AI Vs Predictive Models for CRM
- Team @AppRahas
- Feb 20, 2017
- 3 min read
Many vendors(Salesforce.com, Microsoft) have started rolling out Artificial Intelligence modules as part of their CRM (Customer Relationship Management in Sales and Contact Center) offerings, they are prioritizing AI features in their road map. Customers are excited about these offerings would improve their current CRM applications. How effective will they be?
Let's consider broader areas where intelligence can be embedded
a) Automation of repetitive tasks; examples include appointment reminders, automatic replies for sales proposals etc.
b) Deflection of contact center calls with self service help; virtual agents that have access to knowledge repository will provide self help answers
c) Intelligence in Sales Process or Service Process; Examples include Lead Scoring, Service Request Assignment etc.
There are multiple approaches to providing intelligence in Marketing, Sales and Support processes.
1) AI Capabilities embedded into the application by vendors (Chat Bots, Virtual Assistants etc). We can also refer to these as black box AI solutions
2) Highly Customized Predictive Models and rules engines tailored towards customers business (Predictive Analytical models)
Platform based AI modules implement the generic product/technology (black box) approach to improve productivity. It can solve automation problems with ease; however it will need training models (if vendors allow it) to incorporate any customized business context. We recommend customers start enabling these platform features to utilize the new capabilities and improve productivity. For these AI modules to be effective the data should be made available to CRM Platforms.
A highly customized intelligence model may not be easily achieved through platform based AI offerings, however common use cases applicable across customers are already implemented by vendor and customers can take advantage of them.
Highly customized predictive models understand business context and are highly tailored to customers needs. They will take time to build and needs both data domain knowledge and technology stack expertise. The learning curve will be higher, however there are use cases that benefit from these. They can allow a company to gain a competitive edge. Implementation of customized AI models will be need expertise and will have higher time to market, more maintenance, so not every customer may see a need for these. Customers should follow their existing business case to justify such an investment.
Lets look at some examples below and determine the better approach to solving these problems.
Example 1: Consider a Solar Energy company that has a new lead. CRM platform with AI can see that the neighbors are already solar system users and automatically score the new lead higher. This scoring will allow the Sales Agent to pick up the lead before other leads. A generic AI can solve this problem.
Example 2: Consider an energy retailer selling energy contracts. A predictive model that has access to past energy consumption and current energy market trends can score the lead higher than an generic AI. For this specific example, a customized model works better.
Example 3: Virtual agents assisting end customers to FAQs and routing to appropriate agents. A generic AI Chat bot can provide this functionality. A successful virtual agent interaction depends on availability of up-to date and well maintained knowledge repository
Conclusion
As you can see there are scenarios that benefit from a generic AI vs a customized predictive model.
As a starting point; customers should start evaluating vendor provided AI modules and roll it out to a specific set of users to understand the efficacy of the AI modules for their business process. In parallel based on the need and anticipated business improvement, they should build their specific models/intelligence.
Customers should not expect a vendor roll-out of AI modules to solve all their needs, they are just a starting point to solving common problems and can provide a framework to extend for their customized business specific functionality.