Upskilling Ourselves in AI

#1 in a Series for CX and Customer Service Leaders

Back Story

As you might recall I'd spotted a widespread need to better understand AI after hosting a couple of conferences. Especially around Generative AI which caused a huge global response and promised so much in terms of contact centre evolution.

It's already disruptive. There are many decisions that need making. So informed judgement is going to be critical. Especially in CX and Service - all roles included. But the current way we talk about Generative AI suggests too many people relate to it more as alchemy than science.

The fact that an earlier post on this issue received over 17,000 impressions suggested there is widespread recognition of the need to collectively upskill.

So, I decided to co-author a course with a friend, Brian Manusama who’s been a top analyst in Conversational AI and CX at Gartner. We are making good progress. You can request an email alert at the end of this post once the course goes live.

We've discovered there's a lot to cover. Even for a foundation course. Of course, course design is all about what you keep in and what you discard or only mention briefly. Topics needing more space than we could provide in the course have now made it into this series.  

This will cover a range of topics including ‘must haves’ in a CX/CS AI strategy, early best practice in AI deployments, emerging roles and skills, managing AI quality such as bias and hallucinations, preparing for AI regulation, understanding RAG and how LLMs behave.

There is no expectation that you have prior knowledge. Only that you are curious enough to want to find out more.

This is the first article. It’s a scene setter and overview of the issues that are going to be exploring later in the series.

OK. Orientation done. Time to dive in.

The AI Challenge

The growing impact of artificial intelligence has opened a new era of customer experience and service possibilities. A global study by the XM Institute found that after a positive service experience, customers are 5.2 times more likely to purchase more and 5.6 times more likely to recommend a company.

AI offers immense potential to enable the kind of personalised, empathetic, and low effort experiences needed to drive such loyalty.

However, the AI landscape can seem like a "black box" for CX and service leaders. The barrage of technical jargon, hype cycles, and vendor promises makes it challenging to separate fact from fiction.

In fact, Large Language Models (LLMs) which enable Generative AI are literally ‘black boxes’. Both to their original designers and those who build applications on top. Such as the ones used in CX and Customer Service.

There is a growing industry busily trying to catch up with this ‘black box’ challenge and mitigate the downsides. As a sign of the times, there is now a Hallucination Index to guide better LLM purchasing decisions!

In someways this is just a re-run of that age-old data challenge: the quality of data fed into an AI model impacts the quality of output. Especially when that output also has variation built into the model’s core design such as LLMs. The lesson here is that data quality is foundational to the North Star of AI powered CX and service. So, make it central in your AI strategy.

Second point is that we already know from previous generations of customer technologies, they are enablers but cannot deliver full value out of the box. Mindset, behaviour, operating models all need recalibrating. So, pitch for the change budget, invest the time, use the right methodology and tools. You know the drill by now.  

But we are getting ahead of ourselves. To realise value from AI, each of us must first develop a practical understanding of the technology and its business implications.

Defining the AI Landscape

At its core, AI refers to computer systems that can perform tasks normally requiring human intelligence. The field of AI has evolved rapidly since its early days in the 1950s. These AI systems were rule-based and focused on narrow tasks, but the emergence of machine learning in the 1980s allowed AI to start learning from data.

The 2010s saw another leap forward with deep learning, which enabled more advanced AI capabilities like natural language processing and computer vision.

More recently we have seen the rapid adoption of Generative AI which is causing a fascination as deep and compulsive as smartphones and social media have had on our lifestyle and self-identity.   

Each wave of innovation has been triggered by faster and cheaper computing capababities.

The most important types of AI in CX and Customer Service we should have a working knowledge of are:

Machine Learning: Algorithms that enable computers to learn and improve from experience (data) without being explicitly programmed. Supervised learning (learning from labelled data) is often used for tasks like sentiment analysis and ticket classification, while unsupervised learning (finding patterns in unlabelled data) is used for customer segmentation and anomaly detection.

Deep Learning: Advanced machine learning based on artificial neural networks* that enable more complex and human-like capabilities. Deep learning excels at processing unstructured data like text, speech, and images: powering applications like emotion detection from speech, image-based product recommendations, and Generative AI.

*neural networks are collections of machine learning algorithms modeled to replicate the arrangement of neurons in the brain.

Natural Language Processing (NLP): Techniques that enable computers to understand, interpret, and generate human language. NLP enables applications we use every day, from search engines and chatbots to autocorrect and voice assistants.

While these are often used in combination, understanding their distinctions is crucial for aligning AI capabilities with specific CX and service use cases.

Real-World AI Applications

AI is being applied across the customer journey to transform experiences and enable new service possibilities. We are becoming increasingly familiar with a number of these use cases as we learn how to extract their value from piloting to scaling them.

Conversational AI: AI-powered chatbots and voice assistants can handle routine queries 24/7, providing instant support and freeing up customer facing colleagues for more complex issues. Advanced conversational/ generative AI is becoming better at understanding and retaining context which contributes to more back and forth conversation. Juniper Research estimates that chatbots will deliver $11 billion in annual cost savings by 2025.

Intelligent Call Routing: AI can analyse customer data in real-time to route calls to the best available agent based on skills, experience, situation and sentiment. This improves first-contact resolution and customer satisfaction while optimising colleague utilisation.

Agent Assistance: AI can provide real-time guidance to front line teams during customer interactions, suggesting relevant knowledge articles, troubleshooting steps, next-best-actions or coaching tips. This reduces handle times, improves consistency, and helps new team members ramp up faster. A Salesforce case study found that AI-powered agent assistance increased first contact resolution by 32%.

Sentiment and Emotion Analysis: AI can analyse the sentiment and emotional tone of customer interactions across voice, text, and video channels. This allows companies to identify frustrated customers proactively, assess overall customer satisfaction, and monitor agent empathy. Microsoft reports that organisations using AI for customer experience have 2.5 times higher customer satisfaction rates.

Predictive Analytics: Machine learning models can predict customer behaviour, preferences, and lifetime value. This enables proactive customer service, personalised recommendations, and targeted retention efforts for at-risk customers. McKinsey research shows AI-based personalisation can deliver 5-8 times the ROI on marketing spend.

Robotic Process Automation (RPA): AI-powered RPA can automate repetitive post-interaction tasks like updating CRM records using Generative AI conversation summaries, generating shipping labels, or processing refunds. This frees up agents to focus on higher-value activities.

Anomaly Detection: Unsupervised learning algorithms can identify unusual patterns in customer behaviour or feedback that may signal emerging issues or opportunities. This allows companies to proactively address problems before they escalate.

Knowledge Management: AI can help organise and surface the most relevant content from knowledge bases to assist colleagues and customers. Some AI tools can even auto-generate FAQ content and chatbot responses from past interactions.

There are more but enough to show the scope and impact of current AI capabilities.

And just to drum home the point. Getting from your current capabilities to the North Star just described is why an AI strategy is vital. Checklist management is not going to get you there. Just as ‘around town’ trainers are not the right footwear for serious mountain climbing even though some still ignore the obvious!

Key Challenges and Best Practices

As usual there’s a lot to get right to extract AI’s full value. Here are some key challenges and best practices to consider as a sample of what other articles in this series will dive into.

Align AI with CX Strategy: AI should not be pursued as a standalone initiative, but rather as part of a unified CX strategy. Start by clearly defining the customer and business outcomes you want to achieve, then identify where AI can best support those goals. This is key but often missed during AI goldrush moments!

Ensure Data Quality and Governance: Bias, accuracy, consistency and security are all major topics requiring collaboration across diverse disciplines to develop clear data governance policies that support emerging regulation such the EU AI Act and reverse declining consumer trust in AI.

Prioritise Ethics and Explainability: Ensure AI systems are designed and used ethically, with clear policies around data privacy, bias prevention, and responsible use. Aim for explainable AI whose decisions can be understood and audited. This will represent a milestone achievement given the current state of affairs so needs dedicated and effective leadership.

Focus on Human-AI Collaboration: AI should be viewed as a tool to augment human capabilities, not replace them entirely. Therefore design AI systems with the goal of empowering colleagues and enabling seamless handoffs between AI and human assistance.

Although this is the ‘official’ industry and vendor point of view, don’t underestimate the way AI is being viewed at executive level for headcount reduction. This is going to grow into a very public debate with organisations being judged on actions not words. So be clear about your point of view and influence the internal debate.

Invest in Change Management: Implementing AI often requires significant changes to processes, roles, and skill sets. Invest in communication, training, and support to help colleagues adapt and embrace AI as a tool to make their jobs easier. This starts with foundation training which is why we are developing the course-)

Start Small and Iterate: Trying to implement too much AI at once can be overwhelming. Start with a narrow, well-defined use case and focus on delivering measurable value. Then iterate and expand gradually based on lessons learnt. Scaling AI remains challenging even for the most ambitious organisations at this point. So seek out what they have discovered.

Monitor and Optimise Continuously: AI models need continuous monitoring and refinement based on real-world performance. Establish clear metrics to track the impact of AI and processes for ongoing optimisation. This approach also applies to the previously mentioned topics such bias, accuracy, completeness etc. Finally stay focussed on operating costs as well. Generative AI can be expensive. Vendor pricing and business cases need regular auditing to track actual ROI.

Closing Thoughts

In reality, I’m a cautious optimist. I’m all too conscious around the common fate of great tech. For instance, I’ve not even opened the can of worms when bad actors start using Generative AI.

I also know from decades of engagement and observation how shareholder driven organisations think and view opportunities. Gen AI will be no different.

I’m also aware how timid some organisations have become around being willing to fully turn into the slipstream of disruptive forces and just get on with it. I’m hoping that the average speed of travel will be more than just a plodding pace this time. That’s the optimist in me talking!

And yet.

The fact is AI has the potential to revolutionise CX and customer service. Organisations can operate more efficiently, understand customers more deeply, and deliver more personalised, empathetic experiences at scale. Colleague well-being and opportunities also rise with this tide as well. So potentially a win-win all round.

AI’s latest generation is also changing organisational fortunes. Go check out Nvidia’s share price and find out why that’s happened. Look up the story about Google considering charging for an ad free search service and ask yourself why on earth would an organisation be willing to upend a massive money-spinning business model that has made them one of the largest on the planet. What forces are disrupting them to make them even consider this?

Our lives are being changed in front of our eyes. At work and at home. It’s a goldrush moment in the AI timeline right now. I’m keen we all get to influence this future. With that in mind, I’ll sign off this introduction article with a snippet from the narrator’s script in the up and coming AI foundation course.

“A key motivation in us producing this course was to make sure the ongoing debate and decision-making being made about AI is informed and inclusive of those who should have a voice.”

I’m aiming to make the ongoing conversation as diversified as possible by making sense of ‘what’s underneath the bonnet’ for anyone whose technical alphabet used to stop at ‘algorithm’.

Sign up here for your 'Course Now Live' alert.