Challenges & Concerns with AI for CX Leaders

Recall the blockchain buzz in 2017?

Startups were raking in millions through Initial Coin Offerings (ICOs) based on flashy whitepapers and big promises. Yet, many fizzled out or underdelivered, causing a dent in investor pockets. However, there were a few companies that genuinely stood out, defying the skepticism surrounding the industry. 

Similarly, CX leaders today grapple with separating AI reality from the noise.

In the near future, we'll likely see a shakeup in the AI tool landscape. Some tools will fade away, some will merge with others, and others will carve out their niche and find lasting product-market fit. 

You might be lured into investing heavily in a new AI solution marketed as a game-changer. But without due diligence and understanding its practical applications, there’s a risk you’ll find yourself with an expensive tool that doesn’t align with your customers' needs. 

‍

What are the general limitations of AI for CX 

To help navigate AI-infused waters, here are six core areas that only humans can truly master, but in which AI can positively augment the human’s ability:

Judgment

AI aids decision-making by analyzing vast data, offering insights, and automating tasks. While it excels in predictive analytics and natural language processing, human judgment remains crucial. ‍

In retail, AI can predict which items are likely to be in demand based on historical data and current trends. But human judgment is crucial when an unforeseen event, like a sudden celebrity endorsement, boosts the popularity of a product.

Relationships

AI can't replace human emotion in building relationships. 

For example, when a customer gets the wrong dress color, a chatbot's generic response can feel impersonal and aggravate them. However, AI can still enhance customer service: it analyzes communication patterns to guide relationship improvements, assist in scheduling, route customers effectively, and equip agents with better context.

Creativity

AI aids in creativity by suggesting ideas and automating tasks. However, its creative scope is limited to its training data. While AI can rearrange existing "puzzle pieces," humans innovate by crafting new ones, pushing boundaries beyond AI's capabilities. 

AI tools can segment customers for targeted emails based on their purchase history. But when creating a unique marketing campaign that tells a brand story or resonates with a specific cultural event, human creativity is at the forefront, transcending mere data points.

Agreements

AI supports customer interactions with data insights and streamlined communication. While it can propose solutions based on customer data, genuine customer engagement requires human empathy and understanding. 

For example, AI can find the fastest shipping options, but only a human can express genuine commitment and assurance to deliver the right item in time for the customer's big day.

Belief and understanding

AI doesn't hold beliefs. While your beliefs guide your approach to customers, AI functions purely on data. It can't truly connect with customers, but it can provide valuable insights to enhance the customer journey.

AI might suggest marketing strategies based on data from various regions. Still, human judgment is required to ensure campaigns respect and resonate with local cultures, beliefs, and sentiments. An ad that's effective in one country might be viewed as disrespectful in another.

Emotions

AI can't experience emotions like humans. 

Suppose a customer wants to return a gift they bought for someone who has unfortunately passed away. While AI can handle the return by following the set return policy, it lacks the capacity to offer genuine condolences or thoughtful gestures, such as sending a sympathy card or flowers.

While AI can provide valuable assistance in certain areas, it's important to recognize that there are limitations to what it can do. 

Humans should still be involved in areas where human interaction, empathy, and emotional intelligence are necessary for success. By working together, humans and AI can create a more efficient and effective future.

How AI raises new issues for CX teams  

‍Every fresh tech wave offers us both perks and puzzles to solve.

While automobiles revolutionized transportation, making it faster and more efficient to move from one place to another, they also led to challenges such as traffic congestion, environmental pollution, and a rise in road accidents. 

Facebook and X have enabled global connection, instant news dissemination, and created platforms for social movements. However, they’ve also been implicated in the spread of fake news, cyberbullying, and have raised concerns about data privacy and mental health. 

The same give-and-take is true for AI. It's totally changing the business-customer dance, but it's also tossing a few curveballs to CX leaders. 

Selecting vendors with seamless AI integration

Have you ever tried to build a massive Lego set without the manual? I'm guessing not, because it's a near-impossible task. 

Implementing a new AI tool can feel a little like that. You’ve got a ton of pieces, but figuring out how to put them together to build something effective is challenging. 

Your inventory management is stored in your ERP, customer data is saved in CRM, policies and procedures are fixed in your DMS. And this is just the beginning.

‍

As large enterprises have increased their app usage from 8 in 2015 to 447 in 2024, the need for clarity and guidance during the AI implementation is crucial. 

When selecting an AI vendor, it's paramount to find those who effortlessly weave their AI into your CX processes. The ideal partner offers not just stellar AI tech, but also a suite of integrations and top-notch support to ensure a smoother implementation.

“Managers and directors are concerned about the level of effort required to implement AI, especially with documentation scattered across multiple places.” - Larry Barker, Senior Customer Experience & Operations Manager, Teamshares

Addressing AI security and compliance in AI implementation

Like any other software, AI comes with the normal set of security and compliance issues. Navigating through regulations like GDPR, CCPA, and HIPAA is a given, so we won't delve deeper into that here.

Yet, what sets AI apart is its capability to generate content independently. 

AI is about data—how it's gathered, processed, and used. 

Your Apple Watch monitors your heart rate, counts your steps, and tracks sleep, all to provide insights into your health and offer customized fitness advice. Spotify delves into your song choices and likes/dislikes to curate playlists that match your taste.

Similar to this, your AI model will be trained on your proprietary data, such as your knowledge base and your ticket history, in order to make decisions. 

In other words, making the most of your AI implementation means you should do your base vendor evaluation, while also looking to understand how your data is being leveraged. 

‍

‍

Guiding principles for data management

Use only necessary data.

‍Refrain from overloading your AI with irrelevant data. Focus only on data sources that are essential for your specific use-case.

For instance, if your AI is focused on customer support, only provide it with relevant ticket histories and FAQs. HR records or unrelated financial data should remain inaccessible.

Understand model training with your data.

‍While AI models need to be trained on your data to function effectively, it's essential to ascertain how that data is managed.

Data should be logically segregated from data of other clients. Proper metadata tags should be used to categorize your data accurately. Ensure role-based access control to limit who can view or modify the data. Always store data in encrypted form, whether at rest or in transit.

Anonymize data when applicable.

‍Not all data sets need to be presented in their raw, identifiable format. Where direct context isn't crucial, prioritize the anonymization of datasets to ensure privacy.

For instance, while direct queries from a customer might need personal context, CRM data pulled from platforms like Salesforce should be anonymized to ensure no Personally Identifiable Information (PII) is exposed.

Ask vendors direct questions.

Deeply understand your vendor's data practices. The more you know, the better. Key questions include:

• What information is shared, and with whom?
• How is the data stored and encrypted?
• Is my data used to train your AI models?
• What's the data retention policy?
• Can I request data to be purged or deleted upon ending our engagement?

Check out our questionnaire template for more questions.

‍Choose vendors offering opt-out options.

Your vendor should empower you with the choice to opt-out of certain AI services or data practices. This transparency is essential to cater to varying privacy and security preferences.

Treat AI tools similarly to how you would treat new employees

Picture this: you hire a new agent. On their first day, you wouldn’t blindly trust them, toss them a headset, and expect flawless customer service.

You would train, monitor, and continually check their work. The new agent should spend the initial weeks diving deep: learning your company ethos, grasping product intricacies, and getting familiar with what your customers expect. 

Rushing into things could tarnish your brand image and leave customers feeling overlooked.

This onboarding, combined with guidance and time to settle, ensures your new team members represent your brand authentically and deliver quality service. 

This is precisely how AI tools should be introduced into your CX operations. 

‍AI isn't magic; it's a learning machine. It thrives on guidance, much like a new team member.

Here's your AI onboarding checklist:

• Custom training: Just as new agents learn your ticketing system's ins and outs, an AI model needs to learn from your data to do its best job.

• Practice first: Before AI faces the high-pressure hours, let it 'shadow' and refine its skills in a controlled and closed environment.

• Feedback loop: Just as agents benefit from post-interaction feedback, AI’s efficacy should be consistently assessed and tweaked. 

• Controls and safeguards: To address concerns that AI might make an incorrect decision, it's essential to implement robust controls and safety measures.

• Data care: Onboarding a human or digital agent to confidential information requires discretion. Prioritize data security with AI, sharing essential details and leveraging anonymized datasets wherever feasible.

What stakeholders you should involve in the AI implementation 

When rolling out AI, you'll need to bring several stakeholders to the table. Think of this table as your kickoff checklist. But remember, depending on your company's specifics and your industry, you might need to loop in some other key folks.

CX leaders: Vision and strategy for AI implementation

• Define AI objectives and desired outcomes
• Establish communication with your AI vendor 
• Continually evaluate vendor performance against set objectives

Data Scientists: Model development, data analysis, and validation

• Collaborate with the vendor's team for data integration and sharing 
• Ensure vendor uses ethical and unbiased data practices 
• Validate and test vendor-provided models rigorously

IT Team: Infrastructure, integration, and deployment

• Collaborate with the vendor for seamless tech integration
• Ensure vendor meets cybersecurity standards
• Maintain an exit strategy, should you need to switch vendors or bring AI in-house

Legal & Compliance: Ensure all AI practices adhere to regulations and standards

• Work with the vendor to ensure complete regulatory compliance
• Regularly review vendor's data handling and storage practices
• Draft clear contracts detailing responsibilities, data rights, and IP concerns

CX Frontline Staff: Interacting with AI tools and providing feedback

• Ensure vendor-provided tools are user-friendly and effective
• Relay real-time feedback to the vendor for refinements
• Engage in vendor-offered training sessions

Customers: Users of the AI-enhanced CX processes

• Educate customers about the collaboration with the AI vendor and its benefits
• Offer opt-outs if customers are wary of specific vendor-driven AI implementations
• Gather customer feedback specific to AI enhancements for ongoing refinement

‍