All guides
    Marketplaces

    How to automate Pivity support tickets with AI

    Connect Pivity to Arnvio, train an AI agent on your product docs, and get accurate draft replies for every support ticket — you review and send.

    If you sell GMod or Rust addons on Pivity, you already know the pattern: the same five questions, all day, forever. "How do I install this?" "Why is my server crashing?" "Where's the config file?" You wrote the answer in your product description. They didn't read it. You reply anyway.

    Arnvio doesn't take your customers away from you. What it does is read every incoming Pivity ticket, pull the answer out of your own documentation, and hand you a ready-to-send draft reply. You skim it, tweak if needed, hit send. The customer still hears from you — you just stop typing the same paragraph fifty times a week.

    What you'll need

    • A Pivity seller account with at least one active product
    • Your Pivity API token (Pivity dashboard → Settings → API)
    • An Arnvio account (Free plan works for the initial setup; you'll want Basic or higher once you're handling real ticket volume)

    Step 1 — Connect your Pivity account

    Head to Dashboard → Integrations → Pivity in Arnvio and paste your API token. Arnvio stores it encrypted (AES) — the raw token never sits in plain text anywhere.

    Once connected, your product catalog appears in Dashboard → Products. Click Import from Pivity and select the products you want the agent to help with. Import limits scale with your plan.

    Step 2 — Create your AI agent

    Dashboard → AI Agents → New Agent. Give it a name (this is internal — customers never see it, because you are the one sending replies). Pick a model — gemini-3-flash-preview is the default and drafts Pivity ticket replies well.

    Link the products this agent should cover. One agent can handle multiple products; larger stores usually split into 2–3 agents by product family so each one stays sharply focused on its own docs.

    Step 3 — Train the agent

    This is where accuracy comes from. Under Agent → Training, add:

    • Product documentation — paste it directly, or upload a file (1MB per product limit)
    • FAQs — the exact questions customers already ask, with your canonical answer
    • Source code (optional) — for technical questions about your addon's API or hooks. The agent uses it for context but is forbidden from ever quoting your source back to a customer

    Aim for 50k–100k characters of high-quality documentation per agent. More isn't better — accuracy is.

    Step 4 — Import your Pivity tickets

    Go to Dashboard → Tickets → Pivity. Arnvio pulls your open tickets on a schedule and lists them alongside the product they were opened on. You'll see the customer's message, the product context, and any conversation history — the same view you'd get on Pivity, just consolidated with your other stores.

    Retention and how far back tickets are pulled scale with your plan.

    Step 5 — Generate a draft reply

    Open any ticket and click Generate AI reply. The linked agent reads the ticket, searches your training data for the relevant answer, and produces a draft in the reply box within a few seconds.

    You then:

    1. Read it. Most of the time it's accurate enough to send as-is.
    2. Edit if needed. Reword a sentence, add a personal note, drop something that doesn't apply.
    3. Send from Pivity. The reply goes out under your seller account — the customer only ever sees you.

    Nothing is ever sent automatically. Every reply passes through you.

    What to expect

    Realistic numbers based on what creators actually see once training is dialed in:

    • Roughly 70% of drafts are send-ready with no edits — the agent found the answer in your docs and phrased it well
    • Another 20% need a small tweak — a sentence rewritten, a caveat added
    • The remaining 10% are the tickets that genuinely need your brain — bugs, feature requests, edge cases. Exactly what you should be spending time on

    Time-per-ticket typically drops from 4–6 minutes of writing to 30 seconds of reviewing.

    Common gotchas

    • Drafts reference features that don't exist. Fix: tighten your training data. Remove anything vague or aspirational — the agent will confidently repeat whatever you feed it.
    • Tone feels off-brand. Fix: adjust the agent's voice in Agent → Personality, or just edit the draft before sending. A few edits and the agent tends to match your style over time.
    • Rate limits on Pivity API. Arnvio handles retries and backoff automatically, but if you're hitting them repeatedly, check Integrations → Pivity → Rate limit dashboard.

    Next step

    Once Pivity drafts are landing well, most creators add BuildByBit or GmodStore on the same account and let one agent (or a set of agents) draft replies across their entire support surface. See our BuildByBit guide for that setup.

    Questions? Email support@arnvio.com — a human always replies.