Faqs

Yes! Any language is no problem. No special training materials or translation needed. Chatsi AI can train in English --then answer in any language.

Chatsi admin consoles let non-technical users:

Add or modify FAQ entries and knowledge snippets
Define synonyms or alternate phrasings for key queries
Set exclusion lists (e.g., “don’t recommend product X”)
This aligns with the value promise of streamlined, self-serve management.

Yes—Pattern PXM (Amplifi.io) exposes an API you can use to pull product info and assets; their help center documents API integrations to external systems. You can schedule pulls or use your own webhook/polling strategy depending on what you enable.

Yes. Our modern platform lets users adjust the chatbot’s style and boundaries via:

An admin interface to set desired tone, style, colors, voice, etc. (e.g. formal vs conversational)
Custom response templates, few-shot examples and fallback phrasing
Toggleable intent filters or “no-go” response policies (guardrails)
All without developer intervention

Chatsi promotes itself as a “multilingual” AI sales rep. While RTL support (like Arabic or Hebrew) isn’t explicitly streamed RTL, our platform does support multilingual use cases via language-specific styling.

A/B testing of responses or conversation flows

Offline evaluation pipelines that simulate real queries and validate answers

Human-in-the-loop workflows for approval of new content or guardrail adjustments

Chatsi is positioned as a sales concierge that combines AI with real-time data . Recommendations are driven by:

Real-time hooks to your ecommerce backend for availability and pricing
RAG-driven dynamic retrieval from product metadata + pricing tiers
As long as async connectors pull your live catalog data, recommendations can reflect your latest inventory and pricing contextually.

We design to SOC 2’s five Trust Services Criteria (security, availability, processing integrity, confidentiality, privacy) and provide a GDPR Data Processing Agreement (DPA) when acting as a data processor; for U.S. consumers we follow CCPA/CPRA obligations.

Chatsi connects to Shopify (or WooCommerce, SalesForce Commerce Cloud, Miva) via official APIs: Shopify’s Admin APIs with scoped read-only permissions (products, inventory, etc.) plus webhooks for real-time sync; WooCommerce’s REST API similarly supports reading products, orders, customers, etc. We can also sync with other content from trusted public and private sources.

Grounding is achieved by retrieving relevant docs/snippets (e.g., from PXM + store backend, and other trusted sources) and feeding them to the AI agent at answer time; quality is tracked with RAG-focused evaluators/metrics like groundedness (faithfulness), answer relevance, and retrieval recall/precision per Microsoft’s RAG design/evaluation guides and OpenAI’s RAG best practices. We don’t tolerate any hallucinations.

Chatsi offers easy-to-install apps for major e-commerce platforms like Shopify, Salesforce Commerce Cloud, Woo, Bigcommerce, Magento, and WordPress.

We use:

Grounding layers—RAG ensures responses are based on up-to-date product data/policies
Validation hooks—pre-response checks that see if product is out of stock, or policy is correct
Fallback tier—the system says “Let me check…” or routes to support instead of giving incorrect guidance

Exportable knowledge bases (FAQs, intents, synonyms) in JSON/CSV

Exportable embeddings/index data in standard formats for porting

Implementing Chatsi on a site only requires adding the app and/or a script provided by Chatsi. So technically it’s just a minute or two. Most proof-of-concept (POC) trials last 30 days in order to fine-tune the AI persona and genius details with the merchant and tune up the source data that trains your custom AI knowledge base.

Yes - our AI sales agent can also deploy easily on any other site platform like SFCC (Salesforce), Magento, BigCommerce, Miva, Wordpress, etc. We offer app-based installation for the most popular e-com platforms.

Sales-related metrics & definitions

Purchase revenue: Total revenue from purchase events, typically calculated from the GA4 purchase event parameters and net of refunds.
Average order value (AOV): Revenue ÷ number of orders; Shopify’s own guidance also frames AOV around gross sales minus discounts.
Ecommerce conversion rate: % of sessions or users that result in a purchase (GA4 supports session- and user-based conversion rate views).
Add-to-cart / Cart-to-view rate: % of product views that lead to an add-to-cart (powered by GA4 events like add_to_cart, view_item, view_cart).
Checkout conversion rate: % of checkout starts that complete an order (often analyzed via GA4’s Checkout Journey).
Cart/checkout abandonment rate: Share of carts or checkouts that don’t convert; can be built from GA4 audiences for cart abandoners and is widely discussed in Shopify’s guidance.
Items per order (units/transaction): Average number of items per completed order (see GA4 “Items purchased”).
Attach rate: Ratio/percentage of add-on or secondary items sold with a primary item (e.g., accessories per main product).
Returns/refund rate: Portion of orders or revenue later refunded/returned (Shopify sales reports treat returns/refunds as negatives on the date processed).
Attributed revenue from support/chat: Revenue tied to conversations or assist flows in tools like Gorgias.

Built-in dashboards for conversation volume, successful self-resolves, handoffs, deflection rates

Exportable logs (CSV, JSON) or APIs for raw chat transcripts and metrics—perfect for BI ingestion and custom reporting

Chatsi claims to reduce support queries by about 40% by answering low-level support questions and delivering custom answers to questions about returns or other product use issues (according to estimates from our customers). For non-pre-sale or complex support issues, the Chatsi AI Agent can programmatically create a Gorgias ticket (setting the channel, requester, subject/body, metadata) via the Gorgias REST API, which is the standard escalation path from custom apps. Chatsi can also simply route the user to the form or support intake page on the merchant’s site, offering details to guide the customer on their support path.

https://developers.gorgias.com/docs/create-a-ticket-using-api?utm_source=chatgpt.com

Chatsi uses deep product and other data from shop backend and other approved sources, including product attributes, PDF manuals, installation guides, images, videos, and other documents. It also uses public data such as reviews, YouTube, Reddit, and social media, as well as optional customer data like demographic information, purchase history, and preferences. This data is processed using Gen AI, Natural Language Understanding (NLU), and Retrieval Augmented Generation (RAG).

Support via email, ticketing, or chat during business hours (e.g., 9 am–6 pm), with optional 24/7 add-ons

Escalation paths include triage by support → CSM (customer success manager) → technical engineering if needed

Resolution Rate = solved tickets ÷ received tickets × 100 (measures how many issues are actually closed).

Containment/Deflection = % of conversations resolved by the bot without human hand-off. Higher is better for “self-serve” flows.

CSAT = post-interaction “satisfied” responses ÷ total responses × 100 (standard customer-satisfaction metric).

99.9% uptime (just under 9 hours/year downtime), or even 99.95% for enterprise-level

Exact terms would be defined in Chatsi’s contract per needs

Typically, onboarding for Chatsi includes:

Week 1–2: Discovery, setup of connectors to Shopify/WooCommerce, and initial data sync & ingestion.
Week 3–4: Framework completion (knowledge base setup, brand tone configuration, RAG tuning, testing).
Week 5–6: UAT (user acceptance testing), training material delivery, and soft launch.

Training is often provided by an onboarding team or customer success team, including configuration walkthroughs, Q&A sessions, and best-practice guides.

Chatsi pricing is a simple subscription; Starter, Growth, or Enterprise edition. Selecting the right edition is based on storefront’s scale, traffic, product catalog size, features and data depth required. Additionally Chatsi charges an AI token passthrough fee after a free ‘token trial period’ (to ascertain real AI token usage). We charge the same rate as published by the best models. This is better (and cheaper) than other chatbot platform pricing patterns: per chat, per question answered, per resolution, per conversion, or percentages of sales (outcome-based).

We are running on both Azure US West and US East data regions. Your data is not sent to Open AI but is processed on a privately hosted top LLM model within Azure for ultimate data security and privacy.

Chatsi is trained at sales, but support also comes naturally with answers to tough questions. Just like a physical store there may be times when a salesperson is approached with a support question so Chatsi can answer using its deep product knowledge, but it also has logic to then route by detected intent to “answer,” “recommend,” or “escalate to support”.

Building a DIY chatbot may seem cost‑effective, but one hallucinated policy or unsafe response can expose your brand to lawsuits, fines, and viral bad press—just ask Air Canada and Pak’nSave. Purpose‑built platforms shoulder the heavy lifting of content moderation, attacks, legal compliance, and continuous model tuning, so you can innovate without risking customer trust or your bottom line.

Frequently Asked Questions