We're sharing a guest blog post from our community EIR Grant Singleton, highlighting 15+ examples for how marketplaces can leverage AI/agents. This was previously shared as a post in the community here.
‍
Hey EM, I received lots of feedback after our talk the other day and questions from founders about how we’ve leveraged AI/agents at PangoBooks. I thought it would be helpful to put together an updated post from my previous post (here) covering 10+ examples for how we’ve leveraged AI/agents at PangoBooks and you can as well.
‍
I’m going to group the examples based on the order of importance when starting a marketplace and also add a few more recent examples to the list. This is also not exhaustive and I can share more in the replies below if it would be helpful. 
‍

1. Building the marketplace more efficiently 

• Agentic coding: Tools like Cursor, Claude Code, OpenAI's Codex can speed up dev times significantly. Not yet good enough to build long term production marketplaces without the need for engineering knowledge but will speed up developers and give non-devs the ability to get started. 
• Database building: If your marketplace depends on product data but there isn’t a comprehensive API. It’s now possible to build an AI agent that can take a list of products and a schema, and go get the data needed. I’ve personally built out databases with thousands of records using an AI agent. 
  ‍

2. Improving the user/marketplace experience

• AI shopping assistant: An in app chat assistant that has tools to query your marketplace data. There is lagging demand as users switch their habits from traditional to LLM based search. 
• AI shopping agent: An agent that does the shopping for the buyer. New territory, technically possible but have not seen it in production. It’s going to take time to build buyer trust. 
• Review summaries: Take hundreds of product reviews and concisely summarize them so that the buyer can get a sense of what everyone is saying about the product or the seller. 
• Listing data enrichment: Improve listing quality by fetching relevant missing data after the seller lists. 
• Listing from a photo: Seller uploads photos and listings are built out entirely by AI. This will be the listing flow of the future and marketplaces that get there first will likely gain access to unique supply by lowering the barrier to sell. 
  ‍

3. Operational aspects (also moderation, fraud, trust & safety, etc)

• Content moderation: Likely the simplest way to use AI at scale in your marketplace. When a new listing lands, run the data through an LLM with a system prompt that has your listing guidelines and have it flag listings that aren’t allowed or need updates. Allows you to maintain high quality supply without a dedicated listing reviewer. 
• Fraud detection: Works just like content moderation but on orders. Tell it what to look for in the system prompt and have it flag potentially fraudulent orders. Great for marketplaces with high volume that would otherwise require a lot of hours of manual review to find fraud. 
• Internal messaging spam prevention: Marketplaces that allow buyers and sellers to communicate privately on platform will get spam when they scale. This could be anything from soliciting, phishing, flirting, or taking sales off platform. All of this can be monitored at scale with AI. Run all communications through an LLM with a system prompt that outlines your communication rules and have it flag you when it finds something. 
• Seller profile building: Build richer data for your sellers for your own internal analytics, support, or to make onboarding easier. Scrape data about them from social media handles using AI automatically after sign up. Take care to update your privacy policy and give sellers confidence that this is speeding up their own onboarding. 
• Snowflake ops: Every marketplace has unique internal ops that no one is ever going to build software for. There are many opportunities here to build custom AI agents and AI workflows to do this work. This requires a change of thinking from role-based to task-based. In the past we have hired people to take on a suite of tasks. You shouldn’t try to replace people with AI, you should replace tasks with AI. Pick off tasks that can be completely handled by AI to free up your team. 
  ‍

4. Growth & scaling

• SEO: I talked previously about building a product database with AI. Once that’s done, you can leverage that data to build as many product pages as you have records in the database. On their own they won’t perform well but the longtail of those pages will give you more opportunities to show up. We did this at PangoBooks and most pages get no traffic but some get a ton of traffic. We couldn’t have guessed what those pages would have been so often it’s a good idea to cast a large net and capitalize on the power law. In addition to this, you can use AI to help you find opportunities. I built an AI agent that uses keyword data to find those opportunities. You can also use AI to write content but this only works if you add a lot of original context. LLMs are designed to produce the average of all content and Google awards original content so these two are at odds. The way to overcome this and produce original content with AI is to provide original information and have AI put it in the right format for you. 
• AEO: SEO is the foundation of AEO. You need to show up in search engines in order to be mentioned by AI so SEO still matters, a lot. However, AEO is allowing businesses with low or no domain authority to show up in ChatGPT (and other engines). This is possible for two main reasons. First, because AI looks deeper than the first few links, so if you rank on page 2, you still might get mentioned by AI. Second, because AI looks at what people are saying on third party sites like Reddit, so you can get mentioned even if your domain doesn’t show up at all. 
• Agents for growth: My advice for this is to audit what you currently do to get a buyer or seller and see if you can build an AI agent to do it and scale it. Write down the process you personally take to get a user, and ask yourself if it can be done by AI. 
  ‍

5. Customer support 

• Intercom Fin: Or another third party. I don’t recommend building this in-house when there are so many great options at reasonable prices. 
• AI agent for replying to emails: Use something like Relevance AI or Lindy to read your incoming emails to your support team in real time and answer the ones that it can handle. This has huge alpha because the user experience is incredible when done right. Users can get a reply in a minute or two when they expected it to take a day or more. 
  ‍

I would be happy to help expand on this in the replies below. I’m also curious to hear how others are leveraging AI/agents and any new use cases.
‍
You can connect with Grant to discuss this post in the Everything Marketplaces community here. A big thanks to Grant for also being active in the community as an EIR, where he is often sharing his marketplace experience, insights, and helping early-stage founders.