Since our initial post about the Crescendo AI Assistant, we’ve had many conversations with our clients. These discussions gave us valuable insight into how teams actually want to use our Assistant in their day-to-day work.
At the same time, there has been a major shift in how modern AI systems are built. Agentic architectures now make it significantly easier to design and manage complex workflows.
As a result of these two developments, we’ve introduced major improvements to the Crescendo Assistant, both technically and functionally.
In this article, we’ll walk through the most important updates, show you how to get started, and share practical examples of how the assistant can be used.
What’s Changed
Before diving into setup and examples, here are the most significant improvements compared to the previous version:
🔒 No documents shared with external AI providers
Your documents are never sent to providers like OpenAI or Anthropic. Instead, context is built locally on your device. While the assistant may use small, relevant snippets to generate responses, full documents are never transmitted.
📂 Automatic access to your documents
By default, the assistant can use all documents you have access to on your device. There’s no need to manually select which files should be included. If a document is available to you, the assistant can use it when relevant.
🌐 Support for external knowledge sources
You can connect external sources to expand the assistant’s knowledge. For example, we include crescendoapp.com by default, allowing the assistant to answer product-related questions and act as a support agent. You can also connect your own website or other sources.
🧩 Configurable tools and actions
The assistant can use a set of tools to perform specific tasks. Built-in tools provide awareness of context (such as what you’re currently viewing), past actions, metadata, and more.
For white-label clients, it’s also possible to integrate custom tools, giving the assistant domain-specific capabilities tailored to your needs.
How to Enable the AI Assistant
To get started:
🔹 Create an OpenAI account at platform.openai.com
🔹 Create a new project
🔹 Share access to the project with us
We’ll handle the rest of the configuration.
Crescendo AI will use only your OpenAI project, which means you have full control over:
🔹 Costs
🔹 Prompts
🔹 Model selection
Once everything is set up, we can enable the assistant for your entire team or selected users.
Costs
Good news: we don’t charge anything extra for using the AI Assistant.
However, your organization will pay for the tokens consumed by the AI model. The total cost depends on usage, but in our experience, it remains very cost-effective, especially when using smaller (“Mini”) models.
For most typical use cases, highly advanced and expensive models are not necessary.
OpenAI also allows you to:
🔹 Set usage limits
🔹 Configure spending alerts
This makes cost control straightforward and predictable.
Using Other AI Providers
While we currently support OpenAI by default, it’s possible to use other providers.
The AI landscape is evolving quickly, and we’ve found that most of our clients already use or plan to use OpenAI at the company level. That said, if your team needs support for other models (such as Anthropic or Google), let us know and we’ll do our best to enable it quickly.
How the Assistant Works in Practice
The Crescendo AI Assistant is designed to work directly with the content you already have, browsing, searching, and reading files across your cases in a natural, conversational way.
Instead of manually navigating folders or opening documents, you can simply ask questions and the assistant will take care of the rest. Typically our users use the Assistant in three common ways:
🔹 Finding information across all your files
The assistant can search not only file names but also the content inside documents.
For example:
“What files mention the Johnson contract?”
It will scan all available documents and return the most relevant matches, even if the keyword appears deep inside a PDF.
🔹 Comparing information across documents
The assistant can analyze multiple documents and compare key details.
For example:
“Compare the 2024 and 2025 budget reports and highlight the main differences.”
The assistant will locate both documents, extract relevant data, and present a clear comparison of changes, trends, or discrepancies.
🔹 Writing based on your content
The assistant can generate content using information from your documents.
For example:
“Write a summary of the latest compliance policy for a client email.”
It will find the relevant document, understand its content, and produce a clear, well-structured draft tailored to your request.
More details and examples
The Crescendo Assistant includes a flexible and extensible set of capabilities that users can take advantage of. Read below to learn more.
🔹 Working with Context
The Crescendo Assistant doesn’t require you to explicitly navigate files, folders, or cases. It understands context, what you’re currently viewing, what you’ve recently opened, and how your content is structured, and uses that to interpret your requests naturally.
This means you don’t need to be precise with names or locations. You can refer to things indirectly, just like you would when speaking to a colleague.
🔹 Using your current context
The assistant is aware of what you’re currently looking at, whether it’s a case, folder, or document, and uses that as a starting point.
For example:
“Summarize this document”
“What does this page say about termination clauses?”
There’s no need to specify file names. The assistant understands what “this” refers to based on your current view.
🔹 Referring to recent activity
You can reference things you’ve just opened or worked on without repeating details.
For example:
“Compare this with the previous version”
“What changed in the last document I opened?”
The assistant keeps track of your recent context and uses it to resolve these references.
🔹 Navigating naturally across content
You can move across cases, folders, and documents using simple language, without knowing the exact structure.
For example:
“What’s in the Contracts folder?”
“Show me files in the Acme case”
“Open the latest budget report”
The assistant interprets your intent and finds the right content, even if you don’t provide exact names.
🔹 Searching and drilling into details
The assistant can locate relevant documents and immediately dive into the exact sections that matter.
For example:
“Find documents mentioning liability and show me the relevant pages”
“Where do we discuss the 2025 forecast?”
Instead of just returning files, it points you directly to the most relevant parts.
🔹 Understanding indirect references
You don’t need to be exact when describing what you’re looking for. The assistant can interpret vague or partial instructions.
For example:
“Check that contract we looked at yesterday”
“Summarize the report from the finance folder”
“Find the document about the Johnson deal”
It combines context, history, and content understanding to resolve these references accurately.
🔹 Multi-step workflows
The assistant naturally combines these capabilities to handle more complex tasks without requiring step by step instructions.
🔹 Finding and summarizing relevant content
“Find the latest compliance policy and summarize the key changes on the first few pages.”
The assistant identifies the right document, navigates to the relevant sections, and produces a clear summary.
🔹 Exploring what’s new
“What’s new in the Acme case this week? Give me a quick overview.”
It identifies recently added or updated documents and summarizes each one.
🔹 Cross-referencing information
“Which documents mention indemnification and also reference limitation of liability?”
The assistant searches across your content, connects related information, and returns precise results.
🔹 Navigating to the right place instantly
“Open the Q4 earnings report and go to the section about operating expenses.”
It finds the document and takes you directly to the relevant section.
🔹 Combining context and actions
“Sync the Finance case and then find budget documents from last quarter.”
The assistant ensures your content is up to date and immediately performs the requested search.
🔹 Browsing and drilling down
“Show me what’s in the Acme case, then check the Contracts folder and summarize anything mentioning renewal.”
It navigates through the structure, identifies relevant documents, and extracts key information.
