Accelerating the Adoption of AI in CRM

Introduction

Artificial Intelligence has been democratizing since the early 2010s, but the real boost for the public was the rise of generative AI models like GPT, GAN, VAE, and Transformers in the early 2020s. Since then, companies, large and small, have also experienced pressure to adopt AI in their entire business processes to stay competitive.

Many enterprises are struggling with this adoption and are looking for help to accelerate. This article will double-click on a specific domain within the value chain of any company, customer relationship management (CRM). It will discuss how companies can adopt COTS solutions to accelerate the adoption of AI in CRM. At the same time, any company must consider the challenge of fast AI capability innovation, compliance, data security, and data governance.

Note: ‘Company’ refers to the entity experiencing the challenge, while ‘customer’ refers to the customers with whom they interact.

Company Challenges

With the rise of AI, the leadership of any company has change management-related questions they need to address rather soon:

• How can I adopt AI and how do I get it to work for my business?
• How do I deal with data security, compliance, and legislation?
• Where do I find new personnel, what kind of profiles do I need?
• Is my current landscape ready for the AI revolution?
• What is a sustainable AI technology and solution vendor(s)?
• What is the impact on my workforce (people)?

Let us zoom in on the domain of interaction and relationship management with the outside world of a company: CRM. Generally, CRM covers the following domains:

• Marketing - the company is managing the brand, creating product awareness in the market, and trying to convert prospects into customers
• Sales - the company is managing its sales process using lead, opportunity, and customer (account) management
• Service - customers looking for solutions and managing their service requests through various channels
• Commerce - customers exploring and buying products online

In all of these subdomains of CRM, the company is interacting with (potential) customers and partners, while collecting data about these customers, including sensitive information. Every touchpoint through every channel is an opportunity to strengthen the relationship to increase the propensity to do or extend the business with that customer. Customer data collection makes a company accountable for protecting the privacy of the customer subject to privacy laws like GDPR in Europe and CCPA in the US. Adopting AI increases the challenge of staying compliant. Before diving deeper into the core aspect of AI (data), let us look at a summary of AI use cases in CRM.

Use Cases

In each of these domains, many AI-related use cases are common for any company in any industry:

• Sales and Marketing
  • Using predictive models for Lead scoring and Qualification based on historical data and the probability of converting a lead into an opportunity.
  • Using prompt templates grounded with CRM data using merge fields with generative technologies to quickly create call reports, quotes, contracts, and interaction summaries
  • Customer Segmentation, using clustering and classification models
  • Using neural networks or supported vector machines to predict customer churn (the likelihood that a customer stops using a product or service)
  • Use Time Series and Machine Learning to forecast Sales and Demand
• Commerce
  • Analysis of purchase patterns to identify which products are frequently bought together in Basket Analysis and Product Recommendations
  • Use collaborative filtering and machine learning models to advise on the Next Best Action in any given situation of customer interaction
  • Use NLP capabilities for Sentiment Analysis to identify the sentiment in product reviews, social media posts, and other natural language-based publications
• Service
  • Automatic Classification of Service Requests to route them faster to the right team or agent
  • Knowledge Base mining and enrichment generating information based on historical cases and interactions
  • Autonomous Agents using Large Language Models, augmented with the context of the interaction and customer data (stored in CRM systems) generating replies or deciding on actions to take for service agents or online customers (RAG application)
  • Using language models and generative AI capabilities to generate conversation abstracts

As mentioned, these use cases collect data from and generate data about the customer, transactional, demographic, and personal. Let us dive deeper into the aspects of data in adopting, developing, and deploying AI in the CRM domain of the company.

The AI Data Fundament in CRM

Data Management and AI go together. There is no AI without data. Let us investigate the most important aspects of data in AI, related to CRM:

Data Quality and Platforms

For AI to work properly, assuming the right models and technologies are in place and deployed, you must possess large amounts of quality data. Therefore, companies adopting AI must develop a data governance practice to manage the quality of the collected data. This responsibility is often combined with a center of excellence and the adoption of a scalable, modern data platform that can manage high-volume datasets of both structured and unstructured data.

Data Integration

Today, many companies still have a patchwork of business applications with customers and related data stored in silos. For example, order data is stored in ERP while ERP is often used as the customer master, especially in B2B business models. Digital traffic is collected from various servers and marketing platforms like CDPs. With customer interactions in CRM platforms, data must be collected, preprocessed, cleansed, and unified before it can be used for analytics and AI models for activation (anywhere). This required again, a scalable data platform and a modern integration approach where data can be collected in real-time to stay as relevant and timely as possible.

Data Trust and Security

The most important aspect of using customer data in AI (and in general) is the company’s responsibility to secure sensitive data and respect consumer privacy. Customers need to move with changes in privacy legislation, next to the data privacy requirements related to ethics, control, secure storage, and transparency.

Data Privacy and LLM

Generative AI introduces new challenges to data, specifically to privacy and security. Many companies leverage external LLMs to generate content using prompting grounded or augmented with internal data to retrieve more reliable and relevant results. When the prompt template is merged with internal (customer) data, the result is often sent to an external LLM. It is the responsibility of the company to ensure no customer data is retained at the site(s) hosting the LLM. This can be done by specific encryption and masking technologies.

In addition, the result coming from the LLM (resolution) must be validated against harmful language that could harm either the customer, the user, or even the (reputation) of the company.

Finally, there is a hallucination (LLM generating incorrect, unrelated, and/or unstructured results). Techniques to defend against LLM hallucination include chain-of-thought prompting or RAG (retrieval augmented generation) to ensure the resolution is coherent and relevant.

Adoption Strategy

In my view, the best way for companies to adopt AI in CRM quickly is to partner with a leading supplier of CRM solutions with pre-built AI capabilities instead of building everything yourself. The development and deployment of AI solutions require the following:

• Collect, cleanse, and store data on a scalable, open-data platform
• Hire skilled AI resources to build AI models for the right use case (and there are many)
• Monitor and adapt existing AI models continuously
• Select the right tools and keep them up-to-date to meet the latest business requirements
• Continuously keep an eye on AI innovations to stay ahead of the curve and stay competitive

The benefits of adopting Commercial-Off-The-Shelf (COTS) solutions for CRM and Data Platforms and partnering with a leading vendor; are numerous compared to building it yourself:

• Always the latest AI capabilities, tools, and innovation
• Best practice-based embedded AI capabilities for each common use case
• No need to hire an army of data scientists and AI technical specialists
• Faster time-to-market and time-to-value
• Less maintenance overhead

Conclusion

Every company is using CRM to manage its customer interactions and every leader is asking herself how to adopt AI quickly and sustainably while remaining compliant, ensuring the right technology, and being able to support every use case, while gaining and maintaining a competitive advantage. The question managers are asking themselves “Should I hire and build or partner and buy?” In my humble opinion, in CRM, a company should partner, buy, and adapt to specific needs with as little effort (coding) as possible leveraging industry best practices and common use cases. Especially in CRM, the use cases for AI are very common and are best provided by leading AI solution vendors. Not every company should reinvent the artificial intelligence wheel.