Instruct 2.5

Instruct 2.5

06/01/2026
Instruct is a General AI agent that connects to your apps and executes tasks on your behalf. Just describe the work in plain English, and watch it get done in real-time. Enterprise-grade security, seamless integrations, and reliable automation at scale.
instruct.ai

Instruct 2.5: The Autonomous AI Generalist for Your App Stack

Instruct 2.5 is a next-generation autonomous agent platform that officially launched on December 4, 2025, following its achievement as the #1 Product of the Day on Product Hunt. By shifting the paradigm from static bot building to dynamic live execution, it effectively eliminates the “build-run-break” cycle that has historically plagued traditional robotic process automation (RPA). Instead of requiring complex logic trees or brittle scripts, Instruct 2.5 utilizes a generalist AI agent that navigates application interfaces—including Salesforce, Slack, Gmail, and HubSpot—exactly like a human employee.

Built for transparency and adaptability, the platform prioritizes user observation and real-time correction. It serves as a “digital employee” capable of understanding high-level business goals and executing them across disconnected software tools. This architectural shift ensures that automations remain resilient to UI changes and evolving business rules, as the agent can interpret the screen and adjust its behavior rather than relying on fixed coordinates or selectors.

Key Features

  • Watch It Work (Live Execution Hub): Describe a task in plain English and watch the agent navigate through your business software in real-time. This provides immediate transparency and builds trust by showing exactly how the AI interacts with your data.
  • Iterate Live Correction: If the agent encounters a new scenario or an error, you can provide corrective feedback via chat or upload relevant files to guide the outcome instantly. This interactive coaching transforms the agent into a self-improving tutor for your specific workflows.
  • One-Click Save as Workflow: Once a task is successfully completed under your guidance, you can convert that run into a permanent, recurring automation with a single command. This effectively turns “one-off” work into an automated process without manual coding.
  • Cross-App Navigation & Execution: Unlike isolated integrations, Instruct 2.5 pulls context from one app (e.g., Google Drive) to perform work in another (e.g., Salesforce), allowing for seamless multi-app coordination.
  • Dynamic Task Handling: The agent is designed to handle non-linear tasks—such as filtering job candidates in Ashby or drafting personalized sales follow-ups based on company news—that typically break traditional, rule-based bots.
  • Secure Infrastructure & Context Mapping: Connects securely to your tools using enterprise-grade encryption and only pulls the necessary data context required to execute the specific task requested.
  • Human-in-the-Loop Orchestration: The system is built around the concept of “authoring-by-tutoring,” where the human oversees the execution and provides “hints” that the agent internalizes for future autonomy.

How It Works

The Instruct 2.5 workflow begins with a natural language instruction, such as “Find all invoices in my Gmail from this month and add the totals to my ‘Expenses’ Google Sheet.” The generalist agent immediately begins navigating the browser, accessing your Gmail, identifying relevant attachments, and extracting data. You watch this process live in a dedicated execution window.

If the agent misinterprets a document or skips a specific vendor, you can interrupt via the “Iterate Live” chat to say, “Include the VAT in the total for these specific invoices.” The agent adjusts its behavior on the spot. After the run is finished and verified, you simply click “Save as Workflow,” and the system creates a recurring automation that will perform the same task every month or on a custom schedule. This “teach once, run forever” approach removes the friction of traditional bot development.

Use Cases

  • Recruiting & Talent Filtering: Automate the processing of hundreds of inbound applications in tools like Ashby by instructing the agent to filter for specific experience (e.g., “4+ years of Python”) and move candidates to the “Phone Screen” shortlist.
  • Sales Intelligence & Outreach: Monitor HubSpot for inactive leads, research their latest company news via web search, and draft personalized follow-up emails saved as drafts for human review.
  • Finance & Operations Automation: Extract data from various invoice PDFs received in email and centralize the vendor, date, and amount into a Google Sheet while uploading the source files to a specific Drive folder.
  • Academic & Knowledge Management: Read lecture notes from a shared Drive, generate a 10-question revision quiz based on key themes, and email it to a study group automatically.
  • Customer Support Triage: Navigate through support tickets to categorize them based on urgency and draft initial responses based on internal knowledge base articles.

Pros and Cons

  • Pros: Replaces brittle code-based automations with resilient, UI-aware AI; Eliminates the technical barrier to entry for complex cross-app workflows; Provides unprecedented transparency via live execution monitoring.
  • Cons: Execution speed is currently limited by the time it takes for a UI to load and for the “cursor” to navigate, making it slower than pure API-based integrations; Performance reliability can vary depending on the stability of third-party web interfaces.

Pricing

  • Starter Plan: $0/month. A free tier designed for individual experimentation, supporting basic task execution and initial AI interaction testing.
  • Professional Plan: Custom monthly pricing (typically starting around $99/user/month in similar 2026 SaaS markets). Unlocks unlimited workflow executions, priority processing, and advanced cross-app coordination.
  • Enterprise/Team Plans: Quote-based pricing. Includes dedicated support, high-volume credit pools, enterprise security compliance (GDPR/SOC2), and white-glove onboarding for department-wide automation.

How Does It Compare?

  • Adept: While Adept pioneered the “large action model” space, Instruct 2.5 focuses more on the “Live Iteration” and “Save as Workflow” loop, making it easier for non-technical users to build reliable processes without engineering oversight.
  • UiPath (Autopilot): UiPath is the leader in traditional RPA. Instruct 2.5 is more “agentic” and less “script-based,” meaning it doesn’t break when a button moves on the screen, unlike UiPath’s legacy selectors.
  • MultiOn: MultiOn is excellent for general web-browsing tasks. Instruct 2.5 is more specialized for the “enterprise app stack” (Salesforce, Gmail, Slack), focusing on professional data entry and business intelligence workflows.
  • Zapier Central: Zapier Central connects apps via APIs and triggers. Instruct 2.5 complements this by handling the “last mile” of UI-based tasks that don’t have public APIs or require visual reasoning to complete.
  • Manus (acquired by Meta): A rising competitor in 2026 that handles complex market research and coding. Instruct AI maintains a lead in the “live monitoring” and manual “coaching” aspect of business process automation.

Final Thoughts

Instruct 2.5 represents the realization of the “digital employee” vision, where the barrier between a human describing a task and an AI performing it is virtually erased. By introducing “Watch It Work” and “Iterate Live” features, the platform effectively solves the trust problem that has hindered autonomous agent adoption in corporate environments. While it may not yet match the raw throughput of API-to-API integrations, its ability to navigate the messy, non-linear world of human software interfaces makes it an essential tool for teams looking to reclaim thousands of hours spent on repetitive manual execution in 2026.

Instruct is a General AI agent that connects to your apps and executes tasks on your behalf. Just describe the work in plain English, and watch it get done in real-time. Enterprise-grade security, seamless integrations, and reliable automation at scale.
instruct.ai