Frequently Asked Questions

The Connect platform is designed for anyone working with data, conducting research, or seeking to automate tasks and gain insights. With its powerful analytics, automation capabilities, and user-friendly interface, Connect enhances decision-making and streamlines processes, making it an invaluable asset across various industries and roles. Anyone who uses tools like MS Excel or Google Sheets, web search, and wrangles with data will get great benefits from using the platform.

The platform works on any commonly used web browser. Other than creating an account, it doesn’t require any other implementation steps. Users can start generating value right away.

For enterprise customers with more sophisticated integration needs, let’s talk.

Connect is LLM-agnostic. However, most customers let Connect select the underlying LLM.

We have a scalable RAG that is pipelined into our own proprietary knowledge graph and a vector database.

Connect allows users to upload files into the chat or to connected data sources even in the free version. The main advantages are in the underlying technology:

• Autonomous operation: Unlike competitors who rely on LLMs for planning and reasoning, Connect leverages its proprietary cognitive planner foundational technology. This technology mimics human-like planning and reasoning.
• Enhanced capabilities: Connect's technology allows for more autonomous operations, enabling the completion of highly complex tasks with greater accuracy and minimal human intervention.
• Proven performance: Connect has demonstrated a 10x improvement on the most complex tests (e.g., Gaia) for autonomous agents.
• Personalization technology: The platform gets to know the user’s role, preferences, and objectives to deliver a unique experience for every user
• Thought leadership: While the industry has focused on LLMs over the past two years, Connect has concentrated on cognitive planning technology, positioning itself ahead of the market.
• Theoretical limitations of LLMs: It is theoretically infeasible for LLMs to achieve real planning and reasoning. Connect's focus on cognitive planning technology addresses this gap effectively.

In simple terms, other solutions do not understand how the world actually works. They are just memory units that are really good at predicting the next word in a sentence. LLMs are unable to break a problem into various subtasks, plan the tasks required to solve the problem, analyze outputs on the fly and change direction if needed, nor understand the impact that its outputs or recommendations will create.

The bottom line is that for business users Connect provides the easiest and most straightforward way to create accurate and relevant outputs for questions involving highly complex analytics, even with very simple and high-level human directions.

When the user uploads a file, the platform automatically syncs the file or files. The syncing process includes the creation of metadata to understand the contents and enable easy finding of relevant data points to generate user outputs. The relationship between the files is deduced from the knowledge graph created by the connected sources.

Files, web search, various Python packages (e.g., yFinance), Gmail, Google Calendar, Outlook Email and Calendar,no-code connectors to various enterprise applications.

If there is conflicting data in the available data sources, the platform will prioritize the most recent data. The search is using our proprietary knowledge graph that also helps keep track of data relevancy and freshness.

Yes for some connectors, for others we have a periodic sync.

It will be consistent with some deterministic data analytics results, but depending on the use case, the generated text may not be consistent.

For PDFs, very high proficiency in about 30 of the most common global languages and strong proficiency in over 100 other languages.

Extracting information from pictures is currently not available.

The main interaction types are text-based, but Connect also has OCR capabilities to extract text from file formats like scanned PDFs.

For interactions from Connect to the user, the platform can also create a wide range of visualizations like charts or geospatial graphics, for example.

It will provide the same response as it would without the uploaded documents. If the user specifies that the answer should be found in the uploaded document, Connect will reply that the answer is not in the document. If Connect is working with ‘Web Search Enabled’ it will also try to find the answer using a web search.

Yes.

Connect has various data visualization features, including word clouds and all commonly used chart types. If Connect creates the data visualization, it will be able to produce text descriptions and provide insights related to the visualized data.

Graphs can currently only be interpreted using OCR, which extracts any text data. If the graph does not have any text, Connect will currently be unable to interpret it.

Parameters are selected based on context, available data, and user input. If the user requests specific parameters, those will be implemented. The user also has the option to let Connect determine the best parameters (e.g., variables visualized, chart type), in which case Connect will refer to best practices within the context of the conversation.

Yes. There are various ways to accomplish this, including a Saved Connecty (user-defined custom application) and Conversation Starters (saved prompt available at the push of a button).

Connect uses Apify to scrape websites and the resulting content is saved as plain text.

Every output generated by AI (not just Connect) should be validated by a professional human if it is to be used as a legal, financial, or medical  recommendation.

We do not identify PII specifically; all data is secured and encrypted.

Yes.

Connect does not own nor maintain any LLMs as it is an LLM-agnostic platform. If a customer uses publicly available LLMs, the owner and operator of the selected LLM is responsible for training and maintenance. If a customer uses a private/proprietary LLM, it is the customer’s responsibility to train and maintain.

Yes, but due to the significant engineering effort involved in such a deployment, it is only available for enterprise customers that are interested in a broad deployment across their organization.

Connect’s pricing and packaging is based on number of users and usage. Performance is the same for all subscription tiers, but the tiers differ in terms of the number of messages, volume of training data, connectors, and Shared Connects, etc.