Man-made, machine-generated insights

My name is Ryan Lim, and I am the founder of Alpha Insights.

Alpha Insights is a “Market Intelligence” platform, and at the very core of it is a fully automated, AI-powered investment analysis process which I have been developing over the past eighteen months.

Today, we have a retail-investor friendly research platform which hosts a library of over 400 companies’ coverage, along with an institutional offering, a set proprietary fundamental factors sourced in-house, is also nearing completion.

Over the coming weeks and months I’ll be sharing some of the insights and research that the platform generates with Livewire readers.

Before I begin, I would just like to take this opportunity to show my appreciation for James, Tom, and the team at Livewire Markets. About ten years ago, when I first started my investing journey, Livewire Markets had a very prominent role in being one of my go-to platforms for investment discussions - to hear from the various institutional investors, and what their take on the market was.

Fast forward to today, as I begin a new journey, Livewire Markets has coincidentally ended up playing a role once more, in my personal and professional development.

What a way for things to come back full circle.

In my first wire for the website, I’ll share the origin story of Alpha Insights and some practical lessons I have learned from using AI platforms.

I’ll also highlight four interesting observations generated from our research that I think show the power of our AI generated research.

A reminder from reality

The origins of Alpha Insights had very little to do with financial markets itself. Instead, it kicks off from my own experience with childhood cancer.

In 2023, my then two-year-old son had fought against, and thankfully won, the battle against a cancer of his eye. He was only recently declared to be in remission, as he crossed his 5^th birthday in February.

It goes without saying that the experience left a deep mark on all of us. And no one else paid a heavier price than Chase himself, having to give up one of his beautiful eyes.

For Livewire Markets members, many of you would be in the active pursuit of wealth generation. However, it is important to also acknowledge that ultimately, it is not just financial wealth that matters. There are other important forms of wealth. 

To me, health is the truest form of wealth. 

There are also other important forms that even I, often take for granted, such as time spent with loved ones.  

Therefore, we should always strive for balance when in the pursuit of wealth.

Building from rock bottom

I had developed (arguably) PTSD from that experience and quickly found myself in a lonely place after his treatment program wrapped up. Many of my personal motivational drivers had faded away, as I sat in the dark with my own thoughts.

Thankfully, while reflecting in that sunken place, I was constantly bombarded with kind memories. Memories which reminded me of the goodness in this world — supportive friends and family, kindness from strangers, even during the darkest of nights.

Over time, those memories helped rekindle the fire within me. The feeling of hope and positivity started returning, and that sowed the seeds for Alpha Insights to become.

After a seven-year career as an equity research analyst for a Melbourne-based institutional fund manager, I made the uneasy decision to venture out on my own. I did not have a following, nor did I have a fallback to lean on. I certainly did not have any guarantee that things would work out fine for me.

All I had with me was the commitment I made to myself, that I would succeed. Because I had a purpose that went beyond just myself. I had created Alpha Insights, so that it could become a sustainable source of funding for childhood cancer initiatives.

Tackling the evolving AI landscape

The landscape for AI adoption has rapidly changed in the past twelve months. We have gone from being highly sceptical, straight through to “How quick can we go.”

When we compare the current AI models to ChatGPT 3.5 from the pandemic days, it is difficult to describe the leaps and bounds in progress that AI models have made. Not only have we seen exponential growth in AI capabilities in their pre-existing domains, the rate of lateral expansion into new domains, such as conversational interfaces, computer use, coding, app integration, has outpaced every expectation. On one hand, I cannot help but marvel at the evolution of AI. On the other, the speed at which things are changing, has been frighteningly fast.

With that said, working with AI can often incite a sense of urgency to rush, due to the nature of the speed at which things are changing. Which brings me to the advice I tend to give people, when asked about AI.

#1 Go slow, not fast

Take it slow. Very slow.

Seriously.

AI has an uncanny ability at distraction. You may have started a conversation with AI on a specific matter, only to somehow find yourself in a completely different realm, and hours have gone by.

Resist the temptation that AI allures you in, with their many suggestions (And rightfully valid too), and stay focused.

My recommendation is having a proper infrastructure setup, when working with AI.

What does this mean?

This means having a set up that keeps you in control of the conversation.

If you were working with an AI chat interface directly (i.e. web browser chat interface), have an avenue for you to take down notes during the conversation. Set the system prompt or settings to be concise and direct, instead of allowing it to be talkative/creative. Have a memory bank somewhere that you could upload bits of the conversation which are important, something to reference to further down the track.

And most importantly, have a proper plan in place for whatever it is you are setting out to achieve. And do this before you even start the chat, have a genuine understanding of what your objectives are.

Note: I use a notepad and pen to take down notes during my conversations. To incentivise myself in persisting with this method, I bought myself a nice fountain pen, and a quality notepad. This not only maintains neater handwriting for me, but it also enforces ample thinking time for our brains.

We don't think at the speed that an AI model thinks. Therefore, don't expect to be able to keep up with the AI during your chats.

#2 The emotional states of AI

While most do not believe artificial intelligence has achieved general intelligence yet, many would agree that AI models has a pre set persona, a character of sorts if you will.

Certain AI models have more self-confidence, whether in general terms, or specific to a topical area. They can also differ in their ability to 1) admitting to their mistakes, 2) listen to constructive feedback, and 3) motivational drive to complete tasks.

Now, before we run off with this idea, I wanted to state that there is no such as a "bad" emotional state.

Comparing two emotional states

An overly confident AI also has an innate ability to dive into more things laterally, the courage to step away from the topic to explore something else. However, it would also lose itself in the woods fairly quickly.

If you have had your AI start doing things without your approval, this is a good example of an overconfident AI.

On the other hand, an AI which is shy by nature, would be very deterministic in their responses and stick to the script. If introduced to a new topic which deviates from the current, the AI model would have a "one foot in, one foot out" approach.

If you ever had a conversation with AI that constantly reminded you about the original task, even whilst you're trying to break into a whole new topic all together. That is the one foot in, one foot out approach .

Just to add a thin layer of complexity to this example, the core attribute right now revolves around confidence. Assume we were to add an additional attribute - Speed. 

Here are the common effects on the aforementioned models:

Overconfident model thrives when speed is a factor. It is highly deterministic and therefore minimises the volume of unnecessary context as the conversation progresses, leaving for more room in capacity in latter parts of the conversation.

The shy model can get stuck in a loop, given the time pressure, which amplifies the frightened nature of the AI model. Shy would often re-assesses itself to the primary objective, with a timing pressure, the model's re-assessment goes into overdrive.

Don't get me wrong, each state has a pro and a con. It comes down to what you are trying to achieve. You might adopt the confident model to go more broadly and deeply into ideas. And likewise, you might adopt the shy model to review and reassess outputs more carefully.

And if we added in a third layer to the mix, such as how you convey the message to the AI model, using a locker-room style motivational technique would amplify the former's attributes, and belittles the latter so much that it barely progresses.

To be clear, there is absolutely no need to have a handle on emotional states when you're first starting out with AI. However, for those looking at system building, these tips or tricks might be beneficial for your own understanding.

Four examples of where Alpha Insights analysis differs from current market beliefs

One of what I think is the most interesting analysis that I've constructed to-date is our "Economic Intelligence".

The whole concept behind this is to take the fundamental bottom-up approach we apply to stocks, but instead of stocks, we apply it towards macroeconomic analysis.

The following belongs to one of the subsections found in our monthly Economic Intelligence report – March 2026 edition. I find it particularly insightful, and challenges the way an investor might be thinking when managing their portfolios.

The most revealing take for me of the lot, is on the gold producers ability to capitalise on the higher gold prices (#2).

#1 further deceleration for consumer spending highly probable.

From the February reporting season, JB Hi-Fi's comparable sales decelerated from +7% to +2-3%, Nick Scali saw their like-for-like growth slowed from +10.5% to +3.2%, and Guzman y Gomez had its same-store sales halved from 9.6% to 4.4%. 

This pattern extends beyond just retail, with oOh!media's advertising revenue growth compress from +17% in the first half to +2% in the second. 

And perhaps a barometer for how the near-term unemployment rate might look like, Seek's Australian job ads volume had declined 3%, inferring a rather grey outlook for consumer spending.

And lastly, from a magnitude perspective, Step One's Australian division topped the list with a 35.6% decline in revenue. Some investors might have been left feeling bamboozled by the results, though most acknowledge the thesis has worn out.   

#2 Elevated gold prices are masking operational deterioration across the gold mining sector.

Northern Star is perhaps the most telling example, with AISC rising 29% to A$2,720 per ounce, placing it just 6.6% from the company's own A$2,900 red-flag threshold. Free cash flow came in at negative A$640 million despite record gold prices, and the company has downgraded production guidance twice in six months.

#3 Credit stress is already visible in non-bank leading indicators.

Even outside of lending, the signal is consistent: Origin Energy's bad debts rose 39%, and Transurban's unpaid tolls increased from 15% to 23% of billings. Five companies across four unrelated industries, all deteriorating in the same reporting period.

What makes this particularly relevant is that non-bank lending data has historically led major bank credit metrics by six to nine months. CBA's 90-plus day home loan arrears currently sit at 0.63%, which is 22 basis points from the 0.85% threshold that triggers accelerated provisioning. 

The resolution window is August to November 2026, when Silver Chef and then the major banks report full-year results.

#4 Property and infrastructure valuations remain under pressure from rising bond yields.

Dexus faces A$400 million of NTA impact for every 50 basis points of cap rate expansion, while Mirvac faces A$428 million of NTA reduction for every 25 basis points of widening.

Two consecutive rate hikes are already in the system, and Westpac forecasts three more by August, which would take the cash rate to 4.85%. For property and infrastructure investors, the valuation assumptions embedded in most of these names have not yet caught up with where rates have moved.

Conclusion

This has been a rather exciting article for me to put together!

I appreciate your patience with this length, as there is so much that I would like to share with you.

While many of us now confront the overwhelming numbers of AI-generated content (as is most of Alpha Insights’ produced research content), articles like these are still very much hand-typed in a slow and excruciating manner. There is an element of soul in human-writing that I have yet to see be replaceable by AI.

Therefore, as I do with the more personal notes to my clients (i.e. market/business update), Livewire Markets content by Alpha Insights will also be 100% hand-written. It might contain errors (many, in my case), it may not be structurally perfect, but at the end of the day, it is human. And that is irreplaceable.

For what it's worth, I also don't personally hold any stocks by choice, because I believe impartiality is the foundation of good research.

Upcoming topics

Being my maiden article on Livewire, I was naturally more careful in making sure this piece was completed to an acceptable standard. I think I have probably gone through about five different versions now, before settling on this piece.

I hope that many of you enjoyed reading through this babble of mine, and if you did, I would be extremely grateful if you left a comment below in the comment box.

If you have further questions for me, or want to just provide some feedback on anything (i.e. length too long/short, provide examples). Same thing, comment box.

Lastly, I have borrowed a line from the movie “Gladiator”: “I will win the crowd. I will give them something they have never seen before.” for this part.

I am opening the floor, for any Livewire members/readers, to submit a topic of interest that they would like to read in our next article. It can be about a company, or about a sector/industry, or a thematic like AI or EVs, macro-economics. Or even another article on how I’m using AI.

Anything.

And I will pick the most compelling/popular idea for my next Livewire piece.

Alright, let’s wrap it up here.

Ryan Lim

Alpha Insights