In the evolving landscape of artificial intelligence, Large Action Models (LAMs) offer a new approach compared to Large Language Models (LLMs) like GPT-4. While LLMs have advanced text generation and natural language understanding, LAMs focus on performing actions within software environments. This makes them suitable for tasks requiring real-time, action-based assistance, particularly in complex software like Bentley Products.

I propose prioritizing the development and integration of LAMs within Bentley products. These models, trained on workflows using screen recordings from YouTube and Bentley Learn, can help users tackle complex design tasks more efficiently.

Key Advantages:

• Real-Time Assistance: LAMs can guide users through complex processes and suggest optimizations, reducing the need for specialized knowledge.

• Efficient Workflow: They can automate tasks, offer suggestions, and assist with troubleshooting, making workflows more efficient.

• Learning and Onboarding: LAMs can help new users learn faster by providing interactive, contextual assistance.

• Adaptation to Changes: They can adjust to changes in design parameters or project requirements, keeping the design process flexible.

• Customized Assistance: LAMs can provide different levels of help based on the user's skill level, improving the experience for all users.

Proposed Implementation Features:

• Diverse Training Data: Train LAMs using a wide range of scenarios from YouTube tutorials and Bentley Learn sessions.

• Interactive Feedback: Allow users to ask for suggestions or help, with LAMs offering relevant advice.

• Integration with Tools: LAMs should enhance existing tools in Bentley Products without replacing them.

• Continuous Improvement: Update LAMs regularly with new data and feedback to keep them effective.