Our Technology

Meeting Selection Algorithm

AI: Natural Language Processing & Understanding (NLP & NLU)

While the CalendarHero web app and browser plug-ins are super easy to use, our power users love using CalendarHero inside their favourite chat applications like Microsoft Teams and Slack. To understand our users, we built an in-house NLP stack, to achieve an unmatched domain accuracy for intent and entity recognition.

• Pre-trained entity and intent classification

• 95%+ accuracy intent classification

• Deep learning for language modelling

• Semantic and syntactic-driven recognition

• Per-user & per-organization continuous learning

• Auto-language-detection and translation

• Conversational context-aware

• Intent engine is trained with 12,000 pieces of utterances (ie. sentences)

• Robust “human-input” date parsing

• Slang, emoji, and swearing detection

Contextual Conversation Flow

AI: Machine Learning Models

Machine Learning helps CalendarHero predict 50 workday patterns such as working hours and days, meeting types, and time and location preferences (for your afternoon coffee or your sales calls). It also keeps tabs on your connections and your strength to them, so it always knows which “John” you would like to meet with or get some insights on.

Observing user intent allows us to limit product complexity and remove the need for any required configuration at the time of registration. All the learning are done in under 10 seconds, allowing the user to start using the assistant the moment they finish the onboarding.

• Non-parametric learning; No assumptions made on user behaviour, allowing AI to learn without bias

• Unsupervised learning; drawing inferences from datasets without human guidance

• Continuous adaptive learning

• Contact/User disambiguation

• Predictive and intelligent knowledge tree traversal (for FAQ automation)

People Insights

Being prepared for a meeting is the difference between a successful meeting and a bad one. Knowing who you are about to meet is also part of your meeting preparation when you don’t know the person, or you want to be able to speak to them effectively.

• A recursive algorithm ensures that all of your attendee’s information is found regardless of the email that it is attached to or where that information resides

• Only requires the attendee’s email address

• Searches several structured user databases for their work history, social media accounts, location, tweets, blog posts, and more

• Searches through all of your authorized integrations and returns relevant information on that person, including a link back to that 3rd party software to see more information (eg. the CRM deal that that person is associated with)

CalendarHero’s exclusive People Insights gives you 2-3 pages of relevant information on your meeting attendee(s) right before the meeting or anytime that you ask.

Intelligent Contact Merge

Knowing who to invite to a meeting is crucial as you have the option for CalendarHero to reach out and invite them. Which Joe do you want to meet with? Should we use Joe’s personal or work email? Which work email?

Our contact merge algorithm runs continuously on about 20+ server instances, synchronizing contact information from all of our user’s authorized data sources, including Google G-Suite and Microsoft Office365, as well as HubSpot, Greenhouse, and Zendesk.

The end result is that our user’s have one contact record for each of their known contacts from all of their business applications.

Architecture

Our powerful and flexible architecture allows us to add integrations rapidly, which provides the data, we use in our Machine Learning models, and allows us to build automation on top of these integration. This architecture is actually our secret-sauce.