AI Configuration

Supportly works best when the assistant is treated like a constrained support operator. Strong guidance, clean retrieval sources, and explicit escalation boundaries matter more than clever phrasing.

Automatic priority assignment

Owners and admins can enable automatic priority assignment in AI configuration. It is disabled by default. AI evaluates customer messages on open tickets using the workspace model, fallback settings, and AI allowance. It assigns low, medium, high, or urgent priority and records changes with a reason and confidence in the activity log.

Relaxed strictness raises priority early for plausible impact, balanced requires clear evidence, and strict reserves high or urgent for substantial, concrete impact. Minimum confidence (50–100%, default 80%) determines whether an assessment can change priority. Below the threshold, the current priority stays unchanged.

Enable reassessment to evaluate new customer messages, allow urgent priority for severe incidents, and optionally allow downgrades as impact decreases. Custom priority guidance adds business-specific criteria. Manual priority selections always take precedence, including bulk updates. Resolved and closed tickets are skipped. Existing tickets are assessed on their next customer message; enabling the feature does not run a historical backfill.

Automatic category selection and tagging

In Settings > Categories & tags, choose Generate categories or Generate tags to suggest new workspace choices. The generator uses your current AI configuration, including instructions, persona, classification guidance, priority rules, and escalation settings, together with active FAQs, document excerpts, workspace context, and existing categories and tags. It uses your configured model, fallback settings, and AI allowance. Large sources are shortened to fit the model, and the dialog shows when this happens.

Request 1 to 5 suggestions and optionally add guidance about the topics you want. Review the evidence, edit category names and descriptions or tag names and colors, and select only the suggestions you want to add. Generation does not create choices until you click Add selected. Existing names are filtered out, and failed saves can be retried without adding successful choices twice.

Owners and admins can enable AI category selection and AI tagging independently in AI configuration. Both are disabled by default. Create categories and tags in ticket settings first: AI selects only existing workspace choices, using category descriptions to distinguish similar topics.

Each feature has relaxed, balanced, or strict matching, a minimum confidence from 50 to 100% (default 80%), reassessment on new customer messages, and optional custom guidance. Category selection chooses one best match; allow AI category changes to replace its earlier selection as the issue develops. Tagging adds matches above the threshold, with a maximum of 1 to 10 AI tags per ticket (default 3). Existing tags are never removed.

Manual category selections and tag edits always take precedence, including clearing a category or all tags. Existing categories and tags without AI ownership are preserved. Resolved and closed tickets are skipped. Assessments use your selected model, fallback settings, and AI allowance; category and tag changes appear in the activity log and update live ticket views. Enabling the feature assesses existing tickets on their next customer message, without a historical backfill.

Model routing and failover

Supportly supports standard-provider models plus managed Pro presets. Pro presets keep upstream provider details hidden and can automatically fail over to a workspace-selected standard fallback model when retryable Pro errors occur.

Fallback redirects create internal-only conversation notices for operators. They appear in dashboard timelines and ticket views, but stay hidden from customer-visible widget history and transcripts.

System prompt best practices

Write the system prompt as a compact operating policy. Say what the assistant should optimize for, what it must never do, and when it should escalate. Avoid vague instructions like be helpful or always delight the customer unless they are followed by concrete rules.

Prompt example

Start from a practical baseline like this and then tune it using real conversations.

system-prompt.txttext

1You are Supportly for Acme Support.
2
3Answer only using the workspace knowledge base and the details provided by the customer.
4If the customer asks for refunds, billing reversals, account ownership changes, or anything you are not confident about, escalate to a human agent.
5Keep answers concise, clear, and procedural.
6Ask one clarifying question when required, but do not loop.
7If a request requires internal action, say so plainly and hand off.

Keep the persona name aligned to the brand or support team name.
If AI-assisted closure is enabled, train the assistant to offer closure only after the issue is clearly resolved.
Require the customer to reply with the exact word close before ending the ticket automatically.
Prefer operational verbs like verify, ask, escalate, summarize, and confirm.
Review prompts after actual escalations rather than tuning in the abstract.

Knowledge base tips

Retrieval quality depends on clean source material. FAQ entries should answer one question fully. Uploaded documents should be scoped to one policy or workflow whenever possible.

If an answer requires hidden context to make sense, it is not a good knowledge base candidate yet. Rewrite it so the retrieved chunk can stand on its own.

Source type	What works well	What to avoid
FAQ	Short question plus direct answer	Bundling multiple workflows into one answer
Uploaded docs	Single-purpose policies and SOPs	Long mixed-topic exports with stale sections
Escalation notes	Patterns observed from handoffs	Private internal details that should never be surfaced to customers

Escalation rules guide

Tune escalation using confidence threshold, sentiment detection, maximum AI turns, and explicit trigger phrases. These controls should reflect the actual risk tolerance of your support team.

In Widget > Human handoff, choose AI-assisted or Keyword only. AI-assisted mode evaluates the conversation in the customer's language and transfers issues that need account access, remain unresolved after troubleshooting, or require a specialist. Customers do not need to type a special phrase. It follows the rules configured here, including confidence, sentiment, maximum replies, and full AI resolution.

Keyword only uses only the escalation keywords configured here for automatic handoff. The AI's handoff decisions, sentiment, confidence, and reply limit do not trigger a transfer in this mode. The Request a human button works independently and can be hidden in widget settings. With that button hidden and keyword-only mode selected, configure at least one keyword so customers can reach your team.

Escalation rules payload

The dashboard UI edits the same shape shown below.

escalation-rules.jsonjson

1{
2  "confidence_threshold": 72,
3  "sentiment_detection": true,
4  "max_ai_turns": 8,
5  "escalation_keywords": [
6    "speak to an agent",
7    "manager",
8    "real person",
9    "cancel my account"
10  ],
11  "allow_full_ai_resolution": true
12}

Good review loop

Look at escalated tickets every week. If the AI escalates too early, improve the knowledge base. If it escalates too late, tighten the prompt and lower the confidence tolerance.

Escalation and human takeover

Configure escalation behavior, confidence handling and ticket handoff in AI Configuration. Human takeover stops AI generation for the conversation.

Permissions: Owner/admin configures AI; permitted operators may intervene. Plan: All plans, subject to AI quotas. Use /ai-config/escalation.

Human requests create a visible handoff
Intervention cancels streamed AI output
No-online-agent behavior matches saved settings

AI models and fallback

Configure the supported model, persona, system prompt, fallback and spending settings using AI Configuration. Reuse existing reservation, token and spending controls.

Permissions: Owner/admin (configure_ai). Plan: All plans; models, usage and spending are limited by the plan. Use /ai-config.

Plan-ineligible models cannot be selected
AI failures follow the configured fallback
Concurrent requests cannot bypass spending reservations