Tisane

@Tisane analyze this customer feedback for sentiment and key topics: 'The new dashboard is confusing and slow, but the support team was incredibly helpful in getting us set up.'

This example would return a structured NLU breakdown: overall sentiment is mixed (negative on product, positive on support), with detected entities like 'dashboard' and 'support team', topics including 'usability' and 'customer service', and potentially flagged attributes like 'speed' (negative) and 'helpfulness' (positive). The response would include confidence scores for each detected feature, allowing you to quantify sentiment patterns across multiple feedback entries.

This showcases Tisane's core NLU capability — going beyond basic sentiment to extract entities, topics, and attributes in one pass. Requires an API key with text analysis permissions. Most useful when processing batches of user-generated content where you need structured insights, not just positive/negative labels.

@Tisane translate this product announcement to French and Spanish: 'We're excited to launch real-time collaboration features next quarter.'

This example would return two translations: French ('Nous sommes ravis de lancer des fonctionnalités de collaboration en temps réel au prochain trimestre.') and Spanish ('Estamos emocionados de lanzar funciones de colaboración en tiempo real el próximo trimestre.'). Each translation preserves the tone and intent of the original, with Tisane handling context-aware phrasing rather than word-for-word substitution.

Demonstrates Tisane's transformation tool for localization workflows. You can also use this for paraphrasing within the same language. Check supported languages first with the get-languages tool — not all language pairs have equal coverage. Translation quality depends on domain-specific terminology in your API plan.

@Tisane compare these two feature requests for semantic similarity: 'Add dark mode to the mobile app' vs 'Support night theme on iOS and Android'

This example would return a similarity score around 0.85–0.92 (scale 0–1), indicating these requests are semantically nearly identical despite different wording. The AI could then use this score to deduplicate feature requests in a backlog, grouping 'dark mode', 'night theme', and similar phrasings under one canonical item. Lower scores (below 0.6) would flag genuinely distinct requests.

This highlights Tisane's semantic similarity tool for deduplication or clustering tasks. Useful when consolidating user feedback, support tickets, or survey responses. The numeric score lets you set thresholds programmatically — but context matters: two texts about 'apple' (fruit vs. company) might score high on surface similarity yet differ in intent.

When Tisane beats keyword filters for support triage

A 6-person support team handling 200+ tickets daily across three languages needs automatic routing before agents see them. Tisane's Analyze Text tool detects sentiment and topics in one pass, so you can route angry customers to senior agents and billing questions to finance without writing regex rules for every edge case. The language detection runs first, then sentiment scoring flags priority. This works when your ticket volume justifies API costs (roughly $0.01 per analysis at scale) and you're already using an MCP-compatible workspace. If you're under 50 tickets a day, manual tagging is cheaper. For teams drowning in multilingual support chaos, Tisane turns triage from a 20-minute morning ritual into a background task.

Tisane for real-time comment filtering at community scale

A 3-person community team moderating a forum with 10k daily comments needs to catch abuse without reading every post. Tisane's sentiment and entity detection flags toxic language, personal attacks, and spam patterns faster than human review. The Analyze Text tool runs on each comment at submission, scores it for abuse signals, and holds flagged content for manual review. This setup works when you have structured text input (not images or video) and can tolerate a 200ms API round-trip per comment. If your community is under 500 comments daily, simpler keyword blocklists are enough. For high-volume forums where one missed threat tanks trust, Tisane gives you automated first-pass filtering that doesn't burn out your mod team.

When semantic similarity beats keyword search for docs

A 5-person product team maintains help docs in four languages and fields the same questions repeatedly because users can't find answers. Tisane's Calculate Semantic Similarity tool scores user queries against your doc library, surfacing the right article even when phrasing doesn't match exactly. You run Get Supported Languages once to confirm coverage, then pipe every search query through similarity scoring against pre-indexed articles. This wins when your docs are under 5k articles (larger sets need vector databases) and your team already uses an MCP workspace for support handoffs. If your knowledge base is English-only or under 50 articles, basic full-text search is simpler. For multilingual teams tired of rewriting the same answer in Slack, Tisane turns your docs into a self-service system that actually works.