🚀 Getting Started with AutoBrain
Welcome to AutoBrain — your no-code AI assistant built for monday.com.
AutoBrain makes it easy to create and deploy powerful predictive machine learning models directly from your monday.com boards, no coding, no spreadsheets, and no data science background needed. Whether you're trying to forecast sales, predict customer churn, or optimize workflows, AutoBrain brings data science capabilities into your team's daily routine—intuitively and instantly.
This guide will walk you through everything you need to get up and running—from installation to building your first model and automating predictions. In just a few clicks, you’ll be making smarter, data-backed decisions right inside monday.com.
1. 🔧 Installation & Setup
• Add AutoBrain to Your Board
Open any board, click on "Add View," and select AutoBrain Dashboard to integrate it as a board view.
2. 🧠 Creating Your First AI Model
• Initiate Model Creation
Welcome to you dashboard, here you will be able to create ML models. The Create Prediction will take you to an easy 5 steps wizard setting the required info create your first ML model.
• Define Your Business Objective
To the objective you wish to get answers for, what is the context the model should follow by
• Select the Target Column to Predict
The Target is simply the column you wish the system to predict its value on your behalf to act upon. For instance, the "Converted" or "End Status" for an opportunity that comes in and you wish to better understand.
• Choose Feature Columns
These features represent the columns/values of the board items. Past examples, including the Target values, will be used to train the system. The ML model's predictions will be based on these features.
Note: The system automatically identifies and prioritizes impactful columns for training, so including more board columns is recommended
• Apply Filters (Optional)
In case you have board items you rather exclude from the system to learn from.
• Set Prediction Strategy
Decide whether you want AutoBrain to provide a categorical prediction (e.g., "Yes"/"No") or a probability score (e.g., 87.7% likelihood).
3. 📊 Utilizing Predictions
• Run Predictions
Once your model is ready, select the items* you wish to analyze. Click on AutoBrain Predict from the Apps section and follow the prompts.
*A maximum of 100 items can be predicted at once.
• View Results
AutoBrain will populate a new column "AutoBrain Prediction" with the predicted outcomes for each item.
You just prefomred your first prediction!
Leverage AutoBrain's native automations to automatically generate predictions when items are created or updated, ensuring real-time insights without manual intervention.
Creating a Prediction Model: The Full Guide
Welcome to the in-depth guide for building machine learning prediction models in AutoBrain for monday.com. This guide walks you through every step of the model creation process, with explanations tailored for non-technical users, plus tips and best practices to get the most out of your data.
✨ Overview
AutoBrain turns your monday boards into intelligent, no-code ML models to help you predict outcomes like lead conversion, customer churn, and pricing estimates. To start, you’ll walk through a 6-step model builder wizard.
Step 1: Select Board
Choose a monday board with historical data to train your model. This board must include the outcome you’re trying to predict — for example, whether a lead was won or lost.
Requirements:
- Minimum: 1,000 items, more items will improve the model accuracy
- Must include exactly one column that you wish to predict (Status, Numbers, Dropdown, etc.)
- Avoid boards with heavy free text fields only, as they're harder for the model to interpret
ℹ️ Tip: Boards tracking deals, customers, or projects with clearly defined outcomes work best. AutoBrain learns by detecting patterns between your input data and these outcomes.
Step 2: Define Business Objective
Pick a goal that aligns with your prediction intent. This step tells AutoBrain what you're trying to predict — for example, whether a lead will convert, or what a customer will pay.
Templates:
- Lead Scoring — Predict whether a lead will convert
- Churn Prediction — Identify users at risk of churn
- Price Forecasting — Estimate numerical values like deal amount
- Custom — Customize the model preferences to your own needs (for advanced users or unique use cases)
ℹ️ Tip: You can create multiple models for the same board using different objectives — one for churn, another for upselling, etc. Why this matters? The business objective determines the type of machine learning used (among classification or regression and more), and shapes every decision that follows.
Step 3: Choose Target Column
This is the column AutoBrain will try to predict — your "ground truth." It's critical that the data here is accurate and meaningful.
Best Practices:
- Use columns with clear, non-overlapping labels
- For classification: ensure each category has at least 100 items for the model to learn from
- Make sure the columns are filled (low missing data, more than 85% filled is recommended)
⚠️ Avoid using target columns that are incomplete or subjective — the model can’t learn from noise. This is important becuase a poor target column = poor results. The model can only learn patterns if the outcome is clean and well defined.
Step 4: Apply Filters
Use filters to include only the most relevant historical data in your model. This helps prevent misleading results.
Example use cases:
- Include leads created in the last 12 months (to avoid outdated behavior)
- Exclude deals marked as "Test" or internal
- Only include deals that reached a specific sales stage
Benefits of filtering:
- Removes noise / “not realistic” items from your dataset
- Focuses the model on current patterns
- Improves performance and accuracy
💡 Why filter? Including outdated or irrelevant items can confuse the model. A clean, relevant dataset is one of the biggest drivers of model success.
Step 5: Select Input Columns
These are the features the model will use to make its prediction — such as lead source, assigned rep, or deal size.
Among the many supported column types:
- Status, People, Dropdown, Date, Numbers, Tags
ℹ️ AutoBrain automatically prepare (cleans and transforms) your data behind the scenes — converting categories into numbers, handling missing values, and more. You don’t need to code or prepare data manually.
💡 Pro Tip: Choose columns that would make sense to a human making the same decision. If you were trying to guess the outcome manually, what info would help you?
Step 6: Set Prediction Strategy
This step defines how the prediction result will appear — do you want a simple answer, or a probability score?
Options:
- Top Category Prediction: Output the most likely result (e.g., "Will Churn")
- Category Confidence Score: Output the probability of an outcome (e.g., "78% chance to churn")
When to use:
- Use Top Category when you need quick decisions (e.g., auto-close a deal)
- Use Confidence Score for prioritization (e.g., focus on leads with >70% conversion chance)
💡 Why it matters: Choosing the right strategy affects how you interpret and act on the model. Confidence scores give more nuance, while top categories are easier to automate.
✅ Final Step: Review & Train
After reviewing your setup, click Train Model. AutoBrain takes care of the rest:
- Prepares and encodes your data
- Trains multiple models
- Selects the best-performing one automatically
Typical training time: Under 3 hours
Post-training insights include:
- Model Accuracy — how often it gets predictions right
- Precision / Recall / AUC — deeper metrics for advanced users
- Feature Importance Chart — shows which columns matter most
- Confusion Matrix — shows where the model got it right vs wrong (classification only)
💡 Retraining tip: Retrain every 4-8 weeks if your process or data changes. The model improves with new patterns.
💡 Pro Tips & Best Practices
- ✉️ Start simple — Begin with boards that have clear outcomes. It’s easier to understand and improve your model this way.
- ⚠️ Balance your classes — If you’re predicting “Will Convert” vs “Won’t Convert,” make sure you have at least 10% of one category. The more balanced data will help the model learn more effectively yet isn’t necessary.
- ♻️ Retrain monthly — Especially if your workflows, users, or customer behavior shift over time.
- 🤖 Use predictions in monday automations — e.g., auto-assign high-scoring leads to your best reps, or flag at-risk customers. This turns insights into immediate action and impact on your buisness.
- 🔍 Review feature importance — Understand which columns drive predictions. This helps you improve your data over time.
⚙️Automations: Maximize Your Ops
This guide explains how to use AutoBrain predictions to trigger actions automatically in your monday.com workflows. You’ll learn how to connect model predictions to native monday automations, creating powerful no-code decision flows.
🚀 What You'll Learn
- How to automatically trigger predictions when items are created or updated
- How to use AutoBrain prediction results in monday automations
- Common automation use cases (e.g., lead routing, risk flagging)
- Best practices and pro tips to avoid pitfalls
1. Automatically Trigger Predictions
After training a model, you can configure AutoBrain to automatically make predictions:
🔄 Trigger Options:
- On Item Creation – Predict as soon as a new item (e.g., lead, deal, project) is added to the board
- On Item Update – Re-run prediction when certain fields (e.g., status, assigned rep) are updated
🔧 How to Set It Up:
- Go to your trained model in AutoBrain
- Click Automation Settings
- Choose "Trigger on item creation" or "Trigger on item update"
💡 Why this matters: Automated predictions let you scale insights without manual steps — every new item is instantly analyzed and ready for next actions.
2. Use Predictions with monday automations
Once a prediction is made, it’s stored in a dedicated column (e.g., "AutoBrain Prediction"). You can reference this in native monday automation recipes.
Example Automations:
- Basic Logic — "When AutoBrain Prediction is High Churn Risk, change Status to Follow Up ASAP."
- Numeric Threshold — "When Confidence Score is greater than 80%, assign to Senior Rep."
- Multi-step Routing — "If Score is under 30%, move item to Low Priority group. If over 70%, notify team lead."
How to Use:
- Go to your board > Automations
- Click Create Custom Automation
- Use condition: When [AutoBrain Prediction Column] is...
- Define your action (assign, move, notify, etc.)
⚠️ Note: If using numeric confidence scores, use number-based conditions. If using categorical predictions, use exact match conditions.
3. Common Automation Recipes
Here are some popular ways to put AutoBrain to work:
🎯 Lead Scoring:
- Assign High-Scoring Leads
If prediction = "Will Convert" AND confidence > 70%, assign to top rep - Auto-Close Cold Leads
If score < 20%, set status to "Closed – Low Priority"
🔥 Risk Management:
- Flag At-Risk Projects
If prediction = "Will Miss Deadline", set status to "Red" and notify project lead
🗂️ Workflow Optimization:
- Auto-Group Items
Move items based on score to "Hot Leads", "Warm", and "Cold" groups - Custom Reminders
Trigger an email or Slack message based on the prediction result
💡 Best Practices
- ✉️ Choose the Right Trigger Events — Avoid triggering predictions on every minor update. Stick to meaningful events preventing consuming all your plan predictions.
- 🔁 Avoid over automating — Too many automations can overlap or conflict. Start small, test each one, and monitor behavior.
- 🧪 Use Confidence Scores for Prioritization — If you’re using numeric scores (e.g., "78% chance to churn"), you can tier actions:
90% = urgent follow-up; 60–89% = standard queue; <60% = lower priority - 🔍 Review Prediction Accuracy Regularly — Don’t "set and forget", check that predictions are driving useful actions. Adjust triggers or retrain models as your workflow evolves.
🧠 Why this matters: Automation amplifies both good and bad decisions. Make sure your data is solid, and that the prediction logic matches your real-world goals.
🎯 Optimize your Objective & Prediction Strategy
Welcome to the focused guide on the two most critical steps in the AutoBrain model-building process: Defining Your Business Objective and Choosing Your Prediction Strategy. These steps shape the core of your prediction model and determine what insights you’ll get — and how you’ll use them.
Welcome to the focused guide on the two most critical steps in the AutoBrain model-building process: Defining Your Business Objective and Choosing Your Prediction Strategy. These steps shape the core of your prediction model and determine what insights you’ll get — and how you’ll use them.
💡 Why These Steps Matter?
These steps are not just technical configurations — they reflect your operational priorities. Defining what you want to predict and how that prediction is delivered ensures your model aligns with your day-to-day goals, from prioritizing leads to preventing churn.
This guide was crafted with ops leaders and their teams in mind. Your challenges are real: lead overload, churn blind spots, and limited resources and team bandwidth. These decisions are about empowering your team to make faster, smarter moves.
Define Business Objective
This is where you define what you want to predict. AutoBrain offers built-in templates for common use cases, or you can set your own custom goal.
Built-in Templates:
- Lead Scoring — Predict whether a lead will convert
- Churn Prediction — Identify customers at risk of leaving
- Deal Forecasting — Estimate the size or value of a deal
- Project Risk — Flag projects that at-risk vs planned time
- Custom Objective — Tailor your model for unique processes (advanced)
How It Works?
Your selected objective determines whether AutoBrain uses classification (yes/no answers) or regression (numeric estimates). For example:
- "Will this lead convert?" → Classification (a yes/no or categorial question)
- "What will this client pay?" → Regression (a number question)
Example Use Cases:
- Sales: Prioritize high-converting leads to focus rep efforts
- Customer Success: Flag accounts with high churn risk for intervention
- Finance: Forecast revenue from open opportunities
Best Practices:
- Start with a use case that has clear outcomes
- Use a board with at least 1,000 rows of historical data
- Keep the outcome column simple and consistent (e.g., "Won" vs. "Lost")
💡 Why this matters? A clear business objective is the foundation of any prediction model. This is your opportunity to bring strategic clarity to messy processes. Whether you’re drowning in leads or trying to spot churn before it happens — this step puts control back in your hands.
Set Prediction Strategy
Once your model knows what to predict, you choose how that prediction is delivered. Do you need a confident decision? Or a probability score that helps you prioritize?
Strategy Options:
- Top Category Prediction
- Outputs the most likely category (e.g., "Will Convert")
- Ideal for quick decisions or automations - Confidence Score
- Outputs a probability (e.g., "78% chance to convert")
- Great for prioritization and nuanced decision-making
When to Use Each:
Use Case
Strategy
Why?
Auto-assign leads
Top Category
Clear action trigger
Prioritize outreach
Confidence Score
Helps rank by likelihood
Churn risk flag
Top Category or Score
Depends on how aggressively you want to act
Revenue forecasting
Score (Regression)
You need a numeric value
💡 Why this matters? The strategy you choose determines how the prediction value will be present in your team’s monday board and by that, how your team could/will act on the prediction. Think automation vs. judgment call.
If you’re in ops, think of this like configuring your alerts or reports — only smarter. Do you need clear-cut actions or a smart dashboard that lets your team prioritize by likelihood?
➕ Bonus: Tips for Choosing Wisely
- If in doubt, use Confidence Scores — You can always round up/down later.
- Match your automation style — Use Top Category if you plan to trigger automations and it is a timely matter; use Confidance if a human will review first.
Why this matters? Your team will thank you for this. Prediction Strategy isn't just a button — it's how the insights you generate translate into workflows, actions, and impact. Nail this, and you turn your model into a smart teammate.
