Article:
Transforming Complaints into Insights: How AI is Changing the Way We Manage Feedback

Early AI focused on symbolic reasoning – the focus on using symbols and logical rules to problem solve and make assumptions. Since then, AI has continued to evolve, becoming increasingly integrated into various aspects of our lives, whether this be personal (think biometric authentication and photo editing on smart phones) or professionally (data analysis and natural language processing).

​

The Rise of AI in Complaints

AI complaint management started to emerge in the 2010’s, using tech like chatbots and AI-powered analysis. The development of AI for complaint management has been driven by the need for more efficient solutions, as well as the ability to analyse large volumes of customer interactions, thus supporting root cause analysis.

During this paper I intend to explore how AI is transforming complaint management, highlighting key benefits, and sharing my experience thus far.

​

Challenges with Traditional Complaint Handling

The challenges we face today with traditional complaint management tend to be slow response times, overwhelmed staff, inconsistencies, missed insights and the growing volumes being received.

​

Slow response times are often due to lack of correct resource, lack of skill or knowledge to manage the complaint, complex complaints needing detailed investigation and access to essential information.

​

Staff can often feel overwhelmed when dealing with complaints, either by the volume received, the volume they are expected to close or the topic of the complaint itself.

Everyone has their own personality and style, and this comes across in the way they investigate, respond and outcome they provide. This can lead to inconsistencies across the end-to-end journey and further complaints. It also effects the insights the resulting data provides.

​

In group exercises, I’ve seen how the same complaint scenario results in widely different responses – varying in tone, content and detail.

​

Finally, we have seen an increasing trend since Covid where consumers have greater expectations of products and services, expecting more, expecting faster. Consumers are also more in touch with how they feel and sharing this with their providers, resulting in more complaints due to expectations not being met. Websites and influencers also contribute to this, often sharing tips on how to complain and get something, often financial, as part of the outcome.

​

How AI is Changing the Game

AI-powered complaint management systems provide efficiency and quality improvements through automating and improving the handling of customer complaints, leading to faster resolutions and improved customer satisfaction.

​

This involves using machine learning, natural language processing, and data analytics to categorise complaints, analyse sentiment, provide real-time insights, and personalise customer interactions.

​

Automation and Efficiency: AI is triaging and resolving issues instantly by using techniques like Natural Language Processing (NLP) and sentiment analysis. I’m not going to pretend I knew what that meant so let me share what I found out. NLP algorithms can be trained on large datasets of text and speech to learn patterns and relationships in language. This enables computers to understand, generate, and manipulate human language.

​

It essentially bridges the gap between human communication and computer comprehension, allowing machines to interact with language in meaningful ways. Sentiment analysis is an NLP technique that analyses text to determine the emotional tone or sentiment expressed, classifying it as positive, negative, or neutral. E.g. It would recognise key words and tone, and flag that a complaint is Urgent.

​

Most AI complaint management systems available can either integrate with existing platforms or be used in their own right. They can have API’s that link into your systems, pulling key information across to support with investigations, thus reducing the time colleagues spend searching multiple systems for information. The system we use can listen to calls, read documents and answer questions, enabling colleagues to get to the answers more efficiently.

​

AI can then help write the customer response, focusing on the issues raised by the complainant and reduce quality fails by correcting spelling and grammar errors, learning your brand tone to improve quality of response.

​

Data Driven Decisions: AI systems can often collect better information to be used not only during the end to end complaint journey but also to offer guidance on what the outcome should be, manage teams and productivity more effectively and provide strategic insights  and root cause analysis that identify recurring problems to prevent future complaints.

​

By utilising what AI has to offer, cases are closed faster enabling customers to receive a faster response, quality and customer outcomes are improved, regulatory risks can be reduced, and you have the ability to ensure internal processes and decisions align with best practice.

​

My Experience with AI in Complaint Handling

We started to use AI last year to support our colleagues with writing customer responses, with a focus on quality. With a few tweaks and examples provided, the system learnt our brand tone fairly quickly and continued to learn through engagement with colleague feedback. This alone provided efficiency gains as instead of spending, on average, 30-60 minutes creating a response, the AI produced one in minutes using the complaint detail and investigation notes.

​

The challenge we faced was getting the colleagues to use it correctly to retain those quality gains. Many colleagues would use the AI to generate the response and then manually edit instead of asking the AI to make the amendments. Through lots of coaching and feedback we managed to solve the majority of this.

​

Once we gained confidence, we started to use it to support with identifying complaint issues based on the complaint submitted to ensure colleagues were covering all points in their responses. We could also tailor the response to fit the customer. E.g. customers sending 3-page complaints would receive a longer more detailed response vs a customer sending in a short, sharp, no nonsense complaint would receive a snappy response.

​

We’re now in the process of developing the system based on our requirements to provide more tailored efficiency gains. E.g. automate classification selection, vulnerability flags, branding (through API feeds), pulling of customer records, calls and expanding the root cause functionality. This last one I am especially looking forward to as it will enable us to gain better insights into the deeper root causes quicker. This will be done not only through the reporting, visuals and verbatims but also the ability to ask AI questions, getting it to do a lot of the leg work to get to the issue so we can focus on the actions to improve.

​

Balancing AI with the Human Touch

I understand there are concerns with the loss of human touch and over reliance, but personally I believe this is within our control. It’s the responsibility of those who use it to set guidance on where the human checks are and how much of the functionality is used, thus reducing the risk of AI bias.

Practical ways to achieve balance between automation and human oversight include;

​

• Human in the Loop Systems,

• Regular Reviews of AI Decisions,

• Hybrid Responses,

• Transparent Handover,

• Empathy Filters,

• Limited Automation Scope

• Customer Feedback Loops.

 

Human in the Loop Systems – Use AI to handle routine or first level complaints, but route complex or emotionally sensitive cases to human colleagues. E.g. If the AI detects frustration or legal language it escalates to a human immediately.

​

Regular Review of AI Decisions – Involve teams on a regular basis to audit AI handled cases to ensure the responses are appropriate and are fair and in the correct tone.

 

Hybrid Responses – This is the approach we are currently using which lets the AI draft the initial response and then colleagues can edit or approve before sending. Thus, using the AI to support with consistency of approach, quality etc, but with colleagues still having ownership of the outcomes and ensuring the responses have the right tone and empathy.

​

Transparent Handover – Clearly communicate when a customer is speaking with AI and when they are being handed over to a human. This helps to build trust and avoid frustration. Many online services already do this with online chat bot services.

​

Empathy Filters – Utilise colleagues to review AI handled complaints for emotional nuance – sometimes the AI resolves the issue but misses the emotional need.

​

Limit Automation Scope – Define clear boundaries for what the AI can and shouldn’t do. E.g. as above, don’t let it handle serious cases without a human touch.

​

Customer Feedback Loops – Gather feedback on AI handled interactions and compare it to human handled ones and use the feedback to continuously fine tune both AI performance and when human intervention is needed.

​

Looking Ahead

So, what does all of this mean for the future of complaint handling. I believe that AI is reshaping the way forward on how we handle complaints – streamlining processes, uncovering insights and improving quality and response times. While automation brings undeniable benefits, it’s clear that maintaining a human touch is still important, especially when empathy, nuance or judgment are needed.

​

My experience has shown that the most effective approach blends the strengths of both: letting AI handle the routine while empowering colleagues to step in where it truly counts. As AI continues to evolve, the aim shouldn’t be to replace human involvement, but to enhance it, creating smarter, faster and more compassionate ways to respond to customer concerns.

​

​