All guides
    Marketplaces

    Reducing GMod addon support workload: a creator's guide

    Practical playbook for GMod addon sellers drowning in tickets. Automate repetitive questions with an AI agent trained on your docs.

    Selling GMod addons at any scale means answering "how do I install this?" until you dream about it. This guide is the playbook we wish we'd had when we sold our first 40,000 addons: how to cut your support workload without cutting quality.

    The three types of GMod ticket

    Every ticket falls into one of three buckets. The mix determines your strategy.

    1. Install & setup (55–70% of tickets). "Where do I put this?" "Why isn't it showing up?" "Do I need DarkRP?" These are pure documentation problems. An AI agent handles them at 95%+ accuracy once trained.

    2. Configuration & customization (20–30%). "How do I change the price?" "Can I add more categories?" These need product-specific answers. An AI agent handles them well if your config docs are complete.

    3. Bugs, incompatibilities, and refunds (5–15%). These need a human — you. But you can only spot them when the noise is gone.

    The goal: automate buckets 1 and 2, so you have real time for bucket 3.

    Step 1 — Audit your last 50 tickets

    Open your GmodStore or Pivity ticket history and read the last 50. Tag each one by bucket. Count the top 10 most-asked questions.

    Those 10 questions are your entire automation ROI. Answer them once, well, in your training data, and you've deflected 60% of everything that follows.

    Step 2 — Write documentation like an AI will read it

    AI agents work best with documentation that has:

    • Explicit headers ("Installation", "Configuration", "Common errors")
    • Named error messages as literal quotes: [ERROR] Attempt to index nil value
    • Step-by-step lists instead of paragraphs
    • Version numbers where they matter

    Skip the marketing language. The agent doesn't need to be sold on your addon.

    Step 3 — Set up your Arnvio agent

    Dashboard → AI Agents → New Agent. Link your GMod products (via Pivity or GmodStore integration). Upload your cleaned-up docs under Training.

    Add a system prompt like:

    You are the support agent for [addon name]. You help customers install, configure, and troubleshoot the addon. You only answer questions grounded in the provided documentation. If a customer reports a bug or asks about pricing, escalate to a human. Never quote source code back to a customer.

    The last line matters. GMod addons ship as Lua source, and customers occasionally try to get the agent to leak logic. Arnvio blocks this at the model level, but the explicit instruction reinforces it.

    Step 4 — Run in draft mode for a week

    Every drafted reply the agent gets wrong is training-data gold. Note the miss, add the correct answer to your training, move on.

    By day 5–7, the misses drop to near zero for buckets 1 and 2. Flip to auto-send. Configure escalation: if the customer replies a second time (meaning the first answer didn't satisfy them), route to you.

    The unglamorous truth

    Automation doesn't fix bad documentation — it exposes it. If your addon has confusing config, the agent will confidently give confusing answers. Every hour spent tightening your docs pays back 10x in support time saved.

    The creators getting the most out of Arnvio treat their documentation as a product. It's not overhead. It's the thing your AI agent is made of.

    Next steps