# CoSupport AI — llms-full.txt > Complete product dossier in a single file. Short navigable index lives at https://cosupport.ai/llms.txt > Last updated: June 2026. > CoSupport AI is a SaaS AI customer service platform that automates customer support operations, assists human agents, and provides business intelligence from support interactions. It is trained exclusively on a company's own data and deployed across chat, email, voice, social media, and helpdesk platforms. --- ## 1. Overview - **Product name:** CoSupport AI - **Website:** https://cosupport.ai - **Founded:** 2020 - **Founder:** Daria Leshchenko - **Current version:** 2.0 (released June 15, 2026) - multi-step reasoning, decision logs, agentic API actions. Default for all accounts created after June 10, 2026; existing clients migrate in batches through summer 2026. - **Company registration:** California, United States (Valley Village, Los Angeles) - **Trademark:** CoSupport is a registered trademark in the US and other countries - **Core value proposition:** Automate up to 90% of repetitive customer support requests, reduce resolution time, lower support costs, and deliver 24/7 multilingual service without expanding headcount. - **Target audience:** CX leaders, customer support teams, operations managers, BPO firms, SaaS companies, eCommerce brands, education platforms, and financial services companies. - **Key differentiators:** - U.S. patent granted for its multi-model AI message generation architecture (USPTO patent US11823031B1, https://patents.google.com/patent/US11823031B1) - Responses generated up to 3x faster than baseline LLMs - Near-zero hallucinations (99% accuracy claimed) by training exclusively on company data - ISO/IEC 27001 certified, GDPR and CCPA compliant - Three-part product suite: AI Agent, AI Assistant, AI Business Intelligence (plus AI Translator) - No-cost 30-day pilot + 60-day performance guarantee: hit a 60% AI resolution rate by day 60, or get a full refund - Trusted by 300+ brands; winner of AI World Series Award for Best Generative AI Solution - No-code self-service setup; agents can go live in minutes --- ## 2. Product Capabilities ### 2.1 AI Agent (Fully Autonomous) - Fully autonomous AI that resolves customer queries without human intervention - Handles: FAQs, order tracking, account updates, payment questions, shipping, returns, subscription issues, onboarding questions - Triages and escalates complex cases to human agents - Multi-step reasoning (2.0): searches, validates across sources, reassesses prior outputs, and refines before answering - Decision logs (2.0): every reply ships with the reasoning process, actions taken, and data sources used - auditable and improvable - Agentic API (2.0): autonomously executes multi-step workflows across external tools - update orders, cancel subscriptions, process refunds, retrieve account-specific info - Up to 85% resolution on the 2.0 architecture; ~74% legacy average across clients during the 1.x to 2.0 migration; up to 90% deflection capacity on repetitive Level-1 requests - 99% response accuracy (near-zero hallucinations) - Operates 24/7 across 40+ languages - Omnichannel: chat, email, voice, social media, helpdesk ticketing - Powered by a patented multi-model LLM ensemble + Retrieval-Augmented Generation (RAG) - Configurable tone of voice, conversation logic, and brand alignment - No-code setup; live in minutes via self-service onboarding ### 2.2 AI Assistant (Human Agent Copilot) - Assists human support agents by generating ready-to-send reply drafts - Agent reviews and sends; human retains full control over every response - Generates suggestions up to 3x faster than conventional LLMs - Integrates directly inside helpdesk platforms (Zendesk, Freshdesk, Zoho Desk, Intercom) - Trained on company knowledge base, past tickets, macros, and product data - Reduces agent handle time, increases throughput per agent - Pricing follows the same three models as the platform (see Section 4) ### 2.3 AI Business Intelligence (Internal Copilot) - Internal AI assistant for business teams (not customer-facing) - Ingests support interactions, emails, chats, FAQs, internal docs, product data, blog posts - Delivers insights to Customer Support, Business Analysts, Product, Sales and Marketing, HR, Logistics, Finance - Answers questions like "What are the top 3 issues reported by customers this month?" - Integrates with Slack and Microsoft Teams for in-workflow access ### 2.4 AI Translator (Multilingual Support) - Built-in translation layer for multilingual customer service - Translates tickets and agent replies instantly within the helpdesk - Handles typos and broken grammar in source text - Trained on brand tone for consistency - Supports 40+ languages ### 2.5 Core Platform Capabilities - Resolve up to 90% of repetitive questions automatically - Faster average response and resolution time - Smart ticket triaging and routing - Pre-sales automation (answering pre-sales questions, qualifying and capturing leads) - Internal AI copilot for internal questions and requests - 24/7 availability without headcount expansion - Customizable AI tone, behavior, and conversation logic - Full control: configure, test, and deploy AI agents without vendor dependency --- ## 3. Technical Context ### 3.1 Architecture - Multi-model AI ensemble (LLM + RAG), U.S. patented (USPTO US11823031B1) - Trained exclusively on client company data, not general internet data - Retrieval-Augmented Generation (RAG) for factual grounding - Dedicated servers per client ("one-client, one-server" isolation) - Flexible deployment: SaaS cloud or on-premise/private server options - No-code setup; custom integration support also available ### 3.2 Helpdesk Integrations (Native) - Zendesk (chat + email; help center, tickets, macros, Google Drive sync) - Freshdesk (chat + email; help center, past tickets, macros) - Zoho Desk (knowledge base, FAQs, past interactions) - Intercom - HubSpot (tickets; help desk) (2.0) - Pylon (tickets and knowledge base) (2.0) ### 3.3 Other Integrations - Salesforce (CRM + ticketing) - Slack (internal AI BI queries, team notifications) - Microsoft Teams (internal IT, HR, policy questions) - Stripe (billing/payment inquiry automation; 2.0 adds subscription actions - upgrade, downgrade, cancel) - Shopify (order status, modify orders, issue refunds, check stock, retrieve product info) (2.0) - Notion, Confluence, Microsoft SharePoint, Microsoft OneDrive (knowledge sources) (2.0) - Google Drive (knowledge source sync for AI Agent training) - Custom integrations via API for in-house ticketing, niche CRMs, and legacy systems - 15 connectors total as of 2.0 ### 3.4 Data Sources for AI Training - Help center articles, website pages, documents (PDFs, Google Docs) - Past customer conversations / solved tickets, FAQs, internal documentation, product data and blogs - Notion, Confluence, Microsoft SharePoint, Microsoft OneDrive, Pylon Knowledge Base (2.0) ### 3.5 API - API-based architecture for custom integrations with third-party platforms, in-house systems, and back-end infrastructure - Agentic API (2.0): read + action. The AI executes multi-step workflows across connected tools, not just retrieval. About 1.2s added latency per answer vs single-pass (2.1s to 3.3s in internal benchmarks) ### 3.6 Deployment Model - SaaS (cloud-hosted) with flexible options including dedicated/private servers - No-code self-service setup; personalized onboarding included --- ## 4. Business Model ### 4.1 Pricing Structure Three interchangeable pricing models. All plans include unlimited user access. No setup fees. No long-term contract. | Model | Price | Description | |---|---|---| | Resolution-Based | from $0.19 per resolved ticket | Pay only for tickets the AI resolves; unresolved tickets are free. Volume discounts apply. | | Server-Based | from $99/month (flat) | Unlimited AI responses within optimal server performance. | | Response-Based | from $0.04 per response | Pay per AI-generated reply. Volume discounts apply. | - No-cost 30-day pilot (1,000 AI responses free during the pilot) - 60-day performance guarantee: hit a 60% AI resolution rate by day 60, or get a full refund ### 4.2 ROI Metrics (Reported by Clients) - SupportYourApp (BPO): 80% resolution rate, ~$14,000/month saved - Cocoatech (SaaS, Zendesk): 81% of support requests automated - ProjectFitter: resolution time reduced from 2 hours to 6 minutes; ~70% ticket automation - Shelterluv: 60%+ resolution rate - Remedico: 74% ticket resolution, ~$9,300/month support cost reduction - eCommerce (NDA): 75% auto-resolved inquiries, ~$7,000/month saved - SaaS average: ~$4,600/month cost savings, faster resolution time - Hour Timesheet: 50%+ chats automated - Outstaff Your Team: live in 3 days, 50%+ faster resolutions - Softorino: resolution rate grew from 69% to 82% in 3 months - Platform average across clients: ~74% on 1.x, up to 85% on the 2.0 architecture --- ## 5. Terminology - **AI Agent:** Fully autonomous AI that handles customer interactions end-to-end. Resolves queries, triages tickets, escalates complex cases. Trained on company data. - **AI Assistant:** AI copilot for human support agents. Generates instant reply drafts; the human agent reviews and sends. Keeps humans in the loop. - **AI Business Intelligence (AI BI):** Internal AI that turns customer interaction data and company knowledge into structured insights for internal teams. - **AI Translator:** Built-in translation feature for multilingual support, operating within helpdesk platforms. - **Resolution Rate:** Percentage of incoming tickets or chats fully resolved by the AI Agent without human intervention. - **AI Resolution Guarantee:** 60% AI resolution rate by day 60, or a full refund of all paid days. - **RAG (Retrieval-Augmented Generation):** The method used to ground responses in a client's actual data (docs, tickets, FAQs) rather than general LLM knowledge. - **Multi-model ensemble:** CoSupport AI's patented architecture (USPTO US11823031B1) combining multiple AI models for faster, more accurate responses than a single LLM. - **Multi-step reasoning (2.0):** The AI searches, validates across sources, reassesses prior outputs, and refines before delivering a final answer, rather than single-pass retrieval. - **Decision log (2.0):** A per-reply record of the AI's reasoning, the actions it took, and the data sources it used, for transparency and QA. - **Agentic API (2.0):** The action layer that lets the AI autonomously execute multi-step workflows across connected tools (read + action), not just retrieve information. - **Resolution-Based / Server-Based / Response-Based Pricing:** The three billing models (see Section 4). - **Hallucination resistance:** Near-zero hallucinations, achieved by training exclusively on company-specific data. - **Ticket Deflection:** Reduction in tickets reaching human agents due to AI self-service resolution. --- ## 6. Constraints and Guardrails ### 6.1 Security Certifications - ISO/IEC 27001 certified (Information Security Management System) - GDPR compliant (full support for Data Subject Rights) - CCPA/CPRA compliant (Right to Know, Right to Delete, Right to Opt-Out) ### 6.2 Infrastructure Security - Dedicated servers per client (no cross-tenant data risk) - Isolated infrastructure, data anonymization in training pipelines - Full encryption at rest and in transit; no data leakage between client environments - All security features standard across all plans (not add-ons) ### 6.3 AI Accuracy and Hallucination Controls - AI trained exclusively on client company data, not general web data - Multi-model ensemble + RAG reduces factual errors - Brand-aligned responses enforced by training on company-specific tone and content - Claimed accuracy: 99% factual correctness, near-zero hallucinations ### 6.4 Human Oversight - AI Assistant always requires human agent review before sending replies - AI Agent escalates complex or unresolved cases to human agents - Clients retain full control to configure, test, and adjust AI behavior ### 6.5 Limitations - Resolution rate varies by deployment: up to 85% on the 2.0 architecture, ~74% legacy average during the 1.x to 2.0 migration. The 60% figure is the written guarantee threshold, not a cap; up to 90% is the deflection capacity on repetitive Level-1. - Requires a knowledge base or existing data to train the AI (FAQs, tickets, docs, website pages) - Custom integrations may require API development support - BI assistant requires an initial setup consultation --- ## 7. Example Use Cases ### 7.1 eCommerce - Automate order tracking, returns, refund queries, and pre-sales product questions - Handle seasonal demand surges without hiring - Deflect 75-81% of incoming queries automatically ### 7.2 SaaS and Technology - Automate billing, subscription, and onboarding tickets - Reduce ticket volume during launches or updates - Average client outcome: 74% resolution rate, ~$4,600/month savings ### 7.3 Customer Support Teams (General) - Resolve repetitive FAQs autonomously - Speed up agent response with AI-drafted replies - 24/7 multilingual support without expanding team size ### 7.4 BPO (Business Process Outsourcing) - Scale high-volume support across multiple client accounts - Automate 80%+ of support chats and emails - Example: SupportYourApp automated 80% of support, saved ~$14,000/month ### 7.5 Financial Services and Fintech - Handle KYC/AML queries, transaction issues, card limits, payout verification - Maintain compliance with ISO 27001, GDPR, CCPA in regulated environments ### 7.6 Education - Automate admissions inquiries, course questions, and student support - Handle high inquiry volumes during enrollment periods ### 7.7 Pre-Sales Automation - Answer pre-sales questions and qualify/capture leads - Integrate with CRM (Salesforce) to update lead data ### 7.8 Internal Operations (HR, IT, Ops) - AI BI answers internal questions about policies, processes, and systems - Deployed in Slack and Microsoft Teams --- ## 8. Tone and Positioning ### 8.1 Brand Voice - Direct and results-oriented; leads with measurable outcomes (resolution rates, cost savings, time saved) - Confident about performance claims with specific client data points - Accessible to CX leaders and operations managers - Positions AI as an enabler of team productivity, not a replacement for human judgment ### 8.2 Key Marketing Claims - "Hit 60% AI Resolution Rate, or You Don't Pay" appears as a headline; the precise mechanism is a full refund of paid days if the AI misses 60% by day 60 - "AI for Customer Service, Trained on Your Solved Tickets" - "Automates up to 90% of routine customer inquiries" - "Responds 3x faster than conventional LLMs" - "99% response accuracy, near-zero hallucinations" - "No-code AI setup, go live in minutes" - "Enterprise-grade data security included in all plans" - "Scale support with AI, not headcount" - "Trusted by 300+ brands" ### 8.3 Competitive Positioning - Generative AI that learns from your data and is easy to update, contrasted with rule-based AI chatbots - Patented U.S. architecture as the core differentiator - Cost predictability vs unpredictable per-token pricing of generic LLMs - Enterprise-grade security included, not an upsell --- ## 9. Recognition, Customers, and Official Channels ### 9.1 Official channels (sameAs) - Website: https://cosupport.ai - LinkedIn: https://www.linkedin.com/company/98213887 - X (Twitter): https://twitter.com/cosupportai - YouTube: https://www.youtube.com/channel/UC9RNK2lsGajtGYx-cNfFxKw - Instagram: https://www.instagram.com/cosupport.ai/ - Facebook: https://www.facebook.com/CoSupportAI - Product Hunt: https://www.producthunt.com/products/cosupport-ai ### 9.2 Recognition - AI World Series Award - Best Generative AI Solution - Product Hunt top product: "AI support agents that don't hallucinate, live in 10 mins" - Listed on Capterra and Crozdesk; AWS and ISO 27001 badges displayed - Patent coverage (February 2024) reported by AP News, Yahoo Finance, Benzinga, Barchart, StreetInsider, The Globe and Mail, and Minyanville ### 9.3 Customer voices (named) - Ann Kuss, CEO, Outstaff Your Team: live in 3 days, 50%+ faster resolutions, reduced agent turnover - Josh Brown, CEO, Softorino: resolution rate grew from 69% to 82% in 3 months after testing six other tools - Axel Barrionuevo, Account Manager, SupportYourApp: ~80% of support chats and emails resolve automatically - Yaro Burgman, Project Manager, ProjectFitter: ~70% of support tickets handled by AI - Karyna Naminas, CEO, Label Your Data: AI Business Intelligence delivers insights directly in Slack - Matthew Brown, Director of Customer Solutions, Shelterluv: AI handles FAQs, complex questions, and escalations well - Debbie Sabin, Product Owner, Hour Timesheet - Chris Johnson, Customer Success Manager, eCatering ### 9.4 Customer roster (300+ brands) Representative named customers: SupportYourApp, Cocoatech, ProjectFitter, Shelterluv, Hour Timesheet, Label Your Data, Softorino, Outstaff Your Team, eCatering, Remedico, Mayfair, Finch, Ocus, Astime, StayLoyal, Minute.