{"id":8412,"date":"2024-08-12T14:23:11","date_gmt":"2024-08-12T14:23:11","guid":{"rendered":"https:\/\/tup.kxe.temporary.site\/meet-my-research-team-gemini-chatgpt-and-perplexity\/"},"modified":"2024-08-12T14:23:11","modified_gmt":"2024-08-12T14:23:11","slug":"meet-my-research-team-gemini-chatgpt-and-perplexity","status":"publish","type":"post","link":"https:\/\/okdesign.ca\/fr\/meet-my-research-team-gemini-chatgpt-and-perplexity\/","title":{"rendered":"Meet my research team: Gemini, ChatGPT, and Perplexity"},"content":{"rendered":"


\n<\/p>\n

\n
\"Gemini\"\"Gemini\"<\/figure>\n

Over the past weeks, I spent hours in front of the TV watching endless commercials punctuated with short periods of sports coverage. Is it just me, or were there more commercials than usual during these Olympic Games?<\/p>\n

I saw the Google Gemini and Microsoft Copilot commercials what seemed like a thousand times. I was completely sucked into the vision of typing in a simple request and Gemini or MS Copilot creating a meeting summary, action plan, and a visually compelling presentation all with the click of a button.<\/p>\n

I regularly use generative AI to polish content, assist with writer\u2019s block, and help with creating presentations (I\u2019m a big fan of gamma.app<\/a>) and have found it both effective and a great timesaver. Recently, I\u2019ve begun using it for research-oriented tasks and have found it to be really helpful in two key areas: product discovery and company\/product positioning. In both cases, not only does it save me a tremendous amount of time, but the end results are better than what I normally achieve doing the work manually.<\/p>\n

When I\u2019m using generative AI for research work, I\u2019ve found that I get the best results by using multiple products simultaneously. I typically have tabs open for Gemini<\/a>, ChatGpt<\/a>, and Perplexity<\/a> and work across all three simultaneously. I haven\u2019t used Copilot<\/a> yet, so don\u2019t currently have that on my list, but it is certainly another viable option. It doesn\u2019t matter which tools you choose to use; what\u2019s important is to use multiple tools to ensure optimal information quality and make it easy to identify hallucinations.\u00a0<\/p>\n

Product Discovery<\/h2>\n

It\u2019s no secret that the MarTech landscape contains a huge number of products. The recent MartechMap<\/a> shows some 14,000. It\u2019s unlikely that this will shrink any time soon; indded, it is likely to keep growing as new categories and products emerge. Finding the right products for a martech stack is hugely challenging for several reasons, two of which directly relate to product discovery:<\/p>\n

    \n
  1. Vendors are notoriously bad at providing a concise description of what their product is and does, which makes it difficult for a technology purchaser to easily hone in on a list of products that might be of interest. In fairness, some are better than others; Knak<\/a> is a good example. It makes it very clear on their homepage what its product does. But for most, you need to triangulate feature descriptions against solution overviews and company profile information, which is time-consuming and frequently frustrating.<\/li>\n
  2. Users tend to think in terms of use cases, and vendors generally think in terms of features, which can cause a complete disconnect.\u00a0<\/li>\n<\/ol>\n

    Generative AI tools are particularly good at synthesizing information and connecting dots, which is helpful in triangulating information and overcoming the use case\/feature disconnect in product discovery. I approach product discovery as follows:\u00a0<\/p>\n

      \n
    1. Generation<\/strong> of an initial query. I start by creating a requirements narrative that is often a mix of use case information and features. For the initial query, I don\u2019t worry about the quality of the request or optimizing the query; I consider it a starting point and just try to put as much relevant information into the query as I can.<\/li>\n
    2. Query submission<\/strong>. I submit the same query to multiple (at least three) AI tools.\u00a0<\/li>\n
    3. Review and Refine<\/strong>. After the initial submission, it generally becomes clear whether I\u2019m heading in the right direction or not. If I am, I follow up with additional query details until I get to either a product category to explore or a set of products to evaluate if I\u2019m not heading in the right direction. Side note: I learned very quickly not to use acronyms in my queries because I found that the generative AI tools would make a random assessment of what they represented and, in acting on that, return completely bogus information. I keep refining the original query until I get meaningful responses. You\u2019ll find that as you experiment with queries, some AI tools produce better results than others; sometimes, an AI tool will even respond with an \u201cI can\u2019t do that\u201d message, which is another reason to use multiple tools in parallel.\u00a0<\/li>\n
    4. Identify<\/strong>. If your initial objective was to identify a product category, and you\u2019ve done that, the next step is to develop a list of vendors to evaluate. At this stage, I recommend keeping the list broad to ensure you don\u2019t eliminate a good fit too early.<\/li>\n
    5. Narrow<\/strong>. With a list of vendors in hand, you can now ask the following questions:<\/li>\n<\/ol>\n