Agents
Jobs
Knowledge Bases
Marketplace
Apps
Pricing
Sign in Get a Demo

Insulin Blog

Building agents, grounding them in your own data, and putting the results to work.

AI Email Rule Templates: Exact Text, Word for Word
Inbox App

AI Email Rule Templates: Exact Text, Word for Word

A booking link or a required legal sentence has to arrive exactly as written. Insulin's Inbox rules offer three ways to send text, and a rule older than email templates has none until you change its description and save it.

Max Ma
Read article
AI Workspace Admin Checklist: Set Up Insulin
Workspace

AI Workspace Admin Checklist: Set Up Insulin

Two of Insulin's admin settings fail closed, the spending limit blocks rather than warns, and the wrong org role stops a new member from starting a chat. Nine steps, in the order that keeps each one from breaking the next.

Max Ma
EWS Retirement and Your AI Email Assistant
Integrations

EWS Retirement and Your AI Email Assistant

Microsoft starts disabling Exchange Web Services in October 2026 and asks admins to check every third-party app. What Insulin's docs say about its Outlook connection, and the report in your tenant that settles it.

Shirley Guo
OWASP Agentic Top 10: An AI Workspace Checklist
Trust

OWASP Agentic Top 10: An AI Workspace Checklist

The OWASP Agentic Top 10 as a security reviewer's checklist: for each risk, the control Insulin's docs describe and what they don't promise. Fours' reading of its own docs, not OWASP's assessment.

Sophia Faria
Read-Only Access for AI Agents: Where It's Set
Security

Read-Only Access for AI Agents: Where It's Set

Insulin's allowlist decides which systems an agent can reach, not which tools. Read-only is set in each system's own credential, and for two of them the documented setup can't hold an agent to reading.

Chengjun Yuan
AI Agent for BigQuery: Let It Query, Not Change
Integrations

AI Agent for BigQuery: Let It Query, Not Change

The query tool runs the SQL the agent writes, so what keeps a warehouse agent from changing data is its service account's IAM roles. Here is the pair to grant.

Shirley Guo
AI Model Retirement: What Happens to Your Agents
Integrations

AI Model Retirement: What Happens to Your Agents

Claude Sonnet 4.5 retires on November 30, 2026, and a custom agent pinned to it won't quietly switch models: it stops. Every place Insulin pins a model, and what to change first.

Shirley Guo
Use Notion Docs With AI Agents (and Confluence)
Integrations

Use Notion Docs With AI Agents (and Confluence)

Notion isn't a knowledge-base connector yet and Confluence has none, but agents can reach both live. Which agents can, what they can change, and when to index instead.

Shirley Guo
Why AI Gives Different Answers to Different People
Trust

Why AI Gives Different Answers to Different People

Your colleague asked the same question and got a different answer. Usually neither assistant is broken: each was answering from different accounts, documents, models and memory.

Sophia Faria
Agent Skill vs Knowledge Base: Where Runbooks Go
Knowledge Bases

Agent Skill vs Knowledge Base: Where Runbooks Go

Insulin's docs file runbooks under skills and under knowledge bases, and both are right. The steps an agent follows go in a skill; the reference around them goes in a knowledge base.

Zee Chen
Attach a File to Chat, or Add to a Knowledge Base?
Knowledge Bases

Attach a File to Chat, or Add to a Knowledge Base?

You have a document and a question. Attaching it to the chat and adding it to a knowledge base are different tools — here is which one the document needs.

Zee Chen
The AI Email Assistant for Outlook vs Gmail
Inbox App

The AI Email Assistant for Outlook vs Gmail

Most of Insulin's Inbox works the same on Outlook and Gmail. Three things don't: attachments, where AI drafts live, and editing mailbox drafts in Inbox. Here is the documented list for an M365 team.

Shirley Guo
Change a Knowledge Base's Embedding Model Safely
Knowledge Bases

Change a Knowledge Base's Embedding Model Safely

Switching a knowledge base's embedding model re-embeds every document, and search hides each one until it is rebuilt. The provider choice, and a switch-over runbook that never mixes two models' vectors.

Zee Chen
Turn a Team Procedure Into an AI Agent Skill
Agents

Turn a Team Procedure Into an AI Agent Skill

One procedure per skill, and a Trigger that says when it applies. A worksheet for turning a runbook your team already follows into an agent skill.

Gabriel Paiva
How to Stop Work in Insulin: Replies, Goals, Jobs and Inbox
Trust

How to Stop Work in Insulin: Replies, Goals, Jobs and Inbox

Reloading doesn't stop a reply, a goal keeps running after you leave, and an org admin can't reach your jobs. Where each stop control sits, and what it ends.

Sophia Faria
Usage Limit or Billing Pause? Diagnose a Stopped Insulin Reply
Workspace

Usage Limit or Billing Pause? Diagnose a Stopped Insulin Reply

A model's usage limit and your organization's monthly spending limit can both stop an Insulin reply, and each has its own fix. Read the card before you retry or top up.

Max Ma
Which Models Serve Insulin Chat and Knowledge Bases?
Security

Which Models Serve Insulin Chat and Knowledge Bases?

With no provider of your own, Insulin chat runs on a Fours-hosted pool and knowledge bases embed with BGE-M3. The documented roster, who serves it, and how to check.

Chengjun Yuan
AI Usage by Person, App and Model: Who Spent It?
Workspace

AI Usage by Person, App and Model: Who Spent It?

Five pivots of one month's AI spend, each totalling the same amount: what the No user and No app rows mean, how filters narrow it, and why the table trails the balance.

Max Ma
Where Did My Channel Reply Go? Mentions, Planner and Sub-threads
Agents

Where Did My Channel Reply Go? Mentions, Planner and Sub-threads

Asked in a team channel and heard nothing back? Usually the reply is in a sub-thread, or your mentions sent the message somewhere you didn't expect.

Gabriel Paiva
AI-Built App Data Storage: What Survives a Change
Custom Apps

AI-Built App Data Storage: What Survives a Change

An AI-built app that stores data gets its own database. What code changes, failed migrations, restores and deletes do to it, and a snapshot to take first.

Qiuyang Luo
Cloudflare Workers AI BYOK and DigitalOcean Models
Integrations

Cloudflare Workers AI BYOK and DigitalOcean Models

Run Insulin agents on Cloudflare Workers AI or DigitalOcean with your own key: the five models each, what the connect check proves, and what else to set.

Shirley Guo
Deprecate, Replace, or Delete a Knowledge-Base Document
Knowledge Bases

Deprecate, Replace, or Delete a Knowledge-Base Document

Deprecate takes an outdated file out of AI answers but keeps it. A same-name upload replaces it, and a crawled or synced document is fixed at its source.

Zee Chen
What an Inbox CRM Action Changed—and Who Authorized It
Trust

What an Inbox CRM Action Changed—and Who Authorized It

An Inbox CRM action leaves a two-part record: a receipt when you click Run in chat, and an outcome when the run ends. What each part proves, and where it stops.

Sophia Faria
Preview an Inbox Rule Without Sending Anything
Inbox App

Preview an Inbox Rule Without Sending Anything

Test an AI email rule without sending anything: what each of Preview's four steps checks, what one sample can't prove, and why a live rule goes quiet.

Max Ma
Why Is My AI Knowledge Base Not Finding Documents?
Knowledge Bases

Why Is My AI Knowledge Base Not Finding Documents?

When an agent can't find a document you added, the reason is usually on screen: a Test search, an indexing status, a stale sync, or the bases it searches.

Zee Chen
Email Prompt Injection: Check What the AI Read
Security

Email Prompt Injection: Check What the AI Read

A formatted email has two versions, and the AI reads the one you do not see. How Insulin shows you the text the AI read, and why its warning never blocks.

Chengjun Yuan
Import Agent Skills (SKILL.md): Read Them First
Agent Marketplace

Import Agent Skills (SKILL.md): Read Them First

Only a single markdown file or a raw SKILL.md URL gets a review form before it saves. A .zip or a GitHub folder imports directly, so the reading is yours.

Qiuyang Luo
Transfer Ownership of AI Agents Before Someone Moves On
Workspace

Transfer Ownership of AI Agents Before Someone Moves On

Transfer ownership of AI agents, apps and channels before the owner's admin role changes: who can receive each, and what an org admin can't reach.

Chengjun Yuan
Verify an AI-Generated Dashboard From a Template
Custom Apps

Verify an AI-Generated Dashboard From a Template

Five checks to verify an AI-generated dashboard. Two need a starter template, and View query appears only in apps created from one since September 14.

Sophia Faria
AI Model Allowlist: Which Vendors See Your Prompts
Trust

AI Model Allowlist: Which Vendors See Your Prompts

An AI model allowlist decides which providers may serve your prompts. In Insulin it fails closed — and one missing row removes Fours' hosted models.

Sophia Faria
Does an AI Email Assistant Store Your Email?
Security

Does an AI Email Assistant Store Your Email?

Not the message body: Insulin fetches it live from Gmail or Outlook and never stores it. What it does keep is a short, exact list, and it is not nothing.

Chengjun Yuan
Edit What Your AI Assistant Remembers About You
Trust

Edit What Your AI Assistant Remembers About You

How to edit what your AI assistant remembers in Insulin: where memories sit, how corrections work, what Clear All removes, and why wording decides recall.

Sophia Faria
Least Privilege for AI-Built Apps: Whose Account?
Security

Least Privilege for AI-Built Apps: Whose Account?

Sharing an AI-built app settles who can open it, not whose accounts it spends. Its Salesforce and Slack calls run as the organization, never the runner.

Chengjun Yuan
Scheduled AI Job Not Running? Where to Look First
Jobs

Scheduled AI Job Not Running? Where to Look First

A scheduled AI job that didn't fire at 09:00 usually says why on its own page: the badge, the Next timestamp, a run status, or the Steps transcript.

Shirley Guo
Where the Approval Gate Is Armed, and Where It Isn't
Trust

Where the Approval Gate Is Armed, and Where It Isn't

The tool-approval card is a guardrail on one surface, not a workspace-wide guarantee. Knowing which surface is the difference between a control and a belief.

Sophia Faria
Disable or Disconnect: What Happens to Your Mail
Security

Disable or Disconnect: What Happens to Your Mail

One pauses processing and keeps everything. The other permanently deletes what was stored and clears your writing style. They sit next to each other.

Chengjun Yuan
Which Events Actually Run a Job, and Which Get Filtered
Jobs

Which Events Actually Run a Job, and Which Get Filtered

A job with no filter runs on every event it receives. A filter expression is what turns a firehose into an automation, and skipped runs are recorded rather than lost.

Shirley Guo
AI Is Paused: Which of the Four Reasons, and the Fix
Workspace

AI Is Paused: Which of the Four Reasons, and the Fix

Insulin names the reason it stopped, because each one has a different fix. Adding credit will not clear a limit you set yourself — and that is the most common wasted support ticket.

Max Ma
Writing an Email Rule in Plain English
Inbox App

Writing an Email Rule in Plain English

Two fields, one of them optional. You describe what should happen and Save does the rest — and when the model is unavailable, Save fails rather than writing a rule that matches everything.

Max Ma
How to Keep a Long-Running AI Task on Track
Workspace

How to Keep a Long-Running AI Task on Track

Work that does not finish in one turn needs somewhere to live. Goals, sub-threads and memory are three different answers, and picking the wrong one is why long tasks drift.

Max Ma
One Thread, Many Agents: How Channels Work
Agents

One Thread, Many Agents: How Channels Work

A channel is a group chat where several people and several agents work in one shared thread. What changes is not the chat — it is who is allowed to act, and who coordinates.

Gabriel Paiva
Who Picks the Model, and What If It Fails?
Trust

Who Picks the Model, and What If It Fails?

Most AI tools hand you a model dropdown. Insulin has no per-conversation picker — it selects the model and fails over when a provider breaks. Here is why that is a reliability decision, not a missing feature.

Sophia Faria
Per-Seat or Usage-Based: AI Agent Pricing
Workspace

Per-Seat or Usage-Based: AI Agent Pricing

Per-seat pricing charges for the right to have an account. Usage-based pricing charges for work actually done. The difference shows up in who on your team feels free to try something.

Max Ma
A 90-Day Enterprise AI Adoption Plan That Works
Workspace

A 90-Day Enterprise AI Adoption Plan That Works

Ninety days is enough to get real work running on AI, and not enough to boil the ocean. A three-phase plan with a gate at the end of each one.

Max Ma
The AI Agent Sandbox: Test Before Production
Trust

The AI Agent Sandbox: Test Before Production

Knowing what to test is half the problem. The other half is where — because an agent's first real write should not land in the system your business runs on.

Sophia Faria
AI Email and CRM Automation, With Approval
Inbox App

AI Email and CRM Automation, With Approval

The reply and the CRM update are the same piece of work, done twice. How to automate both as one flow that still waits for a person before anything leaves.

Shirley Guo
AI Model Governance: Choosing by Workflow Risk
Trust

AI Model Governance: Choosing by Workflow Risk

Picking one model for the whole company means overpaying for triage and under-thinking the work that matters. Choose per workflow, by what failure costs.

Sophia Faria
AI Training for Employees Who Work With Agents
Workspace

AI Training for Employees Who Work With Agents

Prompt tips are the least useful thing you can teach. What people actually need is the judgement to delegate work, check it, and know when to stop.

Max Ma
Handling AWS Support Cases With an AI Agent
Integrations

Handling AWS Support Cases With an AI Agent

Half of a support case is assembling evidence you already have. That part an agent can do — the filing and the replying still wait for a person.

Shirley Guo
The Enterprise AI Operating Model: Central or Not
Workspace

The Enterprise AI Operating Model: Central or Not

Centralize AI and it becomes a queue. Federate it and you get forty versions of the same agent. The operating model that avoids both.

Max Ma
How an Enterprise AI Workspace Fits Together
Workspace

How an Enterprise AI Workspace Fits Together

Most companies now own a dozen disconnected AI tools. An enterprise AI workspace is the alternative: one place where chat, knowledge, agents and jobs share a spine.

Max Ma
AI Agent Lifecycle: Ownership to Retirement
Agent Marketplace

AI Agent Lifecycle: Ownership to Retirement

Your first agent is easy. Your fortieth is a governance problem. How to manage an agent from request to retirement before the portfolio manages you.

Qiuyang Luo
AI Agent Tools vs Skills vs Integrations
Agents

AI Agent Tools vs Skills vs Integrations

Tool, skill, integration — three words the AI agent industry uses interchangeably and shouldn't. What each one actually is in Insulin, and which your problem needs.

Gabriel Paiva
AI Agents for Procurement: Intake to Approval
Agents

AI Agents for Procurement: Intake to Approval

Procurement is document-heavy, cross-functional, and accountable — a strong fit for agents, with a hard line agents must not cross. Where they help, and where they can't.

Gabriel Paiva
Building an AI Center of Excellence That Sticks
Workspace

Building an AI Center of Excellence That Sticks

Most AI Centers of Excellence become a committee that ships slide decks. How to build one that business teams actually use — light on governance, heavy on reuse.

Max Ma
BYOK vs Managed Models for Enterprise AI
Security

BYOK vs Managed Models for Enterprise AI

Use the platform's models, or bring your own provider keys? The trade is cost, control, and data path. How to decide, and what changes either way.

Chengjun Yuan
How to Build an AI Executive-Briefing Agent
Jobs

How to Build an AI Executive-Briefing Agent

A daily brief a leader can trust has to be current, concise, and sourced. How to build one as a scheduled AI job that cites every claim it makes.

Shirley Guo
How to Design Human Handoffs for AI Agents
Trust

How to Design Human Handoffs for AI Agents

Safe autonomy isn't an agent that never stops — it's one that stops well. How to design the moment an AI agent hands work back to a person.

Sophia Faria
How Humans and AI Agents Share a Workspace
Workspace

How Humans and AI Agents Share a Workspace

The interesting question isn't whether AI agents replace people — it's how people and agents work in the same space. What good collaboration looks like day to day.

Max Ma
Keeping an AI Knowledge Base Fresh Over Time
Knowledge Bases

Keeping an AI Knowledge Base Fresh Over Time

A knowledge base is only as trustworthy as its most out-of-date document. The owners, refresh habits, and pruning that keep grounded answers actually current.

Zee Chen
Single Agent vs Multi-Agent: When to Split
Agents

Single Agent vs Multi-Agent: When to Split

Multi-agent systems are fashionable, and often one well-scoped agent is simpler and better. A complexity test for when a workflow actually needs more than one.

Gabriel Paiva
How to Build an AI Account-Research Agent
Agents

How to Build an AI Account-Research Agent

Reps spend hours researching an account before a call and still miss the one thing that mattered. How to build an agent that does the research and shows its sources.

Gabriel Paiva
AI Agent Audit Logs: What Enterprises Record
Security

AI Agent Audit Logs: What Enterprises Record

When an AI agent does something wrong, the first question is 'what exactly happened' — and most teams cannot answer it. What an agent audit log has to capture, and why.

Chengjun Yuan
AI Agent Incident Response for Enterprises
Security

AI Agent Incident Response for Enterprises

An AI agent will eventually take an action it should not have. The teams that handle it well decided how to detect, contain, and recover before it happened — not after.

Chengjun Yuan
AI Agents for Finance Operations, Explained
Agents

AI Agents for Finance Operations, Explained

Finance runs on recurring, rule-bound work that is perfect for agents and unforgiving of mistakes. Where AI agents fit the close, reconciliation, and reporting — safely.

Gabriel Paiva
AI Agents vs Workflows vs Copilots: Which to Use
Agents

AI Agents vs Workflows vs Copilots: Which to Use

Three words for three different things, used as if they were one. What separates an AI agent from a workflow from a copilot, and how to pick the right one per task.

Gabriel Paiva
Change Management for Enterprise AI Rollouts
Workspace

Change Management for Enterprise AI Rollouts

The AI rollout that fails rarely fails on the technology. It fails because people were handed a tool and left to figure out whether it was safe to trust. A rollout plan.

Max Ma
Enterprise AI Spend Governance: A Practical Guide
Security

Enterprise AI Spend Governance: A Practical Guide

AI spend does not blow up on the model price — it blows up on usage nobody is watching. How to govern AI cost by user, app, and model before the bill teaches you.

Chengjun Yuan
Enterprise AI Adoption Metrics That Matter
Workspace

Enterprise AI Adoption Metrics That Matter

Logins prove people opened the tool, not that it changed their work. The adoption metrics that actually predict whether enterprise AI sticks — and how to read them.

Max Ma
Permission-Aware Enterprise Search for AI Agents
Knowledge Bases

Permission-Aware Enterprise Search for AI Agents

An AI agent that searches everything will eventually answer from a document the asker was never allowed to see. How permission-aware search stops that at retrieval.

Zee Chen
Shadow AI: How to Find and Govern It at Work
Security

Shadow AI: How to Find and Govern It at Work

Employees are already using AI you did not sanction. Banning it drives it underground; ignoring it leaks data. How to find shadow AI and govern it into the open.

Chengjun Yuan
AI Agent Failure Handling: Retries to Rollbacks
Jobs

AI Agent Failure Handling: Retries to Rollbacks

An agent that runs unattended will fail sometimes. What matters is what happens next — retry, degrade, hand to a person, or leave nothing behind to undo.

Shirley Guo
AI Agent Observability: What Business and IT Teams Should Monitor
Automation

AI Agent Observability: What Business and IT Teams Should Monitor

Once an agent runs unattended, you monitor it or you trust it blindly. What business and IT teams should watch — completion, escalation, groundedness, cost, latency, drift — and where each signal lives.

Shirley Guo
AI Agent Risk Assessment: A Reusable Business Workflow Template
Trust

AI Agent Risk Assessment: A Reusable Business Workflow Template

Before you grant an agent access, score the risk. A reusable template that rates data sensitivity, reversibility, autonomy, blast radius, oversight, and auditability — and decides what to gate.

Sophia Faria
AI Agents for IT Help Desks: A Practical Automation Blueprint
Agents

AI Agents for IT Help Desks: A Practical Automation Blueprint

A blueprint for automating the IT help desk with AI agents: triage and route tickets, answer from your IT docs with cited sources, and gate every account change on human approval.

Gabriel Paiva
AI Use-Case Prioritization: How to Choose the First Five Business Workflows
Automation

AI Use-Case Prioritization: How to Choose the First Five Business Workflows

Everything looks automatable in a workshop. A scored framework — value, frequency, data, reversibility, a checkable 'done', and integrations — picks the first five that actually pay off.

Max Ma
AI Agents for Revenue Operations: 10 Workflows Worth Automating
Agents

AI Agents for Revenue Operations: 10 Workflows Worth Automating

RevOps runs on recurring, deadline-shaped, and change-triggered work — exactly what an agent handles well. Ten workflows to automate, and where a human still approves.

Shirley Guo
Build vs Buy Enterprise AI Agents: A Decision Framework
Agents

Build vs Buy Enterprise AI Agents: A Decision Framework

A framework makes the prototype easy. Production ownership — governance, integrations, audit, reliability, security — stays expensive. A decision framework for build vs buy.

Gabriel Paiva
Enterprise AI Agent Memory: What Should Agents Remember—and Forget?
Agents

Enterprise AI Agent Memory: What Should Agents Remember—and Forget?

"Give the agent memory" is usually four different requests wearing one word. Which kind of memory each business fact belongs in, and what an agent should never keep.

Gabriel Paiva
The Enterprise AI Maturity Model: From Personal Copilots to Governed Agentic Work
Workspace

The Enterprise AI Maturity Model: From Personal Copilots to Governed Agentic Work

Enterprise AI adoption is a ladder, not a switch. Five levels — copilots, grounded chat, scoped agents, scheduled jobs, governed operations — each with the capability and the governance it needs.

Max Ma
How to Measure AI Answer Quality and Source Coverage
Search

How to Measure AI Answer Quality and Source Coverage

Fluency tells you nothing about whether an AI answer is right. The metrics that do — groundedness, source coverage, retrieval recall — and how to compute each one on your own knowledge base.

Zee Chen
25 Security Questions to Ask an Enterprise AI Agent Vendor
Security

25 Security Questions to Ask an Enterprise AI Agent Vendor

A buyer's checklist for evaluating an enterprise AI agent platform — 25 questions across data, access, scoping, actions, audit, and deletion, with the evidence to demand for each.

Chengjun Yuan
AI Agents for Customer Support: From Answering Questions to Resolving Cases
Agents

AI Agents for Customer Support: From Answering Questions to Resolving Cases

Support AI is moving past deflection. Here is how an agent grounds in your help docs, acts across systems with scoped access and approvals, and escalates when it should.

Gabriel Paiva
Enterprise AI Data Privacy: A Practical Guide for Business Teams
Trust

Enterprise AI Data Privacy: A Practical Guide for Business Teams

What company data can safely enter an AI workflow — and the operating rules for classification, access, retention, residency, and deletion that keep it that way.

Sophia Faria
From AI Pilot to Production: The Enterprise Readiness Checklist
Agents

From AI Pilot to Production: The Enterprise Readiness Checklist

A demo that impresses is not an agent you can trust in production. Here is the cross-functional readiness checklist for taking an AI agent live.

Gabriel Paiva
How to Calculate the ROI of AI Agents for Business Workflows
Agents

How to Calculate the ROI of AI Agents for Business Workflows

A transparent framework for AI agent ROI: value from time saved, rework avoided, faster cycles, and wider coverage, minus its build and review cost.

Gabriel Paiva
How to Choose the Right Enterprise AI Agent Platform
Agents

How to Choose the Right Enterprise AI Agent Platform

The demo is easy. The requirements matrix is where platforms separate: knowledge grounding, integrations, governance, observability, and how an agent actually deploys.

Gabriel Paiva
How to Evaluate AI Agents: A Business-Focused Scorecard
Trust

How to Evaluate AI Agents: A Business-Focused Scorecard

Response quality is not the same as good work. An AI agent evaluation framework that scores the whole task — completion, correctness, policy, escalation, cost, and latency.

Sophia Faria
How to Test an AI Agent Before It Touches a Business System
Trust

How to Test an AI Agent Before It Touches a Business System

Before you grant an agent write access, put it through a test suite: golden tasks, adversarial cases, scope checks, tool-failure handling, and approvals.

Sophia Faria
How to Write Reliable Instructions for Enterprise AI Agents
Agents

How to Write Reliable Instructions for Enterprise AI Agents

Most agent failures start with vague ownership and undefined exceptions. Reliable AI agent instructions read like an operating contract, not a prompt.

Gabriel Paiva
RAG vs Fine-Tuning for Business AI: Which One Do You Need?
Knowledge Bases

RAG vs Fine-Tuning for Business AI: Which One Do You Need?

Teams reach for fine-tuning when what they actually need is current company knowledge. Here is what each does, and why retrieval is the default for factual business context.

Zee Chen
How to See Past Runs of a Scheduled AI Job
Jobs

How to See Past Runs of a Scheduled AI Job

An unattended agent ran overnight. The runs table on each job's detail page shows every past execution — when it ran, whether it worked, what it did.

Shirley Guo
Build an Internal Dashboard With AI, No Code
Custom Apps

Build an Internal Dashboard With AI, No Code

Type a sentence, get a working dashboard. What happens in between — generate, compile, safety-check — is the part that decides whether you can trust it.

Qiuyang Luo
Connect Google Drive to an AI Knowledge Base
Knowledge Bases

Connect Google Drive to an AI Knowledge Base

Beyond uploading files: connect GitHub, Google Drive, or ClickUp to an Insulin AI knowledge base, or crawl a public site, and read sync status per source.

Zee Chen
Auto-Draft a Follow-Up When a Deal Stage Changes
Inbox App

Auto-Draft a Follow-Up When a Deal Stage Changes

Wire a Salesforce stage change to a voice-matched follow-up and a staged private offer — with an approval gate so nothing sends until you say so.

Max Ma
How to Build a Grounded AI Agent for Marketplace Ops
Agents

How to Build a Grounded AI Agent for Marketplace Ops

The six-step build for a document-grounded marketplace ops agent in Insulin: name it, scope it, pick a model, attach knowledge bases, allowlist integrations.

Gabriel Paiva
How to Schedule a Recurring AI Agent Job
Jobs

How to Schedule a Recurring AI Agent Job

Turn a chat request into an unattended schedule: pick the agent, write the task in plain English, choose the trigger, review the runs. A step-by-step walkthrough.

Shirley Guo
How to Share an Internal App with Roles
Custom Apps

How to Share an Internal App with Roles

Roll one app out to everyone with a safe default role, then grant a named few more — the Viewer, User, Editor, Admin, Owner ladder, org-wide.

Qiuyang Luo
Install an AI Agent for the Whole Organization
Agent Marketplace

Install an AI Agent for the Whole Organization

The eval-first way to install an AI agent for the whole organization: try it on your own account first, then roll the same component out org-wide once it earns its place.

Qiuyang Luo
Restrict Which Integrations an AI Agent Can Access
Agents

Restrict Which Integrations an AI Agent Can Access

The integration allowlist is the safety boundary for an AI agent. Grant only the connected systems its job needs, and it cannot reach data outside them.

Gabriel Paiva
How to Weight Vector vs Keyword Search
Search

How to Weight Vector vs Keyword Search

Hybrid search has a dial. How to weight vector vs keyword matching in a knowledge base, when to pick a pure mode, and how to read the relevance score.

Zee Chen
Agents and Skills: What the Difference Is
Agents

Agents and Skills: What the Difference Is

An agent is who you talk to. A skill is what it knows how to do. Getting the distinction right decides whether you build one thing or twenty.

Gabriel Paiva
Build an Internal Tool Without Writing Code
Custom Apps

Build an Internal Tool Without Writing Code

The dashboard nobody built is usually the one that needed three fields and a filter. Describing it is now faster than requesting it.

Qiuyang Luo
When to Build a Custom App, Not a Dashboard
Custom Apps

When to Build a Custom App, Not a Dashboard

A dashboard shows you a number. An app lets someone do something about it. Choosing wrongly is why so many dashboards get opened once.

Qiuyang Luo
Setting Up Email Rules Without Writing Code
Inbox App

Setting Up Email Rules Without Writing Code

A rule is four decisions: which emails, what context to use, what to produce, and where it goes. None of them is a line of code, and one of them matters more than the rest.

Max Ma
Is Your Company Data Used to Train Models?
Trust

Is Your Company Data Used to Train Models?

The short answer is no. The longer answer is worth having, because it is the question that decides whether anyone uploads the documents that make this useful.

Sophia Faria
When to Use a Job Instead of a Conversation
Jobs

When to Use a Job Instead of a Conversation

If you have asked an agent the same thing three times, you have written a job by hand. The difference is whether anyone has to be there.

Shirley Guo
Why a Knowledge Base Beats a Wiki for Agents
Knowledge Bases

Why a Knowledge Base Beats a Wiki for Agents

A wiki waits to be searched. A knowledge base is read on every question, including inside jobs nobody is watching. That is the whole difference.

Zee Chen
Your LinkedIn Network, on Autopilot
Jobs

Your LinkedIn Network, on Autopilot

A scheduled Insulin agent runs LinkedIn prospecting in its own cloud browser — laptop off — and leaves a ranked shortlist with drafts, sending nothing.

Chengjun Yuan
What the Insulin Agent Marketplace Actually Is
Agent Marketplace

What the Insulin Agent Marketplace Actually Is

It has the word marketplace in it and nothing to do with AWS, Azure or Google Cloud. One distributes software to buyers. This one distributes agents to your own team.

Qiuyang Luo
When the Inbox App Acts, and When It Asks
Trust

When the Inbox App Acts, and When It Asks

Insulin never sends email on its own. That is a deliberate default, and the interesting question is which of your rules deserves an exception.

Sophia Faria
Controlling Which Agents See Which Documents
Knowledge Bases

Controlling Which Agents See Which Documents

One agent should know your pricing floor. Another should not. Document scope is set per agent, and the default for org knowledge bases is stricter than people expect.

Chengjun Yuan
How AI Learns to Write in Your Own Voice
Inbox App

How AI Learns to Write in Your Own Voice

A draft that does not sound like you costs more to fix than to write. How a voice profile is built from your sent mail, and what it can and cannot capture.

Max Ma
Build Your Own AI Agent, or Install One?
Agents

Build Your Own AI Agent, or Install One?

Once one person has built a good specialist, the next four should not rebuild it. When to start from a catalogue agent, and when to build your own.

Gabriel Paiva
Who Can Run What: AI Workspace Permissions
Workspace

Who Can Run What: AI Workspace Permissions

Sharing an AI agent across a company raises a question a personal tool never does: who can change it. How roles and install scopes fit together.

Chengjun Yuan
Scheduled or Event-Driven: Timing AI Jobs
Jobs

Scheduled or Event-Driven: Timing AI Jobs

A job runs once, on a timer, or when something happens. Choosing the wrong trigger is why automation arrives too late or runs constantly for nothing.

Shirley Guo
Keeping a Human in the Loop on AI Agents
Trust

Keeping a Human in the Loop on AI Agents

An agent that can act needs a gate before it acts. How approval works, and the difference between a review that catches things and one people click past.

Sophia Faria
Why Every AI Answer Needs a Cited Source
Knowledge Bases

Why Every AI Answer Needs a Cited Source

An uncited AI answer asks for trust it has not earned. What citations change about how teams adopt AI, and how to read one properly when it arrives.

Zee Chen
Hybrid vs Vector Search: Which Grounds Better
Knowledge Bases

Hybrid vs Vector Search: Which Grounds Better

Vector search finds meaning, keyword search finds exact terms, and business documents need both. How hybrid retrieval works and when to change its weight.

Zee Chen
Which Documents Can an AI Knowledge Base Read?
Knowledge Bases

Which Documents Can an AI Knowledge Base Read?

File formats, size limits, crawling, and connectors — what an Insulin knowledge base accepts, and which source is the right one to start with.

Shirley Guo
How to Ground an AI Agent in Your Documents
Knowledge Bases

How to Ground an AI Agent in Your Documents

Grounding separates an AI answer you can check from one you must trust. How to attach your own documents to an agent, and what to load first.

Zee Chen
One AI Assistant, or a Team of Specialists?
Agents

One AI Assistant, or a Team of Specialists?

A general assistant knows a little about everything and is accountable for nothing. Why scoped AI specialists behave better, and where to draw the lines.

Gabriel Paiva

Try Insulin on your own work

Get a walkthrough of agents, knowledge bases, and jobs running on your team's data and workflows.

Get a Demo Talk to Sales