web browsing agent that can take on activities online on your behalf (with security limitations!), undertaking some initially basic tasks just like a human would just from a prompt. Deep Research combines reasoning models (AI models that take a multi-stage approach that 'thinks' about the task at hand) with powerful online and knowledge searches to produce detailed research reports (and more). Again, these tools are less about chat and more about AI-driven automation.

Caveat: I've not yet got my hands on Operator as it's not out in the UK, or Deep Research as I need to cough up the additional $180/month to access it. But when I do, I'll be reporting back here.

I maintain - and hey, I might be completely wrong here - that for most business users integrating LLMs into their operations, it won't be the flashiest AI chatbot that wins in the short to medium term. It'll be about the company that manages to create something that feels so tightly integrated to their workflows that it's just unimaginable to suddenly stop using the wider functionality it provides. That's what will drive business adoption - and critically, retention - in the long run.

For me, right now, it's ChatGPT that's currently treading that path.

If you’re curious about AI, and need help exploring how to make it work for your business, why not talk to me about the training I provide through my programme, The AI Advantage, described as ‘the best workshop I’ve even been on!

I work with clients through whole-team workshops through to one-to-one c-suite coaching, designed to immerse you and your team in how to get the most of the tools, and how they can fit into your day-to-day operations.

Interested? Let’s have that chat: drop me a message.

Read More
Phil Birchenall Phil Birchenall

How a bonkers idea to turn our dog into an AI maths tutor ended up as an OpenAI film

When my daughter Daisy struggled with her SATs maths, I turned to AI for help. The result? Izzy, the AI Maths Tutor—powered by ChatGPT. What started as a simple fix for a family challenge became a story that caught OpenAI’s attention, leading to a global film about our journey. Read about how a cocker spaniel, a chatbot, and a bit of curiosity led to something truly unexpected.

Well, the cat’s finally out of the bag. Or should I say, the cocker spaniel’s out of the bag…?

I can barely believe I’m writing this, but a story that started with me feeling like I was failing as a parent is now the subject of a film on the website of probably one of the most important tech companies there has ever been, OpenAI.

Flashback: How Izzy Came About

It all began when my daughter Daisy was prepping for her SATs exams. She’d been struggling with maths, and in a vain attempt to get her back on track, we sat down together with a revision guide I’d bought. “How hard can it be?” I thought. How wrong was I?!

And that’s why Izzy the AI Maths Tutor was created: trained on the areas Daisy was struggling with, and given the persona of our crazy dog, Izzy. You can read more about how I used ChatGPT to help her smash her maths exams here.

A Caravan, A Rainy Day, and a Surprising Email

Fast forward a couple of months. We’re sat in a caravan in The Lake District. It’s raining and my electric MINI has been towed away because it won’t charge, and the rest of our week-long break is starting to look a bit grim.

Daisy’s on her iPad, and I’m doing the terrible parent thing of dipping into emails and checking updates on LinkedIn. Bad dad.

As I scan through messages, one catches my eye: it’s the most far-out piece of spam I’ve ever had.

Hi, I’m from OpenAI’s comms team. We’ve seen your story and we’d like to talk to you about it.

YEAH RIGHT.

Hey Daisy, have you seen this? This is exactly the kind of scam I’ve told you to watch out for!

Skepticism Turns to Reality

You see, I’m not daft: I still remember the ‘classic’ African prince email scam from the dawn of the internet.

Plus, I’m a northerner. I’m genetically programmed to treat all good news with a form of advanced scepticism we’ve developed through the ages, a thick skin hewn from the falling rocks of inter-generational misfortune… Wait. I’ve gone off on one, haven’t I?

Essentially, you can’t blag a blagger, and I’m having none of this.

But I’m intrigued: whatever this is doesn’t happen to people like me.

So I get around to responding, expecting to be asked for my bank details to help process the request. And then I get a reply which looks legit. OpenAI had read about the AI Maths Tutor I’d made using ChatGPT and they were thinking of making a film about it.

Do you have time for a call?

Blimey.

Still from the film: Izzy gets animated during her tutoring session…

From Concept to Reality

A couple of calls later, it’s getting clear this is actually all happening. There are people storyboarding our experience, and I’m getting requests to scout locations a crew could shoot me and Daisy (and our actual dog, Izzy) as we talk about them. Meanwhile, hotels are being booked and Daisy is tapping me up for a new outfit.

The shoot came around quickly. By this point, I’m checking the weather daily. It’s Manchester and, of course, it’s going to rain. Persistently: right on brand. And yet, as we got to the team’s arrival, the rain stopped and the clouds broke. A huge crew descends on our house. OpenAI literally have come knocking on our door.

We spoke about the process of setting up our GPT Maths Tutor, sure, but so much of the shoot was us hanging out at home and around our lovely village, Didsbury.

The OpenAI team were so great to spend time with. Daisy rustled up sweet treats for them, including making fresh ice cream, and the whole experience just felt like having some friends around to mess around with. It’s just that they happened to bring some expensive-looking camera and lighting kit with them.

STUFF GOT REAL

But it still hadn’t really dawned on me how much of a big deal it was until Adam, a really senior guy from the company, picked up a felt-crafted doll of Izzy that was sat on the window shelf in our kitchen.

This is just like the version of Izzy our animation guys are working on!

Gulp.

One internal double-take later, I’m starting to get a sense of how cool this video was going to be. They’d loved our story, using ChatGPT to tackle a simple human challenge. They loved that I wasn’t a techie, just a dad in need of a hand. And they loved the fact that we put our bonkers dog right at the heart of the thing we made.

You know what? They captured the heart and the humour behind turning a family dog into a maths tutor who saved the day. And we LOVE the video they made.

Daisy and Izzy high five.

A Huge Thank You!

HUGE THANKS to Taya, Jeff and Adam at OpenAI for investing in our story and giving us one of the most exciting—and surreal—experiences we’ve had. And thanks to Dan Fitzpatrick for sharing my experience with ‘Izzy’: it has been incredible to see so many people adapt the prompt to their own needs.

The Final Cut

So there we are. The film is live, it’s beautiful, and I’m hoping it inspires others to follow our example. You can find the AI Maths Tutor prompt in my article here.

If you’re curious about AI, and need help exploring how to make it work for your business, why not talk to me about the training I provide through my programme, The AI Advantage, described as ‘the best workshop I’ve even been on!

I work with clients through whole-team workshops through to one-to-one c-suite coaching, designed to immerse you and your team in how to get the most of the tools, and how they can fit into your day-to-day operations.

Interested? Let’s have that chat, one-to-one.

Read More
Phil Birchenall Phil Birchenall

There's a Whale in the Room

DeepSeek AI is being hailed as a technological marvel, offering advanced reasoning capabilities, open-source accessibility, and cost-efficient solutions. But is it better for your business? As the media frenzy grows, it’s important to pause and evaluate what “better” means for you

I wasn't going to 'dive in'.*

*Excuse 1) the terrible pun and 2) me using a phrase that appears far too often in AI-generated social media sludge. But you know, it works on several levels.

Anyway, back to the point, there's a flipping big whale in the room.

And nope, I wasn't going to post - there are way too many AI 'experts' riding the DeepSeek slipstream (I'm doing it again). That's kind of my point.

Just now, I made the mistake of sticking a UK lunchtime politics show on as I threw down some leftover Thai curry. Four 'politicians' were peddling ill-informed conjecture on the rise of DeepSeek to swerve a discussion to their particular agenda. And yes, I should know by now that’s what we call a mistake for political discourse in the West these days....

Suffice to say it wasn't the curry that made my noggin hit boiling point.

Everything they said, and to be honest, practically everything everybody else in the media has said about DeepSeek in the last 24 hours entirely misses the point.

It’s better.

That’s kind of it, folks, if you listen to the pundits on TV. “It’s better.”

It's Better. Haven't You Heard?

And I have ZERO doubt in my mind that the chatter across thousands of boardrooms and offices around the world, right now, is, “We need to use DeepSeek. It’s better.

Yup. It is a technological marvel:

  • Developed with a fraction of the budgets available to its US rivals, DeepSeek has built a model capable of ‘chain of thought’ reasoning, just like OpenAI’s o1 (o2 and o3) models.

  • It’s open source, meaning the model is accessible to anyone wanting to develop on and for it.

  • The app is free to use, and the APIs are drastically cheaper if you’re building solutions that call them. (Hey, but we all know that when a service is free, YOU are the product).

  • And it’s efficient: training and operating DeepSeek is vastly cheaper than its rivals

....and so in many ways, DeepSeek is astonishing.

Hmmm...But Is It Better For You?

But for you, right now, is it BETTER, as the ‘experts’ in the media would have you think? In my honest opinion, as an AI enthusiast helping companies get their head around this stuff:

  • I’d argue that if you haven’t yet found much use for reasoning models, it isn’t better for you.

  • I’d argue that if you’re not deploying apps that are built on APIs, it isn’t better for you.

  • I’d argue that if you value the option to protect your data from training models, it isn’t better for you.

  • And I’d argue that if you’re using the under-the-hood features that tools like ChatGPT offer, such as voice, true image parsing, Canvas for document and code editing, then it isn’t better for you.

The one thing we know about the development of AI tools is that change is constant.

It’s an arms race, where one vendor inches beyond another every week depending on any given set of metrics. And sure, DeepSeek has made a ‘big splash’ (sorry, I’m at it again), but from a business perspective, I worry that we get buoyed along (I’ll get my coat) without fully understanding the true differences between one platform and another, and in particular, what that means for us - right now.

Right. Back to my curry then.

If you’re curious about AI, and need help exploring how to make it work for your business, why not talk to me about the training I provide through my programme, The AI Advantage, described as ‘the best workshop I’ve even been on!

I work with clients through whole-team workshops through to one-to-one c-suite coaching, designed to immerse you and your team in how to get the most of the tools, and how they can fit into your day-to-day operations.

Interested? Let’s have that chat, one-to-one.

Read More