CEO: Brain eBox | Workplace Learning Solutions | Host of the HR on the Brain Podcast | CEO: Lead with Sheree | Speaker | Passionate about developing leaders | Get in touch: sheree@thebrainebox.com

You’ve built a reputation for delivering great L&amp;D work - facilitating training, designing programs, and running onboarding. People trust you to get things done.But to grow into your next role, it’s not just about great work anymore.Now, it’s about influence, alignment, and leadership.You’ve got great ideas, but they’re not always landing.You want to be seen as a strategic partner, but you’re still treated like a service provider.You’re ready to grow, but what you need isn’t tucked inside another course.You need real-world support from someone who understands what L&amp;D excellence looks like - and how to lead it.That’s where I come in.I’m Sheree Comninellis, L&amp;D Executive and mentor for emerging L&amp;D leaders.I help L&amp;D professionals like you step into leadership.If you’re ready to lead, start here:📬 Executive Moves — my 5-minute weekly leadership edge for ambitious L&amp;D professionals.Each edition gives you:📌 1 bold insight to reframe how you lead📌 1 practical move you can apply immediately📌 1 growth resource to help you go deeper➡️ Get Executive Moves here: https://dashboard.mailerlite.com/forms/589169/98676779613423573/share.💡 Want to go further?Join the waitlist for The L&amp;D Professionals Collective - a monthly membership that helps you grow your confidence, visibility, and strategic influence, so you can be recognised as ready for L&amp;D leadership.Inside, you’ll get monthly live sessions, coaching, and practical tools designed to help you bridge the gap between delivering learning and leading it.➡️ Learn more and join the waitlist here: https://leadwithsheree.com/ldprofessionalcollectivewaitinglistMore about me:I live in Cape Town with my wife and beautiful children. I love historical fiction, good coffee, and my all-time favourite series is Gilmore Girls.

Sheree Comninellis (she/her)

Product Manager @ Flexa Careers

I thrive in messy and ambiguous situations where you need to move fast and figure things out as you go. An eclectic background demonstrates my aptitude to up skill fast, my appetite for growth and my ability to adapt.
At UnlikelyAI, I lead client engagements for neurosymbolic AI solutions in financial services, accountancy and insurance where the stakes are too high for the probabilistic outputs generated by LLMs. I run discovery, scope technical solutions and translate between what the technology can do and what the business needs. 
Before this, I was Head of Product at Flexa where I owned the full B2B2C product lifecycle. I've founded two businesses and have a PhD in exercise physiology - the analytical rigour carried over even if the domain didn't!

Dr. Ailish McLaughlin

Solutions Lead @ UnlikelyAI | Enterprise AI for Regulated Industries | Product and Technical Pre-Sales

We’re delighted to welcome Dr. Ailish McLaughlin to the SoundBoard Advisor Panel.

Ailish’s career journey is anything but linear - from exercise physiology and working with Olympic athletes (and competing internationally herself), to founding multiple health businesses, and now pivoting into tech with a strong focus on helping individuals and organisations navigate and implement AI.

It’s a rare mix of deep technical expertise, lived experience, and genuine curiosity about what’s next.

Here’s what Ailish shared with us on why she’s excited to join the SoundBoard community:

“Over a long and winding (and eclectic) career, the people who took time to advise me, make introductions, and help me figure out my next steps have been instrumental. Soundboard is a chance to return the favour. Linear careers are no longer the norm - and I'd love to offer what I've learned to others navigating the twists and turns of their own.”

Welcome, Ailish - we’re so glad to have you!

Dave O' Mahony Eimear O'Halloran MBA

AI prompting is dead.

Not even six months ago, I had a whole framework for writing the perfect prompt. I even shared it in AI workshops I ran during the summer.

Then last week, a friend said she felt like she wasn't very good at prompting.
And I found myself saying: just give it context about what you're trying to solve. Then let it ask you the questions it needs.

And I realised that how I engage with AI has completely changed since I held those workshops.

It got me thinking. Is this because my understanding of how to use AI has evolved? Or has something fundamentally changed in the tech?

Turns out, it's mostly the latter. Models have genuinely improved at understanding what you mean without needing rigid prompts. Anthropic even talks about the shift from "prompt engineering" to "context engineering"; it's less about finding the perfect words, more about giving it the right information.

Six months ago, prompting felt like a skill to master. Now it's your ability to brief well that matters.

The shift from "ask AI questions" to "get AI to ask you questions" is small. But it changes everything about how the interaction feels.

Another reminder that the way we use these tools is evolving as fast as the tools themselves.

Everyone's talking about AI replacing jobs or creating new roles.

But I think the biggest change will come from something else entirely: speed.

Anthropic built and shipped Claude Cowork in 10 days. Using Claude Code.
That's the new baseline.

AI will compress timelines in ways we've never experienced. 

But human brains haven't evolved to match that pace.

"Sleeping on it" is a thing for a reason. We need time to process, to let ideas marinate, to make good decisions. Our best thinking often happens in the gaps between the work.

When AI compresses those gaps to almost nothing, we lose something. And I think we'll see two things happen:

🤿 Stress massively increases. The speed will bring a constant feeling of being underwater, where no one can ever quite catch up.

🦥 Poor decisions and badly thought out plans. Because speed often doesn't equal quality. Sometimes slow is how you get it right.

But I think this will also eventually lead to another evolution.

We saw a resurgence for in-person time after everything could feasibly be done online. People craved what they'd lost.

I wonder if we'll see the same thing here. Purposeful slowing down. Thinking time built into delivery plans. Strategic pauses as a feature, not a bug.

What do you think?

Americans spend more on OnlyFans than ChatGPT and the New York Times combined.

I know nobody who (openly) admits to using or paying for OnlyFans. But almost everyone I know uses and pays for AI.

This graph was a great reminder of the echo chamber I exist in.

It's also a reminder of what actually drives human behaviour. Research consistently shows that our deepest psychological need is connection. Social isolation is more harmful than alcoholism or smoking. And s*x often isn't really about s*x; it's about reducing alienation and loneliness through physical connection.

So it shouldn't surprise us that people turn to technology for connection. P*rn, social media, AI companions - the delivery mechanism changes but the need stays the same.

With social media, we didn't act fast enough to protect vulnerable users. That need for connection made people easy to exploit. We're already seeing echoes of this in AI - companions designed to keep you engaged, responses tuned to tell you what you want to hear. But this time, we have a chance to build it differently.

I used the Claude Chrome plugin to build an app a few weeks ago. It took a few hours.

I didn't question it. I just assumed this was the cost of using AI agents. Building things takes time. Makes sense.

Then I tried Claude Code. And did something similar in minutes.

How? Why?

Listening to the hard fork podcast the other day I learned that the Chrome plugin is essentially blind.

It works by taking screenshots of the webpage, analysing what it sees, identifying which pixel to click, then performing the action. Every single interaction requires this loop. 

Did everyone already know this? I didn't.

Claude Code works directly in the terminal. It reads and writes code as text. No visual interpretation needed. No pixel hunting. It just does the thing.

Another reminder that understanding how this stuff works impacts how you use it.

OpenAI just announced ads are coming to ChatGPT's free and Go tiers.

Cue the outrage. 

But is this a betrayal? Or a clarification?

AI is a (very) helpful tool. But it is built by companies who have incentives.

Some estimates project OpenAI to have >$115 billion in cash burn by 2030. They've committed $1.4 trillion to infrastructure. That money has to come from somewhere.

This is actually useful information. It's a healthy reminder that understanding what the business who builds the technology cares about matters.

Because when you understand what a company is optimising for, you can use their product more strategically. You know what they want from you. And you can decide what you want from them.

I'll continue beating the drum that AI is incentivised to capture your attention. Because this is crucial for the companies that make it to win the AI race. And now, with ads, they can monetise that attention directly.

We talk a lot about understanding AI. But we should be talking just as much about understanding the businesses that build it.

I genuinely believe we're heading towards a future where apps as we know them no longer exist. One super interface, customised for you, where you do everything.

And that excites me.

But Clawdbot, the latest explosion in AI, just gave me a glimpse of how messy the road there is going to be.

It's an open source, proactive AI assistant that connects to your email, calendar, Slack etc. It has full terminal access to your machine and you talk to it through WhatsApp or iMessage. It's been described as "what Siri is supposed to be." And if you know me at all, you'll know properly AI enabled Siri was on my 2025  Christmas wish list!

But there's no company behind it that must comply with security laws, there's no accountability. And it's most useful when you give it access to everything.

Security researchers have already found 900+ exposed instances leaking credentials. The creator's own docs say: "There is no 'perfectly secure' setup."

I would love to try it. But I know my limits. I can't set up a sandboxed environment or understand the technical documentation well enough to harden it properly. So it's too risky for me right now.

And this made me think about how much trust we place in companies and governance. Sometimes that trust is misplaced. But it definitely offers more protection than this.

En route to an exciting future, there is so much to figure out. And so much potential for disaster.

Hugely exciting opportunity for my product network to join our very talented team on a very exciting mission as an IC.

Ethan Mollick recently shared a study saying role prompting (telling the AI to act as a data scientist, copywriter, etc.) had no impact on the quality of output.

Meanwhile, Parth Patil and Reid Hoffman were on a podcast talking about how assigning roles to AI agents gives you wildly different outcomes and perspectives. Parth even generated 100,000 unique expert personalities to tap into different viewpoints.

So which is it? Does role prompting work or not?

I think the reality is that there are no blanket truths in AI yet.

The tools are changing faster than anyone can keep up with. What worked last month might not work this month. What works for one person might not work for another.

All we can do is keep experimenting, stay curious, and decide for ourselves what actually works. There's no "right" way to AI....

The future is fully customisable. But what if we don't actually want that?

I've been building apps and tools with Claude Code. They're pretty good. They mostly work. But I've noticed I'd rather experience friction when using them than spend time perfecting them.

And this made me think about the friction baked into the software we use today. The apps and tools that aren't designed specifically for us. They're built for everyone and as a result we have to bend to fit them.

The promise of AI is that everything becomes fully customisable. Built just for you and perfectly fitted to your workflow.

Which, in theory, sounds amazing.

But do we even have the cognitive bandwidth to know exactly what we want? Every decision about what it should do is another decision we have to make.

Decision fatigue is real. There's a reason Steve Jobs wore the same outfit every day - he was reducing meaningless decisions so he could focus on the important ones. 

Maybe the best software does the same thing and customisation won't provide the freedom we think it will.

AI can now read, sort, and respond to your emails for you.

It’s one of the things Clawdbot (OpenClawd? Moltbot?) gets praised most for.

But what if there’s a different way…..? 😅

taplio

The Battle

Sheree Comninellis (she/her)

Dr. Ailish McLaughlin

Followers

Posts

Engagements

Likes

Performance ScoreDr. Ailish McLaughlin's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Sheree Comninellis (she/her)

Dr. Ailish McLaughlin

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Dr. Ailish McLaughlin's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDr. Ailish McLaughlin's LinkedIn Performance