The LLM-powered ALICE Agent helps you analyze your schedule via a chat interface, allowing you to quickly discover insights and make informed decisions without needing to review a Gantt chart.

Use the Agent to:

• Discover insights in the schedule that are hard to see when you review the data line by line.
• Access key schedule details without reviewing the entire Gantt chart.
• Ask questions and get recommendations to improve the schedule.
• Understand what influences duration, the critical path, and more.
• Understand cascading effects due to changes, e.g. in delay analysis.
• Find soft/preferential logic to explore alternative sequencing.
• Quickly gather metrics to share with stakeholders. and create summaries for different audiences, such as executives or planners.
• Create customized tables about data that’s important for you and export them to Excel.
• Compare schedule updates to see what changed and where.
• Ask questions in your preferred language (the agent will respond accordingly)

How to Use It:

1. Go to the Analyze page and click the Insights Agent button in the lower right corner.
2. You can ask questions about your current schedule or compare two schedules if you enable the comparison mode.
3. The chat will offer prompts to help you get started, or you can type your own question.
   • You can always find these suggestions under the lightbulb icon in the lower left corner.

Chat Features and Controls:

• Above each answer, there is a “Show reasoning” option where you can view the sources and a brief explanation of how the answer was generated.
• Each blue text in an answer is a hyperlink and will lead you to the selected item when you are in the Gantt view
• When there are too many items to display (tens or hundreds of tasks), the Agent will show up to 20 items. You can download the full answer as an Excel (file .xlsx)  using the action below the answer.
• Below each of Insights Agent’s answers, you’ll see these actions:
  • Copy text – Copies the shown answer and reasoning
  • Download source data (.xlsx) – Download one Excel file (.xlsx) with the full data behind the answer. The workbook has one sheet per query.
  • Good response – Let us know if this answer worked well.
  • Bad response – Opens a feedback form where you can share what didn’t work. You can also submit quick feedback without details. Your feedback helps us improve future answers.
• Use the menu (…) in the top right of the chat to:
  • Clear Chat History – Erase previous chat history to start fresh
  • Knowledge base – Opens this help article
• Use the Stop button to cancel an input you already submitted. 
  • This button appears where the submit button used to be and is only visible while the agent is actively generating a response. 
  • Stop is useful if the user changes their mind after submitting a question or if the agent is taking too long to respond and the user wants to start over.
• Use the chat window controls to:
  • Minimize (X) – Click X to minimize the chat window. The agent keeps running, so you can continue working while it answers.
  • Resize – Drag the left edge of the chat window to make it wider. This helps you see large tables better.
     

How it works:

The Insight Agent is driven by a large language model (LLM) designed to analyze your schedule and deliver clear, traceable responses. 

• Your data stays private. It is never used to train the model
• Every answer includes its source and reasoning, so you can follow the logic behind it
• It only reads the selected schedule and it does not access other customers’ projects or search the internet.

What the Agent Can Analyze:

Project Level Metrics: 

• Duration, cost, workdays, start/end dates, resource cost
• Schedule name, schedule number, schedule type (Imported, ALICE or CPM schedule) and data date
• Direct cost, indirect cost and idle cost
• Average crew, equipment and material utilization, and peak crew and equipment units
• Whether the schedule uses Resource Flow Criticality
• Calendars and workweeks, overtime intervals.
• Work breakdown structure (WBS) and hierarchy

Activity Data (task, milestone, level of effort): 

• Cost, status, constraints, identification of critical/longest path, and WBS grouping
• Planned and actual start/end dates
• Original, actual, remaining and at-completion durations
• Total float and free float, in hours and working days
• Budgeted Labor Units: planned, actual and remaining
• Duration formulas
• WBS summary tasks

Task Relationships: 

• Logic types between tasks (e.g., finish-to-start, start-to-start) with reasons for dependencies, lag time, and additional gaps
• Resource-driven tasks, with the name of the driving crew, equipment, material or crane
• Hard logic vs soft logic. Uses name similarity (above 80%) for preferential logic identification.
   

Resources:

• Crews: number, cost, people per crew, calendar, utilization, idle time and cost, task assignment
• Materials: name, quantity assigned, quantity used, type, cost per unit, utilization, task assignment
• Equipment: name, quantity, hourly cost, utilization, idle time and cost, task assignment
• Daily cost and usage per resource, and total persons on site per day
• Unused resources and the reason why they were not used
• Budgeted Labor Units
• Custom Properties assigned to tasks

Analytics: 

• Schedule health metrics
• Any metric included in the Analytics tab
• Historic tracking of crew usage, cost, and idle time
• Equipment usage and costs
• Material consumption and cost, reusable/consumable
• Peak crew utilization

Optimization outcomes:

• Targeted optimization results, including creating a list of accelerated tasks.
• Original and accelerated duration per task, and the original, target and new finish date.

Comparison view (two schedules):

• Tasks added to, removed from, or remaining on the critical or longest path
• Tasks delayed or ahead in the primary schedule compared to the secondary schedule
• Any metric, resource or cost side by side

Prompting Best Practices

Specific questions will produce better results than vague questions. For example, “What is the Total Float for the Deck Pour task in the Construction WBS?” will produce a better result than “What is the Float of the deck pour task?”.

Ask one thing at a time. Do not ask the agent to do many analysis in one prompt. Use separate prompts instead. The agent remembers the recent messages in the chat, so you can build on its previous answers.

Below are some good examples.

Single Schedule questions: 

• How many Critical Tasks are there in the current schedule?
  • The same applies to Longest Path
• What are the critical resources in the current schedule?
  • The same applies to driving resources
• What tasks are driven by resources?
• What are the resource-driven tasks that have a 1 to 5-day time gap?
• What is the total count of FS, SS, FF, and SF relationships being used?
• What is the total cost breakdown by crew?
• Why didn't ALICE use all available Carpenter Crews?
• How can I improve milestone X completion?
• List me the tasks with lags.
• Give me a narrative on what's driving Milestone X.

When comparing two schedules:

• Comparison - when comparing two schedules
• What is the change in total duration?
• What is the change in Critical Tasks? - The same applies to Longest Path.
• Create a table of delta metrics for date, duration, cost, and resource utilization.
• Summarize the main differences between the schedules compared.
• What did ALICE resequence compared to the secondary schedule?
• What changed between the two schedule updates?
• Why did Milestone X move?
• Which tasks moved because Task X was delayed?

Known Limitations:

The insights agent is not currently trained to explain ALICE features or how to use them. Please refer to our help center or contact your Customer Success Manager for assistance.

LLM-generated answers may not always be accurate. We recommend double-checking results, especially for critical decisions.

FAQs:

Is my data safe and secure?

Yes. Your data is never used to train the AI model and remains private.

Do I need to worry about hallucinations with this AI-generated report?

Like all large language models, the agent may occasionally provide incorrect information. Please review answers carefully.

Can the Insights Agent make changes to my schedule?

No. The Insights Agent only analyzes your schedule. It does not make changes. To change your schedule, go to the Plan page.

Can it compare more than two schedules/solutions?

No, Insights Agent can compare two at a time.

Can I upload documents, like a BoQ, to the Insights Agent?

Currently uploading documents isn’t supported, but it’s coming in the future.