Expertise Agent Overview

Overview

Kantata's Expertise Agent helps teams proactively manage projects, deliver more consistently, uncover hidden insights, and boost productivity. With access to key project, resource, and financial features and data plus knowledge of Professional Services industry best practices, the Expertise Agent simplifies daily tasks and streamlines operations—allowing you to focus on delivering successful projects and exceptional client experiences.

expertise agent - while on task tracker page2.png

Note: To access the Expertise Agent, users—including Account Administrators—must be in a custom Access Group with the Expertise Agent permission in the Kantata AI Access Group set checked. Account Administrators do not have access to the Expertise Agent by default and must be added to a custom Access Group.

About Kantata's AI Architecture

Legacy Agents

Kantata's initial AI tools used a network of specialized agents with their own purpose and model. This framework looked like a tree—with the Agents in the user interface delegating tasks to “subagents” and subagents delegating to subagents below them.

Although initially designed to be faster and return higher quality responses, this static structure sometimes resulted in:

Slow responses, leading to timeouts and user frustration.
Inability for Agents to maintain context over long conversations, making it difficult to accomplish tasks.
Agents understanding only Kantata-specific language, requiring users to know and use Kantata terminology for successful prompts.
Agents not having access to all areas of Kantata, limiting the actions users could perform via Agents.

Expertise Agent

The Expertise Agent is built on a new framework that addresses all these points and more: Dynamic Context Engineering. Rather than maintaining a static tree of agents with fixed capabilities, this framework utilizes a split architecture:

An "Architect" Agent that analyzes user intent
A specialized "Worker" Agent that the Architect agent instantly assembles

This dynamic framework allows the Expertise Agent to respond to open ended queries and assemble the right context and tools on the fly.

Additionally, this framework has access to hundreds of Kantata API endpoints as well as Professional Services-specific domain knowledge, such as how to optimize a resource schedule for maximum margin and rules around revenue recognition. Generic AI tools are powerful but they fall short. The information stored in our framework ensures that every interaction with the Expertise Agent follows industry best practices.

Benefits

Our new framework results in the following capabilities and benefits:

The Expertise Agent taps into a store of domain-specific knowledge to produce smarter actions and more accurate responses.
No more prompt engineering—instead of having to find the right words to get the Agent to do what you want, simply ask the Expertise Agent in your own words and it identifies and builds the exact tools it needs to meet your requests.
The Expertise Agent is capable of scaling across hundreds of API endpoints and complex Professional Services domain rules without context fracturing—allowing users to do almost anything through the Agent while following best practices.

Kantata's Expertise Agent is the ideal AI partner for your team members, with the speed, precision, and accuracy required to help you amplify institutional knowledge and drive exceptional outcomes.

What Does the Expertise Agent Know and What Can It Do?

The Expertise Agent has knowledge and context related to Kantata and the Professional Services industry. The Agent's knowledge of the Kantata data model, how Kantata objects are related to one another, and the specific actions required to accomplish tasks allows it to help you complete actions in Kantata. The Agent also has general Professional Services knowledge and context that helps it handle project management and resource management tasks, follow industry best practices, and suggest relevant follow up actions.

The Expertise Agent has knowledge and context for the subject areas listed below, allowing it to answer questions and help you with end-to-end tasks.

Project management
Time and expense tracking
Resource planning
Invoicing and billing
Project financials and revenue
Communication (e.g. status reports and project posts)
Files (i.e. analyzing uploaded files)
Templates
Project snapshots and baselines
Pulse
Rate cards
Skills and roles
Custom fields
User settings and project team management
Holidays and workweeks
Access Groups

Note: We are continuously adding subject areas and capabilities to the Expertise Agent. For the latest information about what the Agent knows and can do, try asking the Agent to "show me what you can do".

Here are just some examples of the tasks the Agent can help you with:

Project delivery tasks

List the top projects that need your attention
Summarize key risks and issues from health reports and activity feeds
Identify what you should prioritize for the day based on your assignments, deadlines, and dependencies
Determine which projects are over-budget and why and suggest mitigation plans
Identify projects where burn rate is outpacing the timeline and why
Provide a health check on a project and recommendations to optimize margin, timeline, and client sentiment
List project changes since a specified date
Review progress for the financial period close checklist
Create a project plan from an SOW file
Extract and summarize information from files, including meeting transcripts
Search project data and files to answer your questions
Collect and organize data in formats such as lists and tables
Write content such as emails and status reports
Create and update project data for you
Deliver insights about the project

Resourcing tasks

Assemble the ideal team for projects based on skills, availability, and financial impact
Identify which roles are most over-booked across active projects and determine if that poses delivery risks
Locate team members with availability over a specified period and recommend assignments
Determine roles you may need to hire for and when, based on future capacity
Perform bulk actions not available in the UI, such as creating or updating assignments or allocations for all resources
Surface risks around budgets, scheduling, and staffing
Clean up resourcing data as work gets confirmed, shifted, or completed
Align resourcing data as team members join or leave the project, dates shift, or work gets completed
Create and update data for you, such as creating allocations, distributing hours, assigning resources to tasks, and more

Billing tasks

Draft an invoice with approved time and expenses
List unpaid invoices past their payment schedule date
Create all fixed fee invoices for a specific month
Identify why forecasts differ from actuals
Confirm data is up-to-date before submitting revenue for recognition
Create invoices with invoiceable billing milestones

Time tasks

Review your task assignments, scheduled hours, and task activity to create a timesheet for a week
Identify who hasn't submitted their timesheet for the previous week
Identify outliers or mistakes in timesheets (e.g. time entries over 10 hours in a single day)

Expense tasks

List all unsubmitted expense reports across your projects
Create an expense report with all submitted expenses in a project and summarize submission notes

How to Prepare Projects for the Expertise Agent

For the best results, the following information should be available in projects:

A concise project description that clearly outlines the project's scope, identifies the clients, and details the project objectives.
Task Tracker is up to date and reflects the actual status of the project.
Relevant documents and deliverables have been uploaded to the project, to give the Agent more context about the project.
Meeting transcripts have been uploaded to the project, to help the Agent better understand the nuances of the project.

Using the Expertise Agent

Note: You must have the Access Group permission to use the Expertise Agent.

Launching the Expertise Agent and Having Conversations

To launch the Expertise Agent, select the Agent (“ ”) icon in the app bar. The Expertise Agent panel opens to the right of the page.

To start a conversation, enter a question or request. You can also select a suggested prompt. The list of suggested prompts changes based on the page you're currently on.

Tip: Ask the Agent to “show me what you can do” for detailed information about what actions it can take related to the current page. The Agent's response to this prompt changes based on the page you're currently on.

Once you've sent a message, the Agent may take a few moments to reply. You may minimize the modal while the Agent is working and your conversation will be retained. If you close the modal while the Agent is working, your request will be cancelled.

The Agent is aware of the project and page you're currently looking at. You can say “this project” or “this page” and the Agent will understand what you mean.

Before creating or updating data, the Agent will ask for confirmation, allowing you to review and verify changes first. Select the Confirmation Mode (“ ”) icon to change the confirmation mode: Ask to confirm or Auto mode.

The Agent always has access to the latest project data and activity, excluding uploaded files. The Agent is updated with data from uploaded files nightly. You can select the Refresh (“ ”) icon to the upper-right to manually update the Agent's data related to uploaded files.

If custom agents have been created on your account, the Expertise Agent may automatically use a custom agent to process your request. To call on a custom agent directly, follow the steps below.

Type / into the chat. The commands list opens.
Select the /agent option.
Select a custom agent.
A tag appears above the text box, indicating that a specific custom agent will be used to process your request.
Enter a prompt.

Managing Chats

Chat threads allow you to preserve context for ongoing conversations or start new sessions for new tasks. Your chat history is retained, letting you revisit and continue past conversations as well as close and reopen the Agent without losing context.

Select the Open Sidebar (“ ”) icon to expand the sidebar and access chats and options.

To start a new chat, select New Chat in the left sidebar. By default, the chat's name is based on your first message.

To see all your chats, select Chats in the left sidebar. Chats are sorted by most recent activity and only the last 100 chats are displayed.

To search chats, select Search in the left sidebar, then enter a keyword. Chat names and individual messages are searched.

To rename a chat, select Chats in the left sidebar, select the More (“ ”) icon for a chat, and select Rename. Type a new name, then click Enter on your keyboard or click outside of the input field.

To delete a chat, select Chats in the left sidebar, select the More (“ ”) icon for a chat, and select Delete. Select Delete in the confirmation modal that appears.

Note: Chats are automatically deleted after 365 days of inactivity.

Interacting with the Agent Panel

The Expertise Agent panel stays open as you navigate across tabs in certain areas—such as the project workspace or Resource Center—allowing you to converse with the Agent and see data on the page simultaneously.

Select the Open Sidebar (“ ”) icon to the upper-left of the Agent panel to expand the sidebar, see a list of your chats, and use search.

Select the Close (“ ”) icon to the upper-right to close the Agent panel. If you close the panel while the Agent is working, your request will be cancelled, but the rest of your conversation will be retained.

To resize the Agent panel, drag the left edge to the left or right until you reach your desired size.

You can Minimize (“ ”) the Agent panel to the lower-right of the page. When you minimize the panel while the Agent is working, the Agent will continue processing your request and your conversation will be retained.

When the Agent panel is minimized, you can enter a new message to resume a conversation. The panel will automatically expand when you send the message.

Select the Expand (“ ”) icon to reopen the Agent panel and access your conversations.

Interacting with the Agent's Responses

You can rate the Agent's responses and share feedback via the Good response (“ ”) and Bad response (“ ”) buttons. This feedback goes directly to the Kantata team and helps us improve the Agent.

Select the Copy (“ ”) icon to quickly copy a response to your clipboard.