How do you see the AI model or platform you are using?

The AI Chronicles - notes from an investment consultant

Artificial intelligence is changing how we work, live, and do business. We cannot simply “hide” from it. AI is already making everyday tasks easier, and its influence on financial analysis, portfolio management, financial planning, and investment advice will continue to grow—whether we welcome it or not.

None of us at Tricio are programmers or software developers. But as investment buddies to professional investors, we need to understand how AI can be used across different areas of investment practice. That means exploring the tools ourselves, testing their practical applications and learning from the experience. By doing so, we can help professional investors navigate their own AI journey with greater confidence.

In this regular blog series, I will share observations, practical experiences and lessons from my own journey with AI. I'll also share articles and resources that can help professional investors use AI. The aim is not to explain the technology in technical terms, but to explore how busy professionals can apply it in the real world.

My friends - not all of whom work in finance or investment - clients and professional investors all use AI differently. They use it for everything from drafting emails and letters, reviewing legal documents, planning holidays, organising calendars and files, and helping with simple coding, to generating pictures and many other tasks.

Two discussions often arise. First, these AI tools don't always work well because they sometimes produce false or unrealistic results. Second, we seem to spend more time than expected trying to get AI tools to work properly.

Many technical experts would argue that we should learn how to write better prompts or even learn some coding. I think something more fundamental is at play than the technical side. That starts with how we see these AI tools and what they do for us.

We often call these AI tools “AI” or “AI models”, but what are they really? Perhaps we should ask the AI tools themselves.

Perplexity describes itself as an “AI assistant”, a “large language model-based AI assistant” or a “generative AI chatbot”. Gemini describes itself as a “conversational AI platform”. DeepSeek gives me a long list of answers, but none of them is definitive. It simply emphasises that none of these AI tools is “Artificial General Intelligence (AGI)”.

Claude gives more detailed and technically better answers:

§ Gemini and Claude are both general-purpose assistants built on large language models;

§ Perplexity is described as an AI-powered answer engine or search engine, so it is particularly useful for web searches;

§ Grok is a real-time assistant that deals with fast-moving events, naturally reflecting its close connection with X;

§ DeepSeek is built around efficiency, so Claude cannot really find a description that matches the others. Then again, you might expect Claude to say that.

So, they are assistants. Unlike some people who treat them as virtual buddies, I see them as virtual assistants.

Just like hiring a human assistant, you should have a clear view of what an AI assistant is set up to do for you. This will affect how you use it and what it can help you with.

Also, these AI assistants may sound intelligent, but this comes from the training they receive and the information they learn from. Much of that information comes from the web. As many commentators have pointed out, the web also contains false and inaccurate information.

Personally, I treat AI tools as fast-learning assistants that can search the internet and perform tasks on computers at astonishing speed. But they are also inexperienced, with a limited understanding of some real-world intricacies.

This means that when you ask AI assistants a question or ask them to carry out a particular task, you should keep asking them to check the latest facts on reliable and official websites. For research, regulatory and legal questions, I go further and ask them to ignore social media posts and messages from influencers. Sometimes it's worth checking the same question with different AI assistants. You may end up with something new, interesting and unexpected.

It also means that these AI tools assist you; they do not replace the real-life experience of professionals, including you. Most AI tools include disclaimers when you ask them to carry out legal, tax, regulatory or financial analysis. I use AI tools to help me formulate better questions before I speak with a specialist for advice. This saves me time and, therefore, cost. But it also encourages me to think and prepare more carefully.

Lastly, using AI is similar to asking a young, inexperienced team member to carry out a particular task. You need to be explicitly clear. The context and the desired outcome are always important. AI tools are smart, but they are still not as capable as a human being who can infer meaning from your messages, unlike a team member who has worked with you for years and understands your communication style.

So, back to the time question pointed out at the start of the blog: be prepared to spend more time organising your thoughts and providing sufficient detail before writing a prompt. Yes, AI tools can eventually save you time. But first, you need to invest the necessary time and effort to make them work for you.

James Chu, CFA

Head of Investment Solutions