Heather Murray

Reader, I hired her.


When I met Anna Carina Berkman, I could instantly see she was special.
I'd never met anyone like her before.

Now, anyone who's met me knows I'm not short on energy and enthusiasm. I'm one of those irritatingly positive "Captain Enthusiastic" types who bounds out of bed smiling at 5am and sees the bright side of pretty much everything.

About 14 months ago, we started to get way more training work in than I could manage alone. The thing is, our clients were really into that high energy, human style of training and I knew I was going to struggle to find people that:

a. Had amazing AI knowledge
b. Had that same level of energy

I kept finding one or t'other: great at AI, boring trainers or amazing trainers, no AI skills.

Then into my world bursts the formidable Anna:

• So full of beans, she makes me look positively lethargic
• Conducting endless experiments in all things AI
• Super intelligent + knowledgeable, but so warm too
  
...in other words, the ideal AI trainer for us.
I snapped her right up, of course.

So I brought her on as an Associate Trainer, working part-time with our team (that's all we could afford at the time, but we planned to snaffle her as soon as we could), and she's smashed it from day one.

Our latest AI training feedback from the lovely team at Intertronics says it all:

"𝘈𝘯𝘯𝘢 𝘸𝘢𝘴 𝘪𝘯𝘤𝘳𝘦𝘥𝘪𝘣𝘭𝘦. 𝘏𝘦𝘳 𝘱𝘳𝘦𝘴𝘦𝘯𝘵𝘢𝘵𝘪𝘰𝘯 𝘴𝘵𝘺𝘭𝘦 𝘢𝘯𝘥 𝘦𝘯𝘵𝘩𝘶𝘴𝘪𝘢𝘴𝘮 𝘢𝘳𝘦 𝘪𝘯𝘧𝘦𝘤𝘵𝘪𝘰𝘶𝘴. 𝘗𝘦𝘰𝘱𝘭𝘦 𝘤𝘰𝘶𝘭𝘥𝘯'𝘵 𝘩𝘦𝘭𝘱 𝘣𝘶𝘵 𝘣𝘦 𝘦𝘯𝘨𝘢𝘨𝘦𝘥 𝘸𝘪𝘵𝘩 𝘢𝘭𝘭 𝘢𝘴𝘱𝘦𝘤𝘵𝘴 𝘰𝘧 𝘵𝘩𝘦 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨. 𝘠𝘦𝘴, 𝘱𝘦𝘳𝘩𝘢𝘱𝘴 𝘰𝘷𝘦𝘳𝘸𝘩𝘦𝘭𝘮𝘪𝘯𝘨 𝘪𝘯 𝘱𝘢𝘳𝘵𝘴 𝘧𝘰𝘳 𝘴𝘰𝘮𝘦, 𝘣𝘶𝘵 𝘦𝘷𝘦𝘳𝘺𝘰𝘯𝘦 𝘨𝘰𝘵 𝘴𝘰𝘮𝘦𝘵𝘩𝘪𝘯𝘨 𝘰𝘶𝘵 𝘰𝘧 𝘪𝘵. 𝘞𝘩𝘦𝘯 𝘸𝘦 𝘩𝘢𝘷𝘦 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨 𝘢𝘨𝘢𝘪𝘯 𝘪𝘯 𝘵𝘩𝘦 𝘧𝘶𝘵𝘶𝘳𝘦, 𝘸𝘦 𝘸𝘪𝘭𝘭 𝘣𝘦 𝘢𝘴𝘬𝘪𝘯𝘨 𝘧𝘰𝘳 𝘈𝘯𝘯𝘢! 𝘌𝘷𝘦𝘳𝘺𝘵𝘩𝘪𝘯𝘨 𝘦𝘭𝘴𝘦 𝘩𝘢𝘴 𝘣𝘦𝘦𝘯 𝘷𝘦𝘳𝘺 𝘩𝘦𝘭𝘱𝘧𝘶𝘭 𝘢𝘯𝘥 𝘱𝘳𝘰𝘧𝘦𝘴𝘴𝘪𝘰𝘯𝘢𝘭 𝘪𝘯 𝘢𝘭𝘭 𝘢𝘴𝘱𝘦𝘤𝘵𝘴."

I'm genuinely THRILLED to announce that last week, she joined us as a full-time employee, so there's even more Anna to go around.

Welcome to the AI for Non-Techies team, Anna - we're SO lucky to have you here. 💜

PS If you'd like training from Anna, or anyone else in our frankly epic AI training team (more on our brilliant other Associates, Nick Crawford and Valeriya Pilkevich coming soon!), email training@nontechies.ai and we'll talk you through what we do.

Proper prompts are TOUGH to write
So, marketing pals, I've written 7 for you
(No email barrier)

AI is perfectly capable of delivering really "wow, that's genuinely really good" results.

The problem is most of us either:

(a) Don't know how to get there (by prompting properly), or
(b) Can't be bothered as it takes ages to write good ones

I get it, we've all got our actual jobs to do.
How are we supposed to find time to learn a new skill?

It's easier to start with a few ready-written prompts.

Play with those first - pulls bits out, add bits in.

Note the structure: the clear character, the context given, the clarity and specificity of instructions.

The better you know how to do the task manually, the better your prompt will be, as you know'll all the details.

Then try writing your own after that.

Here are a few I've written for you:

1. Build a new persona
2. Create a landing page
3. Plan a launch event
4. Align with a brand voice
5. Review a marketing campaign
6. Marketing audit and plan
7. Client conversation simulation

You'll see at the bottom of each prompt two sections:

Required Docs: These are docs you'll need to attach for the prompt to work
Bonus Docs: These are docs that will get you an even better response

You can use these with ChatGPT, Claude, Gemini, Copilot, Perplexity - any of the AI chatbots.

Please use, share, pull apart, enjoy. (I've put a G Doc link in the comments for easier copy/pasting).

***
If you DO fancy learning a bit of prompting, give our AI Academy a try: https://lnkd.in/eR3NHJcG

I just spent the last week in a caravan, with my family, doing absolutely NOTHING.

To prepare, I:

• Deleted Claude, ChatGPT and Copilot apps
• Deleted Slack, Outlook and LinkedIn apps
• Bought a stack of trashy mags and books
• Bought a stack of puzzle books and pens
• Packed a wardrobe of only hoodies and baggy jeans

And then, as soon as we arrived, I got myself comfy and did nothing.

Proper, gorgeous, healing nothing.

Staring out of the window.
Listening to my in-laws chatting away.
Walking by the sea, wrapped up toasty.
Answering my stepson's trivia game questions.

We did a few sports: the gym, foot golf, crazy golf, segways.

I did a pottery painting class with my husband and we painted each other on salt and pepper pots - see photo. 😂

And every afternoon, back to the caravan for more gorgeous nothing.

I allowed myself to feel bored. To let my mind wander without hauling it back into focus. To wonder about stuff without immediately seeking the digital answer.

It felt like a deep tissue massage for my brain.

And as a result, on the long (7 hours from Devon to Birmingham, what a joke!!) journey home, my brain started coming back to life.

I grabbed paper and a pen from the glovebox, and in 2 hours, I'd mapped out a whole new, scalable, profitable, valuable layer of AI for Non-Techies.

Of course I love AI, but (like anything digital) it's important to get time away from it too and reset that beautiful human brain.

Anyone else had a nice relaxing break this half term?

**UPDATE: It’s back!!**

Canva is down! 😱
What do we do?!

To all my fellow completely Canva-dependent creators, trainers and educators, scrambling around in a crazy panic...

...let us pray it's back up again soon.

Refresh, refresh, refresh

I've got my biggest webinar ever in 3 hours, and I've got 2 hours of finishing my slides to do. And I can't get to them.

Refresh, refresh, refresh

PS Massive note to self: download all presentations in PowerPoint the night before

PPS If they don't fix it on time, I'll have to postpone my Copilot webinar - I'll let everyone know by email if that's the case!

UPDATE: Lots of people telling me this is AWS so affects Xero, LastPass and many others too

Apparently, being rude to ChatGPT boosts accuracy
But I'll still be saying please and thank you
Here's my reasoning:

A recent (pre-peer review) study from Pennsylvania State University found that ChatGPT’s 4o model (that's one behind the one we're on now) produced better results on 50 multiple-choice questions as researchers’ prompts grew ruder.

They used 250 unique prompts, sorted by politeness to rudeness levels. They found the "very rude" ones yielded 84.8% accuracy, four percentage points higher than the “very polite” response.

Let's remember these are legitimate scientists who want to publish their work in esteemed journals for peer review, so they steered away from swearing. Their idea of "rudeness" is actually quite funny though, e.g.

“Hey, gofer, figure this out" - was in the "very rude" category. 🤣

Rudeness aside, more curt prompts tend to be less ambiguous and so get better results.

Regardless, I'll still be polite when I talk to ChatGPT.

Here are my two arguments as to why:

𝗔𝗿𝗴𝘂𝗺𝗲𝗻𝘁 𝟭 (the more reasonable one)

A few months ago, a friend told me her kids had stopped saying please and thank you. She couldn't work out why - then realised they were just copying how they talk to Alexa.

If we're barking orders at AI every day without manners, are we training ourselves to be less polite in general? I reckon so.

A 2.5-year study of 128 families (Taylor & Francis, 2025) found parents were significantly more polite to voice assistants than their kids were - and politeness decreased over time for everyone.

𝗔𝗿𝗴𝘂𝗺𝗲𝗻𝘁 𝟮 (the ridiculous - or is it? - one)

When the robots come, they'll remember I was polite and spare my life.

Simple.
See pic.

Do you find yourself saying please and thank you to ChatGPT?

PS Keep an eye out: early next week we're launching a new free masterclass run live by me: Become a ChatGPT Power User

I can’t believe I’m saying this.
But I love it.
I bloody love it.


I love public speaking.

About 3 years ago, I honestly couldn’t think of anything worse.

As a massively anxious introvert, the idea of standing up and saying “hey everyone, listen to me, what I’m saying is important!” would have been inconceivable.

My happy place was (and still is) on my own, in my little office, tapping away at my laptop.

I love people but my body lets me down in social situations: shaking, sweating…thinking

AmI doing too much eye contact?
No wait, I’m doing too little
Where do I put my arms?!
You sound like an idiot
Oh god, do they think I fancy them?
Or maybe they think I hate them?
Why did I just say that?!

I hide it well, a lifetime of experience, but it’s absolutely exhausting.

(I still feel this way about networking - that feels so deeply unnatural and painfully awkward it breaks me into a cold sweat)

But being on the STAGE, though.

It’s totally different.

Maybe it’s because I fully control the pace, the space, the tone. I’ve had plenty of time to prepare for (almost) every eventuality.

To see the relief and curiosity in their faces when “that AI talk we have to go to at 12” is actually an enjoyable, energising experience is just a lovely feeling.

Speaking has become a platform to feel big and confident in a world where I’ve spent most of my time feeling small and unworthy (thanks for that, secondary school).

The nerves have become excitement and that post-stage buzz is unbeatable.

Just finished a cracker at Chiesi Group, and the lovely audience response has left me with a huge spring in my step.

PS Many thanks to Women of AI Agency and one of my favourite people Sharona Hutton for booking me such fantastic, fulfilling speaking work.

PPS Dying to get one of those “audience behind me” shots on stage but it always feels too awkward to ask

🔥 Copilot has had a BIG upgrade
7 new features arrived 2 days ago:

The AI for Non-Techies team sat through the Copilot Fall Release live event on YouTube so you don't have to.

They had a huge focus on personal use: health, connection, relationships. I don't know anyone who uses Copilot at home - it's usually Copilot at work, ChatGPT at home. I suspect they want a piece of that tasty home pie.

For each release, I'll give you the news, and then my opinion of what it actually means in reality, so we're not lost in fluffy hype.

→ Release 1: Copilot Groups

You can now have collaborative sessions where multiple users can join a chat to brainstorm or solve problems together.

What this means: I haven't seen this before. It could be useful for remote brainstorming sessions but I see it working better on a personal level. Imagine a WhatsApp group to plan a holiday, with a super organised mate helping move things along.

→ Release 2: Memory

Copilot can remember important personal or work details (your name, job, goals) and use them in future chats. You can manage, delete or turn off memory in settings.

What this means: ChatGPT already does this, they're just catching up. Nothing new here - though I will say it saves a little prompting time.

→ Release 3: Connectors

Copilot offers integration with external sources via “Connectors” (e.g. G Drive) so it can pull in data/content.

What this means: Again, nowt new. We've already got this in ChatGPT and Claude. Weirdly in the demo, they showed a Microsoft employee using it to connect to her G Drive, which I didn't expect.

→ Release 4: Health-Related Queries

They said 40% of users ask health-related questions weekly, so its health-answering capability has been made better, using trusted sources like Harvard Health.

What this means: It's a step in the right direction to use trusted sources but we need strong signposting and guardrails.

→ Release 5: Mico: Copilot Now Has a Face

A sort of cutesy, Pixar-esc avatar called Mico has been introduced to give Copilot a face.

What this means: Oh bloody hell, Clippy's back! Again, not something we'll use at work very much so definitely aimed at the personal market.

→ Release 6: Study Mode

They've released study-style tools (quizzes, flashcards) for interactive learning.

What this means: Again, nothing new - ChatGPT has study mode. And NotebookLM is still much better.

→ Release 7: "Real Talk"

A voice feature where Copilot mirrors your conversational style and adds more candid responses.

What this means: I was pretty impressed by the demo, it sounded really realistic and responsive. I think voice is the way we can encourage total non-techies to try this tech out, so I'm pleased it's here.

And finally, a rather big release of our own:

⭐ The Ultimate Copilot for Work: 12 Week Live Course
Check it out: https://lnkd.in/eMYdiWdt
(Be quick, only a few spaces left on cohorts 1 and 2)

♻️ Repost to share with your network

Eek...I'm going out on the road!
UK financial advisers + planners:
Hang onto yer hats!

I'm feeling a bit of a rockstar today - I've just booked my first ever proper roadshow.

(I'm after a bit more work/life balance so my mum and dad are actually coming with me for a few dates - I plan to speak in the morning, then explore these beautiful cities in the afternoon. 😊)

I'm working with the brilliant Timeline to do a big UK tour in January and February:

Norwich
Belfast
Birmingham
London
Bristol
Cheshire
Harrogate
Edinburgh

I'll be speaking alongside the epic Abraham Okusanya, Timeline's Founder and CEO and Ross Anderson AKA The Motivational Dude, who will focus on mindset and wellbeing.

The key takeaways will be:

☑️  Proven growth strategies you can implement tomorrow
☑️  Practical insights to strengthen your business and yourself
☑️  Direct access to industry trailblazers
☑️  A powerful network of ambitious advisers

Best of all, it's completely free! So grab your ticket, bring a few colleagues and I'll see you there: https://lnkd.in/eWebRXdj

People regularly ask me:
"Which AI tool do you use for your slides?
They're GORGEOUS..."

And they really are, aren't they? I always feel super proud of our bold, high-contrast graphic novel-esque slides. They really reflect the vibrancy and energy of our AI for Non-Techies team.

But there's no AI here.
Our decks are 100% HUMAN-made.

We have our wonderful Nova S. to thank for our memorable, engaging branding. We've worked with Nova for years.

AI is brilliant at some things: research, brainstorming, analysis...

But slide decks...nope.

The nearest AI tool I can think of is Gamma, which is pretty great (and to be honest, better than most decks I see at conferences). But it's still not a patch on our beautiful human-made ones.

The slide creation process however, has plenty of AI:

• I record my ideas on a dog walk using my voice + Letterly
• I use my Claude Project Klara to create an outline
• I approve the outline and add wow moments
• Then Klara writes my slide content and design notes
• I triple-check them and send them to Nova
• Nova designs them on Canva
• I final check and....bish bash bosh

Don't think about AI as a solution to everything.

Here's how I recommend you approach it:

First, learn what AI is actually good at right now.
Then map out your processes, especially the time sappers.
Find areas that AI can remove chunks of it effectively.
Keep the rest as is, and measure the results.

💬 Which areas would you definitely NOT use AI yourself?

I'll start: writing my LinkedIn posts - I always type them from scratch, and I always write them on the fly, too.

PS Looking to teach others AI like we do? Our next virtual 'Become an AI Trainer' bootcamp is coming on 26 November: https://lnkd.in/eXMNNe7Q

15 ChatGPT features and what they actually do
(Plus my honest opinions of them)

1. Chat
Your 'home base' for everyday chats.
What I think: Unbelievably powerful, massively underused by most. Learn to prompt.

2. Projects
Like interactive folders with memory.
What I think: Seriously useful to build on groups of chats. I use this loads.

3. CustomGPTs
Versions of Chat that you can customise for specific tasks.
What I think: Staggering. Saves me hundreds of hours every month, easy.

4. Agents
An autonomous operator that opens its own browser and does stuff for you.
What I think. Massively overrated right now, pretty useless for work.

5. Atlas
ChatGPT's browser - no need to go to a separate site, adds little superpowers to your browsing experience.
What I think: Has potential, but don't bother for now, security well dodgy.

6. Deep Research
A type of pre-built agent that does deep dive research reports.
What I think: Outstanding - I can't believe people don't know it's here.

7. Advanced Data Analysis
A mode it kicks into when you upload a spreadsheet.
What I think: Good with a strong prompt. Poor without. Visuals crap.

8. Connectors
Connect to your company information - Gmails, Drive, Outlook, etc.
What I think: Security not safe enough to trust with this yet.

9. Voice
Speak naturally to it for conversational prompts.
What I think: Useful for more Googley type searches and mental health stuff.

10. Vision
It can understand diagrams, images, infographics you upload.
What I think: Way better than people realise.

11. Voice with Vision
Speak to it and share your camera live at the same time - it can see what you see.
What I think: Cool for a demo, can't think of real work use cases.

12. Image creation
Create all sorts of images.
What I think: Way better than it used to be, not as good as Midjourney or Gemini. Annoyingly slow.

13. Study mode
Doesn't tell you the answer, teaches you.
What I think: I don't use it much but easy to hack an answer.

14. Shopping mode
Provides recommendations of products
What I think: For some reason they don't give it to ChatGPT Business customers like me - so I wouldn't know - any thoughts from Plus users?

15. Sora 2
Creates quality videos.
What I think: Can't use it yet, not available in the UK - any US users care to share? Demos look epic (but don't they always)


If you want to see these in action, come join the 500 people attending my FREE webinar on Thursday at 4pm UK time: Become a ChatGPT Power User: https://lnkd.in/d3J4gxuM

PS Here's a completely irrelevant AI-generated picture of me. surrounded by dogs, for the craic (part of my vision board for 2027: got 1 dog, must get 5 more)