AI assistants continue to grow in capability. I am particularly intrigued by recent additions of “deep research” that essentially do web searches to collect information and then create an overall summary. Below is my experience trying to use Google Gemini Deep Research to investigate the following question

I am looking to find or create a tool that will help me record events for a bicycle travel blog of an upcoming trip. An ideal system would ask me an initial set of questions, transcribe my responses and then ask follow-on questions to clarify or collect additional levels of detail. What tools are out there that do this task or can be used together to accomplish this task?

After typing this into Gemini, it provided me the following “research plan” and asked for a confirmation

(1) Search for existing all-in-one AI-powered journaling or interview tools designed to transcribe spoken responses and ask dynamic, context-aware follow-up questions to generate narrative content.
(2) Identify and evaluate top-rated voice transcription applications and software. Focus on features relevant to a traveler, such as offline functionality, accuracy, and ease of use on mobile devices.
(3) Investigate AI writing assistants and conversational AI platforms that can be configured to act as an interviewer. Explore their capabilities for generating relevant follow-up questions based on an initial block of text.
(4) Research automation platforms to understand how a workflow could be created to connect a transcription service with a conversational AI, detailing the steps to automatically process transcribed text.
(5) Explore guided journaling and travelogue applications that use pre-set question templates or daily prompts to structure entries, assessing them as a simpler alternative to a dynamic AI system.
(6) Synthesize the findings into potential tool combinations or ‘stacks’. For each proposed solution, describe the workflow, the specific types of tools involved, the likely setup process, and the potential pros and cons.
(7) Compare and contrast the proposed custom AI-driven workflows with the simpler, pre-structured journaling apps based on setup complexity, cost, flexibility, and the potential quality of the generated blog content.

It gave me a chance to edit this plan, but it looked fine so I clicked “start research”

As it went along, it provided some progress steps along the way. For example here were the first lines. It also started to list web site URLs that it looked at.

Understanding the Core Need

I’m starting to explore tools that can help create a bicycle travel blog. The ideal solution would act like an interviewer, transcribing spoken responses and asking follow-up questions to gather more detailed information. This suggests a need for a combination of voice transcription, natural language processing for question generation, and a way to integrate these functionalities.

Initial Research Approach

My initial approach involves looking for existing all-in-one solutions that might fit this description. If a complete solution isn’t readily available, I’ll then investigate individual component technologies like transcription services, AI writing assistants, and automation platforms. I’m also considering simpler, non-AI alternatives like guided journaling applications.

Next Steps in Exploration

I’m beginning my research by searching for terms like ‘AI interviewer for journaling’ and ‘conversational journaling app’ to identify any existing integrated tools. Concurrently, I’ll be looking into ‘voice transcription apps for travelers’ and ‘AI story writing assistant’ to understand the landscape of individual technologies that could be combined to achieve the desired functionality.

As it went along it found some tools that seemed to do a lot of what I was asking (reflectr and ulog.ai). It also pretty quickly honed in on some potential solutions summarizing a bit further

Comprehensive Understanding of Solutions

I’ve now gathered a comprehensive understanding of various approaches to meet the user’s needs. This includes dedicated AI-powered journaling tools, individual components for a custom solution, and simpler journaling applications. I have enough information to compare and contrast these options effectively.

All-in-One Tool Insights

I’ve confirmed that ‘Reflectr’ and ‘uLog.ai’ are strong contenders for an out-of-the-box solution. uLog.ai, in particular, stands out for its ability to use AI to ask follow-up questions and build a narrative timeline, which directly addresses a key aspect of the user’s request.

Refined Transcription Options

For voice transcription, ‘VoiceScriber AI’ is a robust option for offline use, while ‘Otter.ai’ offers a suitable Pro plan for individual users, providing a clear path for integration. I’ve also confirmed that Zapier can effectively connect transcription services like Otter.ai with other services, eliminating the need for complex developer APIs in many cases.

Automation and AI Integration

I have a solid understanding of how to leverage Zapier to automate the workflow between a transcription service and a conversational AI, such as ChatGPT. This provides a clear framework for building a custom solution that can generate follow-up questions based on transcribed input.

Ready for Final Synthesis

I have reached a point of saturation in my research and possess all the necessary information to synthesize my findings. My next step is to compile a comprehensive report that compares the various solutions based on factors like setup complexity, cost, flexibility, and the quality of the output, providing a clear overview for the user.

This led it to create the following report. Which was the result of the research. While it doesn’t solve the complete problem, it seems like an intriguing starting point of pointing at existing tools/approaches described on the web as well as other approaches.

Note: This is an 18-page report, at bottom left you can go to next page…