This article summarises all the releases of Sensai, including details on changes that have been introduced within each version, starting with the most recent Generally Available release.
Release 3.3 | 15th March 2021
- New! Multi-Sequenced Models - Sensai's Winning Opportunity model now supports the use of multiple time-sequenced data for training and predictions. Previously, only one time-sequenced entity could be used in conjunction with static Opportunity information to drive predictions. Now, Sensai can utilise multiple time-sequenced related Opportunity entities, such as Activities, Stage and Field History, as well as Observations, which in turn creates far deeper and high-performing models.
- Fixed Model Metrics - Changes have been made in this release to resolve issues encountered when using specific model metrics, as well as broader changes to how metrics are calculated to ensure that they are accurate.
The following metrics should be removed and re-added before the model is re-trained in order to prevent any errors: Precision Score, Recall Score, Balanced Accuracy Score, Cohen Kappa Score, F1 Score
Release 3.2.1 | 24th February 2021
- Fixed Salesforce Federated Login - An infrastructure issue was identified that was causing federated login to Salesforce orgs to fail, which has now been resolved in production.
- Fixed Record Searching - When searching for records within list views, if no records matched the entered search term the search box was being incorrectly hidden.
Release 3.2 | 13th January 2021
- New! Numerical Regression Models - Support has been added for Numerical Regression models on the Sensai platform. End-users can now use the new Numerical Regression model type in the Admin app to create and train models to predict the outcome of numerical variables. The Numerical Regression model can be used to predict deal amounts and churn rates and other custom numerical fields.
- New! Model Metrics - End-users can now choose which Metrics to use to score the performance of models. Different Metrics are available based on the type of model being used, with an appropriate default Metric assigned to each trained model.
- Improved Model List Views - The Models list view has been enhanced to show a subset of models based on the model type. This enables end-users to perform a like-for-like comparison of models using the same scoring Metric.
- Improved Column Names in Files - Improvements have been made to handle scenarios where special characters and spaces are used in column names. These improvements reduce the likelihood of model training failures and the requirement to change the format of columns prior to training.
- Fixed Model Column Exclusion - For model columns that cannot be edited the Actions button was incorrectly being made available to end-users in the Columns tab within the Model detail view.
- Fixed Cancelling File Uploads - When uploading a File it was possible for end-users to cancel out of the modal whilst the upload was still in progress.
- Fixed Reset User Passwords - For new users that have not verified their password, if the user is subsequently disabled without confirming their password, the 'Resend Password' option was incorrectly available.
Release 3.1 | 17th December 2020
- New! Support for Activity Data - Activities and other time sequenced data can now be used to train Salesforce optimised Winning Opportunity models. Activity data including Tasks and Events, as well as other sequenced data such as Opportunity Field History that are related to Opportunities, can now be utilised when creating and training Winning Opportunity models.
- Improved Admin App List Views - Enhancements have been made to list views in the admin app which include the introduction of pagination, record searching, and sorting. These enhancements have been made across all applicable list views and therefore make it easier for end-users to locate and access records across the admin app.
Release 3.0 | 24th November 2020
- New! Support for Opportunity History - Salesforce Opportunity History data can now be used alongside Opportunity data to train models in Sensai. This time sequenced data contains information on stage changes for an Opportunity, including Amount and CloseDate changes. A new Salesforce optimised model type is provided in this release that enables models to be created and trained using Opportunity History data.
- New! Build and Deploy Classification models - Classification models can now be built using the new 'Binary Classification' model type that is available in this release. This model type enables users to build models using data from any source to predict the outcome of categorical columns, for example, predict the propensity for a customer to churn, or predict the likelihood of converting a Lead.
- Improved Model Creation process - When creating models using one of the SuMo or Salesforce optimised model types, end-users no longer need to specify additional training parameters, therefore simplifying the creation process. Additional context specific help is now provided when creating models, information is provided for each model type that helps end-users identify the correct model type they require.
Release 2.4 | 4th November 2020
- New! Create Datasets from Schemas - You are now able to create Datasets directly from Schema records in the Sensai webapp. When creating Datasets in this way, the Schema is already pre-defined for users, and once created, the Dataset records can be accessed from the Schema.
Release 2.3 | 21st October 2020
- Improved Sensai for SuMo v1.21 Salesforce Package - This latest package contains enhancements that ensure that Predictions and Next Best Action recommendations can always be delivered when different types of Models (and Schemas) are deployed to environments.
- Improved Sensai for Salesforce v1.3 Salesforce Package - This latest package contains enhancements that ensure that Predictions can always be delivered when different types of Models (and Schemas) are deployed to environments.
- Improved Data File Handling - Support has been provided to allow Models to be trained using data files that contain spaces in file names or that contain mixed capitalisation of column headers.
Release 2.2.1 | 19th October 2020
- Fixed Opportunity Forecast Predictions - An issue was identified that was causing Forecast predictions to fail when specifically using a model that had been trained only on Opportunity data.
Release 2.2 | 1st October 2020
- New! Sensai for Salesforce v1.1 Salesforce Package - The Sensai for Salesforce package provides businesses who do not have SuMo for Salesforce installed the ability to access Opportunity Predictions on Opportunity records within Salesforce. The Opportunity Prediction Indicator provides a real-time prediction (0 to 100) of the likelihood of the Opportunity successfully closing, which is driven by your business's unique AI Model.
- Improved Sensai for SuMo v1.18 Salesforce Package - Enhancements have been made to the Sensai Opportunity component to handle scenarios where no-data is to be displayed.
- Improved Dynamic Data Import - The data import functionality has been improved so that customers can now import data from connected Salesforce environments based on the entities defined in Schema definitions. This is far more flexible and powerful, as multiple data Files from different related entities can be dynamically imported at the same time. The data import functionality also respects the permissions of the delegated Integration User, meaning that data Files can only be imported based on their entity access.
Release 2.1 | 21st September 2020
- Improved Sensai for SuMo v1.18 Salesforce Package - Minor display enhancements have been made to the Sensai for SuMo package that can be installed alongside the SuMo package in Salesforce orgs.
- Improved Data Volume Support - The prediction service has been enhanced to support larger data volumes, therefore allowing models to be trained that utilise larger datasets. These enhancements mean that more complex datasets that contain many rows and columns can be processed by the prediction service, which can lead to improved Forecast accuracy.
- Improved Model Creation Process - The model creation process has been further refined in this update, through the introduction of a new model status type and improved user experience when creating and working with model records.
- Improved Sort Order Columns - When a Schema record entity has been defined with a 'Sort Order' column, this column will no longer be listed and set as excluded on the model record detail view (unless the same column has been selected for training).
- Fixed Target Column Exclusion - There was an issue identified where a models 'Target' column exclusion reason was being incorrectly set.
- Fixed Model Detail Fields - There was an issue identified where the 'Target' field value of model records was not being displayed. The previously displayed 'Index' field was also removed as it is no longer required.
Release 2.0 | 1st September 2020
- New! Deal Predictions - Predictions can now be made from day-one on existing deals without the dependency on time-series data provided by SuMo. Realtime Deal Predictions can be delivered to Salesforce users on open Opportunities, as well as Forecast Deal Predictions to Salesforce and to other CRM systems. Deal Predictions unlocks the capabilities of the Sensai platform immediately to all customers regardless of their platform and data requirements, by utilising existing deal data which is modelled by Sensai's Machine Learning algorithms.
- New! Schemas & Datasets - The Sensai platform has been developed to provide more flexibility and support for different types of data that can be used to train models. This has been achieved by the introduction of Schemas and Datasets which enable customers to define data entities and relationships and then map these to specific data files. These new structures enable end-users to create and train models using different CRM and external data, thus allowing the Sensai platform to harness the full breadth of this data to provide better predictions.
- Improved Machine Learning Enhancements - Further enhancements have been made to Sensai's Machine Learning capabilities based on new research and development as well as through collaboration with implementation partners. These enhancements provide customers with more realistic and reliable predictions based on their unique data compositions.
- Improved Architecture Efficiencies - Sensai's serverless microservice architecture has been further optimised to handle large and more complex data volumes and structures. These efficiencies will improve performance and raise limits across the platform, which will in turn benefits customers as they further utilise the capabilities of Sensai.
Release 1.0 | 23rd October 2019
- New! Sensai Guided Selling - provides sales executives with a simple, yet immensely powerful component that guides sales users to perform the most impactful actions and behaviours whilst indicating the impact these behaviours have had and will have, on the likelihood of each individual opportunity closing.
The Opportunity Prediction Indicator is displayed within the component and provides a real-time prediction (0 to 100) of the likelihood of the Opportunity successfully closing. This prediction is driven by your business's unique AI Model. Any activities performed on Opportunities will trigger a dynamic update of the Opportunity Prediction and likelihood of successful closure.
To accompany the Prediction Indicator, a Past Actions indicator and timeline are also displayed within the component. These map the history of how the prediction has changed over time. This sparkline plots the individual Behaviours that have been performed previously on the Opportunity, including when they were performed and the prediction at that point in time.
The Next Best Actions element with the component provide sales users with a context-sensitive list of the most impactful and likely Behaviours that can be completed at that specific point of time for the Opportunity. These recommendations are derived by Sensai’s deep learning capabilities and, when completed, are intended to increase the likelihood of Opportunity closure.
Each recommendation that is displayed provides a name and description as to what the Behaviour is that can be performed at that point in time. The recommendations also include an Impact Indicator that, when revealed, displays the Prediction uplift that will occur after the Behaviour has been performed.
New! Sensai Forecast Predictions - provide businesses with a holistic view of the performance of their open Opportunities in their pipeline. Sales Leaders and Sales Operations can instantly assess the health of the pipeline using the prediction data from the Sensai platform.
The Sensai platform can integrate directly with existing BI and reporting tools, allowing the Opportunity data generated by Sensai’s unique AI algorithms to be surfaced without the need to use additional tools or services. Forecast predictions for all open opportunities can be generated on a frequent basis, allowing sales leaders and operations to benefit from the latest forecast predictions provided by the Sensai platform.
Sensai is able to generate Opportunity Close Indicators across all open Opportunities for specific environments - whether that be production or sandbox. These indicators represent the likelihood of Opportunities closing at that point in time, and this information can be used within existing forecasting reports and dashboards to provide an alternative perspective to the traditional Forecast.
New! Sensai Admin App - is an easy-to-use and intuitive cloud-based administration console provided for Sensai customers, allowing them to manage their Sensai users, data and integrations with 3rd party applications that they may also choose to leverage.
The Sensai console enables users to either manually upload or dynamically import their Opportunity and activity data into the platform from their environment, e.g. Salesforce. Providing the data is in a CSV format, the datasets that are created in the console can be immediately used to create Models tailored to your environments.
Unique AI Models can be created and tailored in the console, then trained and deployed to all of the environments that are linked to your Sensai account. Sensai enables businesses to create and train multiple Models and then deploy them to their environments - both production and sandbox.
Sensai’s environment deployment technology is truly flexible in that it allows different Models to be deployed to both production and sandbox environments at the same time, therefore allowing customers to test new Model configurations in a sandbox before deploying them to live production environments.
Within the Sensai console, users can view the Training Accuracy of the Models that have been created to understand how accurate the model is based on the dataset and training parameters that have been used. Users can also access graphs that display information as to how the Model’s accuracy and loss have changed during the training process.
- New! Sensai for SuMo v1.0 Salesforce Package - The Sensai for SuMo package can be installed in Salesforce orgs that have SuMo for Salesforce installed. The Sensai package provides Salesforce users with access to the Sensai Guided Selling features and data integration capabilities between Salesforce and the Sensai platform.