# StayCharted AI Model Trainer > No-code text and image classification trained on your own labeled examples. ## Comparisons - [StayCharted vs Amazon SageMaker Canvas](https://staycharted.com/compare/staycharted-vs-sagemaker-canvas): Compare no-code classification, data imports, evaluation, deployment and billing. - [StayCharted vs Google Teachable Machine](https://staycharted.com/compare/staycharted-vs-teachable-machine): Compare image-model training, supported inputs, deployment and recurring classification workflows. - [StayCharted vs Labelbox: Classification and Labeling](https://staycharted.com/compare/staycharted-vs-labelbox): Compare category prediction workflows with data annotation, curation and model evaluation. - [Compare AI classification platforms](https://staycharted.com/compare): Choose a workflow using setup, evaluation, review and integration requirements. - [StayCharted vs Nyckel](https://staycharted.com/compare/staycharted-vs-nyckel): Sourced feature and plan comparison, checked October 8, 2026. - [Levity alternatives](https://staycharted.com/compare/levity-alternatives): Classification options and a practical evaluation checklist. ## Product and resources - [StayCharted for AI agents](https://staycharted.com/ai-agents): Private classification models for agent workflows, confidence-based human review, MCP and Prediction API integration. - [AI services for MSPs and VARs](https://staycharted.com/blog/ai-services-for-msps-and-vars): Deliver customer classification projects with training, evaluation, integrations, separate customer workspaces and ongoing model review. - [SharePoint oversharing before Copilot](https://staycharted.com/blog/sharepoint-oversharing-copilot): Compare exported access with reviewed permission history; keep admins in control of changes. - [LLM fine-tuning for private enterprise AI](https://staycharted.com/blog/llm-fine-tuning-guide): A 12-step guide to LoRA, QLoRA, evaluation and human review, with banking, legal and medical benchmark comparisons. - [Investment screening for PE and VC](https://staycharted.com/solutions/investment-screening-pe-vc): Categorize company summaries by mandate fit, investment theme or diligence topic; train on reviewed labels and keep investment decisions with your team. - [Classify to industry standards](https://staycharted.com/product-classification-standards): Train on reviewed ETIM, ECLASS, GS1 GPC or HS codes; compare workflows, data preparation and human-review requirements. - [StayCharted Blog](https://staycharted.com/blog): Practical AI tips, workflow ideas, and company updates. - [How AI Model Trainer works](https://staycharted.com/product): Turn labeled text or pictures into a model that applies your categories to new records. Prepare examples, train, check the results, and put the model to work. - [Choose the model for your examples](https://staycharted.com/product/models): Compare AMT’s text and picture models: what they learn, when to try them, and which plans include them. - [Security and your data](https://staycharted.com/security): Understand how StayCharted handles model data, workspace access, masking, assistant connections, and deletion. - [Work with AMT through Claude and ChatGPT](https://staycharted.com/product/ai-assistants): Use a conversation to inspect your models, understand training results, and work with your labeled data while AMT supplies the classification model. - [StayCharted AI Model Trainer in Claude's connector directory](https://claude.ai/directory/staycharted-ai-model-trainer) - Microsoft Marketplace (SaaS, sold through Microsoft): https://marketplace.microsoft.com/en-us/product/staycharted1791510497761.staycharted-ai-model-trainer - Remote MCP server: https://amt.staycharted.com/mcp (OAuth; you choose the workspace) - [Contract clause classification for legal teams](https://staycharted.com/solutions/legal-contract-clause-classification): Train on clauses your team has labeled to organize contract text into your own taxonomy and direct uncertain results to review. - [Classify banking and fintech customer messages](https://staycharted.com/solutions/banking-customer-message-classification): Learn intent categories from labeled customer messages to help support teams route work and review uncertain cases. - [Product categorization for retail and e-commerce](https://staycharted.com/solutions/ecommerce-product-categorization): Apply your catalog taxonomy to product descriptions or photos using models trained on examples your team has already categorized. - [Expense and invoice classification for finance teams](https://staycharted.com/solutions/expense-invoice-classification): Learn expense categories from approved descriptions and supplier data, then suggest consistent labels for new rows. - [Add customer-specific classification to your software](https://staycharted.com/solutions/for-software-companies): Use StayCharted AMT to train and measure category models, then call them from your application through the API. - [AI classification benchmarks on real data](https://staycharted.com/insights/benchmarks): Compare measured text and image classification results, the data behind them, and the review work that remains. - [How AI Model Trainer works behind the scenes](https://staycharted.com/insights/how-it-works): Explore the design decisions behind AMT’s text classifiers and personal-information checks. - [Plans and pricing](https://staycharted.com/pricing) - [API reference](https://staycharted.com/developers) - [Claude and ChatGPT setup](https://staycharted.com/guides/ai-assistants) - [Release notes](https://staycharted.com/release-notes) - [Privacy Policy](https://staycharted.com/legal/privacy) - [Data Processing Addendum](https://staycharted.com/legal/dpa) ## Blog - [Why taxonomy matters to every business](https://staycharted.com/blog/why-taxonomy-matters): How categories affect reporting, fees and compliance, with practical steps for consistent classification and human review. - [Claude in Google Sheets and your trained AI model](https://staycharted.com/blog/claude-google-sheets-trained-ai-model): How conversational help and trained categories can complement each other. StayCharted’s Sheets add-on is available from the Google Workspace Marketplace; the Claude sidebar workflow still needs validation. ## Articles and measured results - [Claude vs. trained classifiers](https://staycharted.com/insights/assistant-vs-trained-model-benchmark): BANKING77 84.7% vs 93.7%; LEDGAR 76.9% vs 84.9%; Medical Abstracts 65.7% vs 63.6% (Claude vs Dedicated AI Model). Different training inputs; full method and limits. - [Food grading by photo](https://staycharted.com/insights/food-grading-image-classification): 98.7% fruit accuracy; cashew defects found: 65% Dedicated versus 40% classifier. - [Contract clauses on LEDGAR](https://staycharted.com/insights/legal-clause-classification-ledgar): 84.9% Dedicated versus 82.7% words plus meaning across 100 clause types. - [SigLIP 2 versus DINOv2](https://staycharted.com/insights/why-amt-image-model-uses-siglip-2): 91.1% product-photo accuracy for SigLIP 2; DINOv2 base led on pet breeds at 96.0%. - [Words and meaning for text classification](https://staycharted.com/insights/why-ai-classifier-uses-words-and-meaning): 92.6% words plus meaning versus 91.0% meaning alone on BANKING77. - [Banking intent classification](https://staycharted.com/insights/ai-for-banking-support): 93.7% measured accuracy across 77 intents; review outcomes are projections. - [PII Masking for AI Training](https://staycharted.com/insights/sensitive-data-check-before-ai-training): Pattern-based checks on every plan; no measured detection-accuracy claim. Names and free-text health details are not detected. - [Medical abstract classification](https://staycharted.com/insights/life-sciences-medical-abstract-classification): 63.6% fine-tuned versus 60.0% embeddings; 78.7% reviewed for projected 97.5%, assuming flagged errors are corrected. ## Product taxonomy workflows - [ETIM Classification with AI for Supplier Catalogs](https://staycharted.com/blog/etim-classification-ai): ETIM classification for wholesalers and manufacturers: train on approved class codes, classify supplier catalogs, and review predictions with confidence scores. - [ECLASS Classification with AI for Product Master Data](https://staycharted.com/blog/eclass-classification-ai): ECLASS classification for manufacturers and procurement teams: learn from approved material codes, classify new items, and review uncertain predictions. - [GS1 GPC Brick Classification with AI](https://staycharted.com/blog/gs1-gpc-brick-classification-ai): GS1 GPC brick classification for brands and retailers: train on approved brick codes, suggest categories for new products, and review supplier catalog data. - [HS Code Classification: AI Suggestions for Review](https://staycharted.com/blog/hs-code-classification-ai): HS code classification support for importers: train on reviewed import history, suggest tariff codes, and keep final decisions with your compliance team or broker. ## Get started - [Do I need a trained model?](https://staycharted.com/guides/assistant-or-classifier): Compare assistants and trained classifiers using examples, evaluation, review effort and data-control requirements. - [Logistics and transportation](https://staycharted.com/solutions/logistics-transportation-classification): Classify shipment exceptions, 3PL requests, and trucking reports using reviewed examples. - [Returns and reverse logistics](https://staycharted.com/solutions/returns-reverse-logistics-classification): Categorize return reasons, visible conditions, and inspection findings for review. - [Automotive: parts, service and warranty](https://staycharted.com/solutions/automotive-classification): Categorize supplier part descriptions, reported service concerns, and warranty claim records using approved examples. - [StayCharted vs Microsoft AI Builder](https://staycharted.com/compare/staycharted-vs-microsoft-ai-builder): Compare custom text classification in StayCharted with Microsoft AI Builder: training data, Power Automate workflows and confidence scores. - [StayCharted vs Amazon Comprehend](https://staycharted.com/compare/staycharted-vs-amazon-comprehend): Compare custom text classification in StayCharted and Amazon Comprehend: labeled examples, console setup, APIs and application integration. - [StayCharted vs Roboflow for Image Classification](https://staycharted.com/compare/staycharted-vs-roboflow): Compare StayCharted and Roboflow for image classification, annotation, model training and deployment. Distinguish category prediction from object detection. - [Azure Custom Vision alternatives](https://staycharted.com/compare/azure-custom-vision-alternatives): Image classification migration options ahead of the September 25, 2028 retirement. - [SharePoint access with Power Automate](https://staycharted.com/guides/sharepoint-access-power-automate): Suggest groups and permission levels for new files from your permission history; route wider access for approval. - [AI-ready file data](https://staycharted.com/blog/ai-ready-file-data): Build consistent document categories from prepared text; use your existing systems for extraction, indexing and permissions. - [StayCharted AI Model Trainer for Google Sheets](https://workspace.google.com/marketplace/app/staycharted_ai_model_trainer/622961385348): Install the published add-on from the Google Workspace Marketplace. Free of charge with paid features. - [MonkeyLearn alternative](https://staycharted.com/compare/monkeylearn-alternative): Custom text classification in Google Sheets, feedback topics and ticket intent; migration from labeled examples, not trained-model import. - [File & SharePoint permissions](https://staycharted.com/solutions/file-permissions): Learn access categories from permissioning history; apply approved results with your own systems. - [Access permissions benchmark](https://staycharted.com/insights/access-permissions-benchmark): 97.8% accuracy on 72,700 held-out group decisions; 5.9% review gives projected 99.4%. - [Customer reply selection](https://staycharted.com/solutions/customer-reply-selection): Train a private classifier to choose an approved reply type and owner from past messages. Returns the reply name; your system maps it to template text and sends it. - [Partners](https://staycharted.com/partners): For SharePoint administrators, consultants and service providers; customer workspaces, file permission suggestions and partner inquiries. - [AI adoption for SharePoint resellers, VARs and MSPs](https://staycharted.com/blog/sharepoint-partners-resellers-vars-msps-ai): Model customer access decisions and repetitive tasks, connect review workflows, and organize customer workspaces. - [Access request approval benchmark](https://staycharted.com/insights/access-request-approval-benchmark): 79.1% routine; 278 of 381 denials routed to review, 103 missed. Application-resource dataset. - [NTFS permissions and access requests](https://staycharted.com/blog/ntfs-access-requests-ai): Learn past decisions to prioritize review; existing approvers retain authority. - [Best practices for training a private AI classifier](https://staycharted.com/guides/best-practices): Prepare labeled examples, compare models, read category accuracy, set review cutoffs and retrain from corrections. - [Classify records with Make](https://staycharted.com/guides/make): Call your trained model from Make’s HTTP module and route uncertain answers for review. ## Practical classification workflows - [How to categorize Google Sheets rows with AI using your own categories](https://staycharted.com/blog/categorize-google-sheets-rows-ai): Automatically categorize Google Sheets rows using your labeled examples. Train a model, fill new rows, check confidence and review uncertain answers. - [AI email routing: sort a shared inbox using your team’s past decisions](https://staycharted.com/blog/ai-email-routing-shared-inbox): Route shared inbox emails with AI trained on past decisions. Classify messages for Outlook or Gmail workflows, then review uncertainty before taking action. - [AI invoice coding: suggest GL accounts and cost centers from past entries](https://staycharted.com/blog/ai-invoice-coding-gl-accounts): Suggest GL accounts and cost centers from previously coded invoice rows. Train on reviewed entries, measure accuracy and keep approval in your finance workflow. - [How to map supplier products to your own catalog categories with AI](https://staycharted.com/blog/supplier-product-category-mapping-ai): Classify supplier product rows into your catalog categories. Train on reviewed mappings, compare results and review uncertain products before importing. - [UNSPSC classification with AI: learn from your reviewed purchase records](https://staycharted.com/blog/unspsc-classification-ai): Classify purchase and invoice rows into UNSPSC codes using reviewed examples. Choose a taxonomy level, test accuracy and review uncertain spend categories. - [How many labeled examples do you need to train an AI classifier?](https://staycharted.com/blog/how-many-examples-train-ai-classifier): Plan labeled training data for AI classification. Learn how category balance, ambiguity, duplicates and held-out testing affect the examples you need. - [AI confidence scores: when should a person review the answer?](https://staycharted.com/blog/ai-confidence-scores-human-review): Understand AI classification confidence scores. Choose a review threshold, compare errors with review effort and keep uncertain answers with a person. - [How to train an image classifier without code](https://staycharted.com/blog/train-image-classifier-without-code): Train on labeled photos, check held-out results and try a new picture from your computer. - [Spend classification with machine learning](https://staycharted.com/blog/spend-classification-machine-learning): Train on reviewed purchase history, choose your taxonomy and evaluate line-item categories. - [Logistics document classification with AI](https://staycharted.com/blog/logistics-document-classification): Categorize BOLs, PODs, invoices and customs documents from prepared text, with human review. - [AI keyword mapping with Google Search Console](https://staycharted.com/blog/ai-keyword-mapping-search-console): Train on reviewed query–page groups and review suggested mappings in Google Sheets. - [Site Search Query Classification for Ecommerce](https://staycharted.com/solutions/site-search-query-classification): Classify site search queries into your own catalog categories. Review zero-result searches and possible gaps in demand using your search history. - [Automate Google Ads Negative Keyword Review with AI](https://staycharted.com/solutions/google-ads-search-terms-ai): Sort Google Ads search terms with a model trained on reviewed account decisions: add as keyword, add as negative, or leave. Prioritize review with confidence scores. - [AI lead qualification from your CRM history](https://staycharted.com/solutions/ai-lead-qualification): Train on reviewed lead outcomes and route categories through your CRM workflow, with confidence-based review. ## Classification for AI products - [Pre-Label Training Data with Your Own Model](https://staycharted.com/solutions/pre-label-training-data): Pre-label text and pictures using your reviewed examples. Export labels and confidence scores, and focus annotation effort on uncertain cases. - [Custom Content Moderation with Your Own Policy](https://staycharted.com/solutions/custom-content-moderation): Apply your own content policy to user messages and AI answers with a model trained on your moderators' decisions. Use it alongside general safety filters. - [Automatic Metadata Tagging for RAG Retrieval](https://staycharted.com/solutions/rag-metadata-tagging): Tag every document for RAG with your own metadata (type, department, product and who may see it) using a model trained on files your team already filed. - [Triage AI Product Feedback by Issue Type](https://staycharted.com/solutions/ai-feedback-triage): Sort thumbs-down feedback on your AI product into your own issue types and owners, with a model trained on feedback your team already sorted. - [AI Agent to Human Handoff, from Your Own Chats](https://staycharted.com/solutions/ai-agent-human-handoff): Decide when your AI agent should hand a chat to a person, with a model trained on the chats your team actually took over. A confidence on every handoff. - [Intent Routing for AI Agents and LLM Apps](https://staycharted.com/solutions/ai-agent-request-routing): Route requests to an agent, tool or model with an intent classifier trained on your own logs. Measure accuracy and use confidence to choose cases for review. - [Search relevance: trained model vs general AI](https://staycharted.com/insights/search-relevance-benchmark): WANDS 80.2% Dedicated AI versus 72.7% Claude; ESCI 59.0% versus 59.9%. Each test uses 4,000 query–product pairs with queries held out from supervised training. - [Open-ended survey coding](https://staycharted.com/solutions/open-ended-survey-coding): Train on coded survey waves to apply your codebook to new verbatims, with multiple codes and confidence-based review. - [Insurance claims triage](https://staycharted.com/solutions/insurance-claims-triage): Train on past claim documents and intake records to suggest categories, queues and cause-of-loss codes; adjusters decide claim outcomes. - [Jev vs Laya vs a trained classifier](https://staycharted.com/compare/jev-vs-laya): Zero-shot text decisions versus training on reviewed examples; workflow and pricing comparison, not an accuracy benchmark. - [Jev vs a trained classifier](https://staycharted.com/blog/jev-vs-trained-classifier): When to use each and how to verify draft labels before training. - [Open-source Laya](https://staycharted.com/blog/laya-open-source-classifier): Self-hosting, fine-tuning and project-reported benchmarks with their limitations. - [Car damage classification benchmark](https://staycharted.com/insights/automotive-damage-classification-benchmark): 94.0% binary damage accuracy; 98.4% among the 77% above cutoff. Parts and rare damage results, methods and review projections. - [Condition monitoring from inspection photos](https://staycharted.com/solutions/condition-based-monitoring): Train on inspector-graded photos; track asset trends in your own system. - [Visual condition monitoring setup guide](https://staycharted.com/blog/condition-based-monitoring-inspection-photos): Prepare photos, evaluate grades, review uncertainty and compare asset condition over time. - [UAD 3.6 appraisal QC with your own rules](https://staycharted.com/blog/uad-3-6-appraisal-qc-ai): Revision routing, photo checks and human review using trained models. - [Maintenance work order classification](https://staycharted.com/guides/maintenance-work-order-classification): Train on maintenance history; categorize work orders through MaintainX, Zapier or the API. - [Amazon Bedrock fine-tuning for classification](https://staycharted.com/blog/amazon-bedrock-fine-tuning): Costs, prompting comparison and training Nova with StayCharted. - [Fine-tune Amazon Nova: step-by-step guide](https://staycharted.com/guides/amazon-bedrock-fine-tuning): JSONL, S3, IAM, deployment and a spreadsheet workflow. - [Gemini fine-tuning for classification: tune it on your team's decisions](https://staycharted.com/blog/gemini-fine-tuning): Fine-tune Gemini in your own Google Cloud project, from a spreadsheet or Google Sheet with StayCharted Business, or directly through Google Cloud. - [Fine-tune Gemini in your own Google Cloud](https://staycharted.com/guides/gemini-fine-tuning): Fine-tune Gemini in your own Google Cloud project, from a spreadsheet or Google Sheet with StayCharted Business, or directly through Google Cloud. - [Fine-tuning on Microsoft Foundry for classification: what it costs and what it takes](https://staycharted.com/blog/microsoft-foundry-fine-tuning): Fine-tune GPT models in your own Microsoft Foundry resource with StayCharted Business. Connect with Microsoft sign-in, train, test and publish. - [How to fine-tune a model on Microsoft Foundry for text classification](https://staycharted.com/guides/microsoft-foundry-fine-tuning): Fine-tune GPT models in your own Microsoft Foundry resource with StayCharted Business. Connect with Microsoft sign-in, train, test and publish. - [Train AI in your own cloud](https://staycharted.com/product/train-in-your-cloud): Train and test private models in Amazon Bedrock, Microsoft Foundry or Google Cloud with StayCharted Business. - [HS code classification benchmark](https://staycharted.com/insights/hs-code-classification-benchmark): 1,167 later CBP rulings; 22.2%/25.1% exact 10-digit accuracy. AI Classifier: 96.2% accuracy on the 6.7% of rows at confidence ≥0.9. - [Classify images in S3](https://staycharted.com/blog/classify-images-s3-bucket): Train from your categorized bucket photos and sort new ones. - [Auto-keyword photos](https://staycharted.com/blog/auto-keyword-photos-your-categories): Add your own categories to picture metadata. - [Object storage guide](https://staycharted.com/guides/object-storage): Connect storage, train from folders and route sorted copies. - [Part identification from photos](https://staycharted.com/insights/part-identification-from-photo-benchmark): 100 part numbers, 1,000 held-out photos. StayCharted AI Image Classifier Model 87.8%; Dedicated AI Image Model 75.3%. - [Parts classification with AI](https://staycharted.com/blog/parts-classification-ai): Identify part numbers from photos or categorize part descriptions with your own examples. - [Open-weight vs closed AI models for training](https://staycharted.com/blog/open-weight-vs-closed-ai-models-training): Ownership, model retirement, fine-tuning costs, portability and StayCharted’s model options. - [Solution partners](https://staycharted.com/partners/solution-partners): Embed AI classification in your software, with private customer models and the Partner API.