Artificial Intelligence Customer Support

    Evaluating A.I. for Customer Support

    By Divya Susarla on September, 26 2017

    Stay up to date

    Back to main Blog
    Divya Susarla

    Business leaders are becoming more keen to evaluate A.I. solutions. Based on our conversations in the market and being an A.I. provider, here is a list of things to consider when evaluating A.I for customer support.

    1. Control Level of the A.I. system

    When deciding to employ an A.I. system, it is important to think through how control will be given to the A.I. versus the human agent. This will start with thinking through the different types of A.I. services to consider. Several options exist, in the form of platforms or integrations directly into the helpdesk services your team already uses. Weigh the options of taking on a tool that will require training your agents on a new system, versus ones that require no behavioral change from them. Think about where in your workflow and interactions with customers you would like to leverage A.I. Beyond an on and off switch, consider when the A.I. system can operate and when it is more appropriate for a human agent to take over.

    2. Audit Trail: How does the A.I. make decisions

    Once you have decided what type of service makes most sense for your agents and workflow, you will want to also focus on thinking through which type of machine learning models comprise part of the system. Different models have different implications on how the A.I. system makes decisions. You will need to give thought to your ticket volumes, since this is the data that will directly feed into these models. Simpler, retrieval based models will have a higher degree of traceability around the decisions being made and the drafts that your agents will see versus the more complex neural network models.

    3. Specialized A.I.

    With both the type of system and types of models under consideration, you will also want to understand the ability to create specialized A.I. that is specific to business cases within organization. An A.I. system that is trained for one specific business use case is more effective that solving across multiple use cases. Prioritize the business line that you are trying to bring efficiencies to and optimize for.

    4. Complete integration

    In connection to the specialized A.I. approach, give thought to what fully integrating an A.I. system could look like for your organization. Do you want the A.I. system to utilize data pulled from all the systems in your organization or are you looking for a plug and play solution by department? You will need to think about different departments and if they connect through different databases, and what crossover functionalities make the most sense.

    5. Speed of Deployment

    Now, with a more holistic picture of the A.I. services that you are considering, you will also need to evaluate timing of deployment. This comes down to evaluating the competitive landscape for your business and how quickly you want to move. Know that A.I. systems will have an upfront time to deploy that can vary based on the type of service (platform versus directly into your existing systems), the type of models being used, and if you are looking for a complete integration.

    Submit a Comment

    Stay up to date