3
13 Comments

Request for product: AI-powered customer feedback tool

As an indie hacker, I constantly get requests for features from different sources: email, Github issues, Github forum posts, hacker news comments, indie hacker comments, etc. I keep a mental list of what's most requested based on reading all these, but it would be much more helpful to have an AI that ingested all these sources then gave me the top ones, with counts of how many users requested each and links to the exact comments so I could view details and verify the AI was correctly categorizing comments. Then it would be easy for me to prioritize features based on count and maybe even emotional tone of the requests.

I don't have the time or interest in building this myself, but thought I'd ask the IH community to build it. I think it would be a great business for indie hackers and probably startups as well.

As a starting point, I suggest just the ability to ingest emails. I have a ready-made dataset if you want me as a beta tester. 😁

on March 29, 2023
  1. 1

    I have seen someone suggest a similar idea on reddit, for a service where you create a list of customer support subjects and the AI is supposed to take each customer support request you get and put it in the correct subject.

    I thought about building it but this was a while ago before chatgpt became popular and was offered as a service.

    1. 2

      @fle_x Fm, the need is more often than not -- sorely misunderstood.

      While many request 'AI-powered customer feedback tool' they are not saying the explicit pains they wish to eliminate.

      • From a feedback perspective, the pains are not the same across contexts.
      • Triage is king for inbound requests, where the volume has reached overwhelmed.
      • Some only need a data gathering tool (email or video snippet to transaction). Transaction being the desired FR tool used.
      • For example: Apple's feedback tool (a distinct app) really showed this too after using it across my ipad mini, iphone and macbook. Three distinct experiences, for various apps or osx upgrades.
      • the best inbound feedback tools I have seen show state-of-play on a dashboard, so all feedback-ers can come up to speed quickly (at-a-glance) on age of issues, counts, impacts, scope breadth and vertical. The dashboard is needed as the volume of FR is often extensive. There is a side benefit when dev take this approach. Their users (including prospects) learn aspects the app they did not consider. Many times I have voted saying, yes that is a good idea, and had not thought of it. FR are correlated ot the context in which in the app is used AND the maturity of the business using the app. When feedback and ideation is crowd sourced, with transparency, experience has shown me a more robust and greater alignment about any particular issue, occurs.
      • The best capture point of the issue at hand is right there on screen, like google does Except it would be even better if I could click a button to say record.. and shows what I was doing and can voice over to explain. Often when a user goes to email issue, some important specifics are lost. Then an AI model video to text is needed to record as transaction.

      AI that ingested all these sources then gave me the top ones

      Definition of 'top ones' actually will evolve over time in most app launches.

      with counts of how many users requested each

      Using counts is rarely going to guide the most impactful requested features. I have seen apps that use counts, and launch those features and discover later they missed the mark. This is because often the first feature request is articulated from one perspective, versus a person's who understands the full app. We would often curate and join feature requests. A role of any good product manager.

      AI or even simple tags, could certainly. map 'similar' FR with similar', then offer a thread like conversation to the developer/product manager.

      1. 2

        Makes sense, I'm tackling the problem as an engineer I, have never worked as a product manager so it makes sense that my perspective on the problem isn't the best.

        This is my problem with creating SaaS, I don't have enough experience with finding good ideas and applying them to be practical enough to make money

        1. 2

          Understood. Agree, it is wise to pause.

          • Many tactics to validate ideas is way-off from my business experience e.g. LP signup page; newsletter signup, traffic is traffic + nothing else .
          • Fm, the art is understanding 'context'. It has been super interesting using GPT and seeing what context is can handle and what it totally misses. Listing a set of needs is not context. In fact in my many years, I have yet to see a build solution that really gets context. This FR app is an example of what I call a 'build solution'. An app to make building great apps 'more likely to occur'. Increasing the probability of the dev getting the desired context at the earliest point in time.

          Transformation of app build space is underway.

          • Those who pathfind in this shift will do very well as the appetite from the SME side is long overdue. Space is too noisy for most business to interact with. Too effortful.
          • App build play zone from a business perspective. A workspace they can simulate stuff, play to test their ideas. Challenges with cash pots can be sponsored by bigger businesses (medium) or by niche industries to solve their problems.
          • If such a play -zone existed I would subscribed already. Test concept using the 7 steps GPT4 articulated to build my first ever citizen iphone app. Like I am paying for the chatgpt subscription I would pay a similar amount for workspace. Imagine if a bundle of dev became a collective, each offering a needed workspace (as I am sure there are many). Then business, just needs to buy 1 subscription, and gets directed to the workspace they need for their current project. Coaching opportunities would spring forth. Sadly all such style workspaces are from a technical lense not a business lense. Missed opportunity.
    2. 2

      Cool. I recently found kraftful.com which purports to do something similar.

      1. 1

        Yep seems similar enough. this is what mostly discourages me from creating a SaaS there always seems to be a service already released for every single idea, but I guess the key is to make something better than that service.

        1. 1

          Yeah, I guess it can be tough if there's a lot of competition. Then you need a competitive advantage either in your features, pricing, or in your marketing distribution channels.

          But also, there are TONS of ideas that don't have services yet, or have services that only tangentially meet the needs of some customers. I would talk with a lot more people and ask them what their problems are. I found this really inspiring when I did it. Don't get discouraged!

  2. 1

    This sounds great. I am trying to create a feedback tool that analyzes feedback but I can go this particular direction.

    My idea was you can't go through every request so AI is going to analyze it and save that time for you. Let me know what do you think about it?

    Also, give ai all data about user like location, device information and let it figure out what users are facing which problems like mobile users have one issue or mac users have particular issue

    1. 2

      I mean, I do read every request but it's hard for me to categorize and keep track of the counts. I do that process in my head but it's very vague and could do with some precision. That's why I believe I would like categories with counts and also links to the actual request to verify that the AI categorized correctly and get more details.

      1. 1

        Yeah. You will need to go through request but as you said categorize and make relations between different requests from users based on other data is the main thing. The feedback sentiment analysis will also be there.

        1. 1

          Yeah, sounds cool! I've categorized every email I've ever gotten as "bug or feature request" so I have a pretty large data set already. Are you building something yet?

          1. 1

            Just started building planning to put out a product in 3-6 weeks. Aim is 3 weeks for launching MVP.