Artificial-intelligence-based orchestration
AI-based orchestration. In an embodiment, a recommendation engine is applied to a data pipeline representing company accounts. Engagement metric(s) are calculated based on activity data associated with the company accounts, and predictive model(s) are applied to the activity data and/or firmographic data associated with the company accounts to generate predictive output. A tactic recommendation model is applied to orchestration features, comprising the engagement metric(s) and predictive output, to generate recommended tactic(s). In addition, a contact recommendation model is applied to contact data to generate recommended contact(s). The recommended tactic(s) are combined with the recommend contact(s) to generate an orchestration, comprising recommended action(s), to be executed.
1 . A method comprising using at least one hardware processor to:
during a training phase:
train a profile model using machine learning with a first training dataset that associates contact features, comprising one or more job descriptors, with an indication of whether a sales opportunity was opened, won, or lost; and
train at least one of a plurality of modular plug-and-play tactic-specific models using machine learning with a second training dataset comprising labeled feature vectors, wherein each of the labeled feature vectors comprises a set of orchestration features labeled with an indication of whether or not an engagement resulted within a defined time period; and
during an operation phase:
receive one or more orchestration settings;
automatically apply a recommendation engine to a data pipeline representing a plurality of company accounts by
receiving activity data associated with the plurality of company accounts,
receiving firmographic data associated with the plurality of company accounts,
calculating one or more engagement metrics based on the activity data, wherein the one or more engagement metrics indicate a level of engagement by each of the plurality of company accounts,
applying a plurality of predictive models to one or both of the activity data and the firmographic data to generate a predictive output, wherein the plurality of predictive models comprises the profile model and an intent model,
wherein the profile model accepts, as input, the contact features for each of one or more contacts within contact data associated with each of one or more of the plurality of company accounts, and outputs a profile score for each of the one or more contacts, wherein the profile score for each of the one or more contacts represents a relevance of that contact to a sales opportunity, and wherein the profile scores that are output by the profile model are incorporated into the predictive output, and
wherein the intent model accepts, as input, activity data associated with each of one or more of the plurality of company accounts, and outputs an intent score, wherein the intent score represents a likelihood to engage in a sales opportunity, and wherein the intent scores that are output by the intent model are incorporated into the predictive output,
applying a tactic recommendation model to orchestration features, comprising the engagement metrics and the predictive output, to generate one or more recommended tactics, wherein the tactic recommendation model comprises the plurality of modular plug-and-play tactic-specific models, and wherein each of the plurality of modular plug-and-play tactic-specific models is trained to determine whether or not a respective tactic should be used for each of the plurality of company accounts based on the orchestration features associated with that company account,
applying a contact recommendation model to contact data to generate one or more recommended contacts, and
combining the one or more recommended tactics with the one or more recommended contacts to generate an orchestration comprising one or more recommended actions; and
execute the orchestration, wherein executing the orchestration comprises at least one of adding one of the one or more recommended contacts to a marketing campaign, adding one of the plurality of company accounts to a marketing campaign, purchasing a new contact, or initiating a marketing campaign to at least one of the one or more recommended contacts, one or more of the plurality of company accounts, or one or more purchased new contacts.
2 . The method of claim 1 , wherein the one or more orchestration settings comprise an orchestration goal.
3 . The method of claim 2 , wherein the orchestration goal defines one or more entry criteria, and wherein the method further comprises using the at least one hardware processor to select the plurality of company accounts as a subset of a plurality of available company accounts based on the one or more entry criteria.
4 . The method of claim 2 , wherein the orchestration goal defines one or more exit criteria, and wherein the method further comprises using the at least one hardware processor to, after executing the orchestration, measure a success of the orchestration based on company accounts that satisfy the one or more exit criteria.
5 . The method of claim 1 , wherein the one or more orchestration settings comprise a set of tactics, and wherein the tactic recommendation model is constrained to only incorporate tactics from the set of tactics into the one or more recommended tactics.
6 . The method of claim 1 , wherein the one or more orchestration settings comprise, for at least one tactic that is available to be recommended by the tactic recommendation model, an identifier of at least one marketing campaign to be associated with the at least one tactic.
7 . The method of claim 1 , wherein the one or more orchestration settings comprise an indication of whether or not the orchestration is to be executed automatically, wherein the method comprises using the at least one hardware processor to:
when the indication is that the orchestration is to be executed automatically, automatically execute the orchestration without user intervention; and,
when the indication is that the orchestration is not to be executed automatically,
request user approval of the orchestration, and
execute the orchestration only after the user approval has been received.
8 . The method of claim 1 , wherein the activity data comprise representations of online activities that have been mapped to the plurality of company accounts, and wherein calculating the one or more engagement metrics comprises weighting a first type of online activity higher than a second type of online activity.
9 . The method of claim 1 , wherein the orchestration features further comprise at least a subset of the firmographic data.
10 . The method of claim 1 , wherein combining the one or more recommended tactics with the one or more recommended contacts comprises converting a tactic to add a company account to a marketing campaign into an action to add one of the recommended contacts, which is already associated with one of the plurality of company accounts, to a marketing campaign identified in the orchestration settings.
11 . The method of claim 1 , wherein combining the one or more recommended tactics with the one or more recommended contacts comprises converting a tactic to acquire a new contact into an action to acquire one of the recommended contacts, which is not already associated with one of the plurality of company accounts.
12 . The method of claim 1 , wherein the contact recommendation model utilizes the profile model to predict a relevance of each contact, which is associated with the plurality of company accounts, to a sales opportunity, and generates the one or more recommended contacts based on the predicted relevance.
13 . The method of claim 12 , wherein the contact recommendation model further utilizes the profile model to predict a relevance of each of a plurality of contacts, which is not already associated with the plurality of company accounts, to a sales opportunity.
14 . The method of claim 1 , wherein the profile model comprises a random forest algorithm or a gradient-boosting algorithm.
15 . The method of claim 1 , wherein the intent model comprises a Bayes algorithm.
16 . The method of claim 1 , wherein the at least one of the plurality of modular plug-and-play tactic-specific models comprises a random forest algorithm or a gradient-boosting algorithm.
17 . A system comprising:
at least one hardware processor; and
software that is configured to, when executed by the at least one hardware processor,
during a training phase,
train a profile model using machine learning with a first training dataset that associates contact features, comprising one or more job descriptors, with an indication of whether a sales opportunity was opened, won, or lost, and
train at least one of a plurality of modular plug-and-play tactic-specific models using machine learning with a second training dataset comprising labeled feature vectors, wherein each of the labeled feature vectors comprises a set of orchestration features labeled with an indication of whether or not an engagement resulted within a defined time period, and
during an operation phase,
receive one or more orchestration settings,
automatically apply a recommendation engine to a data pipeline representing a plurality of company accounts by
receiving activity data associated with the plurality of company accounts,
receiving firmographic data associated with the plurality of company accounts,
calculating one or more engagement metrics based on the activity data, wherein the one or more engagement metrics indicate a level of engagement by each of the plurality of company accounts,
applying a plurality of predictive models to one or both of the activity data and the firmographic data to generate a predictive output, wherein the plurality of predictive models comprises the profile model and an intent model,
wherein the profile model accepts, as input, the contact features for each of one or more contacts within contact data associated with each of one or more of the plurality of company accounts, and outputs a profile score for each of the one or more contacts, wherein the profile score for each of the one or more contacts represents a relevance of that contact to a sales opportunity, and wherein the profile scores that are output by the profile model are incorporated into the predictive output, and
wherein the intent model accepts, as input, activity data associated with each of one or more of the plurality of company accounts, and outputs an intent score, wherein the intent score represents a likelihood to engage in a sales opportunity, and wherein the intent scores that are output by the intent model are incorporated into the predictive output,
applying a tactic recommendation model to orchestration features, comprising the engagement metrics and the predictive output, to generate one or more recommended tactics, wherein the tactic recommendation model comprises the plurality of modular plug-and-play tactic-specific models, and wherein each of the plurality of modular plug-and-play tactic-specific models is trained to determine whether or not a respective tactic should be used for each of the plurality of company accounts based on the orchestration features associated with that company account,
applying a contact recommendation model to contact data to generate one or more recommended contacts, and
combining the one or more recommended tactics with the one or more recommended contacts to generate an orchestration comprising one or more recommended actions, and
execute the orchestration, wherein executing the orchestration comprises at least one of adding one of the one or more recommended contacts to a marketing campaign, adding one of the plurality of company accounts to a marketing campaign, purchasing a new contact, or initiating a marketing campaign to at least one of the one or more recommended contacts, one or more of the plurality of company accounts, or one or more purchased new contacts.
18 . A non-transitory computer-readable medium having instructions stored therein, wherein the instructions, when executed by a processor, cause the processor to:
during a training phase:
train a profile model using machine learning with a first training dataset that associates contact features, comprising one or more job descriptors, with an indication of whether a sales opportunity was opened, won, or lost; and
train at least one of a plurality of modular plug-and-play tactic-specific models using machine learning with a second training dataset comprising labeled feature vectors, wherein each of the labeled feature vectors comprises a set of orchestration features labeled with an indication of whether or not an engagement resulted within a defined time period; and
during an operation phase:
receive one or more orchestration settings;
automatically apply a recommendation engine to a data pipeline representing a plurality of company accounts by
receiving activity data associated with the plurality of company accounts,
receiving firmographic data associated with the plurality of company accounts,
calculating one or more engagement metrics based on the activity data, wherein the one or more engagement metrics indicate a level of engagement by each of the plurality of company accounts,
applying a plurality of predictive models to one or both of the activity data and the firmographic data to generate a predictive output, wherein the plurality of predictive models comprises the profile model and an intent model,
wherein the profile model accepts, as input, the contact features for each of one or more contacts within contact data associated with each of one or more of the plurality of company accounts, and outputs a profile score for each of the one or more contacts, wherein the profile score for each of the one or more contacts represents a relevance of that contact to a sales opportunity, and wherein the profile scores that are output by the profile model are incorporated into the predictive output, and
wherein the intent model accepts, as input, activity data associated with each of one or more of the plurality of company accounts, and outputs an intent score, wherein the intent score represents a likelihood to engage in a sales opportunity, and wherein the intent scores that are output by the intent model are incorporated into the predictive output,
applying a tactic recommendation model to orchestration features, comprising the engagement metrics and the predictive output, to generate one or more recommended tactics, wherein the tactic recommendation model comprises the plurality of modular plug-and-play tactic-specific models, and wherein each of the plurality of modular plug-and-play tactic-specific models is trained to determine whether or not a respective tactic should be used for each of the plurality of company accounts based on the orchestration features associated with that company account,
applying a contact recommendation model to contact data to generate one or more recommended contacts, and
combining the one or more recommended tactics with the one or more recommended contacts to generate an orchestration comprising one or more recommended actions; and
execute the orchestration, wherein executing the orchestration comprises at least one of adding one of the one or more recommended contacts to a marketing campaign, adding one of the plurality of company accounts to a marketing campaign, purchasing a new contact, or initiating a marketing campaign to at least one of the one or more recommended contacts, one or more of the plurality of company accounts, or one or more purchased new contacts.