Problem Solved: Using AI to Drive Much-Needed Engagement with Customer Reviews

As VP, managing partner at DAC Group UK, Mike Fantis has over 18 years’ experience in the digital marketing industry. His knowledge spans blue-chip brands in every major vertical and he now develops measured media strategies for brands including David Lloyd Clubs, NetJets, Center Parcs and Wickes whilst ensuring high performance across the company’s client portfolio.

The Problem

We found that many brands are neglecting to respond to customer reviews on their Google Map listings. Engaging with customer reviews is essential, as it plays a crucial role in driving success and visibility in local search. Brands that fail to respond to reviews risk lower visibility in non-branded search results, potentially missing out on valuable traffic and revenue opportunities. Proactively addressing reviews can enhance a brand’s online presence, build trust, and positively influence local rankings.

The first challenge we discovered is that Google Maps listings often slip through the cracks between different teams. While these listings are a valuable digital asset, they also function as a retail channel—but they are typically managed by online teams, leaving retail and customer service without proper visibility.

A common issue brands face is a lack of dedicated resources to respond to reviews on Google Maps, in addition to handling traditional customer service channels, leading to missed opportunities for engagement and customer satisfaction.

Ideation

We developed several modules within our multi-location marketing platform TransparenSEE to save time for brands. For example, we created pre-populated responses that users can edit to speed up the response time, as well as automated responses for reviews that do not have any content.

Some users went to great lengths to leave detailed reviews, which required equally thoughtful responses to address all the points raised. As such, our pre-populated responses still needed significant manual effort to produce a great response. To solve this, we developed an AI-powered tool for responding to reviews. This tool ensures that every aspect of the review is addressed.

However, the response isn’t published until it’s reviewed and approved by a team member. This approval process guarantees that the response aligns with the brand's tone and style. In cases where the AI-generated response missed the mark, we could train the system to improve its tone and language for future interactions.

Building on this AI initiative, we also created a solution that allows reviews to be uploaded and analysed for common themes. This product helps identify trends and provides recommendations for operational improvements.

The process was straightforward once we identified the need. We asked questions of how the AI could be utilised and we scoped out a solution that went straight into development.

When building our AI review response product, we drew inspiration from a variety of sources. We examined how leading brands engage with customers in their reviews, focusing on personalised, thoughtful responses that address all aspects of customer feedback. We also looked at customer service best practices across industries, understanding the balance between efficiency and empathy in handling large volumes of reviews.

Additionally, we studied successful AI applications in conversational platforms and chatbots to understand how natural language processing could be harnessed to generate relevant, on-brand responses.

Finally, we took cues from sentiment analysis tools to ensure the AI could accurately interpret the tone and emotional context of each review.

These insights helped us design a solution that not only produced high-quality AI responses but did so in a way that enhanced customer engagement and maintained brand integrity.

Prototype & Design

One of the most interesting aspects of the designing and prototyping process for the AI review response product was creating a system that could assist teams in drafting personalised responses without fully automating the process.

The challenge was to ensure the AI provided helpful suggestions that captured the customer’s sentiment and addressed all key points, while still allowing human oversight and input. It was fascinating to design a tool that could intelligently interpret complex reviews—such as those with multiple concerns or emotionally charged feedback—and offer structured responses that felt thoughtful and aligned with the brand’s tone.

However, final control always remained with the user, ensuring that every response could be customised to fit the specific situation.

One of the most interesting challenges that emerged during the development of the AI review response product was the balance between human input and AI assistance. There were many discussions around how much the AI should "suggest" versus how much control the human team should retain.

We wanted the AI to offer useful, structured responses, but without crossing the line into fully automating customer interactions, which could feel impersonal or off-brand. This raised some fascinating conversations about the nature of customer service and how much personalisation is needed to maintain trust. 

Another challenge was ensuring that the AI could understand the nuance of language, especially in emotionally charged reviews. It wasn’t just about responding to the content but also picking up on tone, context, and subtext.

The question of how deeply the AI could interpret human emotion sparked some of the most thought-provoking conversations, as we explored the limitations and potential of AI in these delicate scenarios.

At this stage of developing the AI review response product, we had to collaborate with a range of specialists to maximise the capabilities of an existing AI tool while tailoring it to our needs. First, we worked with natural language processing (NLP) experts who helped us adapt the pre-built AI for our specific use case.

Their expertise was essential in fine-tuning the AI to understand and respond appropriately to the nuanced language of customer reviews.

They assisted in configuring the tool to better capture sentiment and context, making sure the suggestions were accurate and relevant.

We also partnered with UX/UI designers, who were key in creating an intuitive, user-friendly interface. Since the AI was already built, their focus was on ensuring that users could easily review and adjust the AI-generated responses.

They helped streamline the workflow so that the product felt seamless and accessible, even for those without technical expertise.

One of the creative risks we encountered was integrating an existing AI technology into a process where human oversight was still essential. The novel aspect wasn’t in building the AI but in how we adapted and designed around it to fit our specific needs—balancing automation with personalisation.

The challenge was making sure the AI could generate useful, on-brand responses without replacing the human touch, which is critical in customer interactions. We took a creative risk by relying on the AI to handle complex, nuanced reviews, but we mitigated that risk by incorporating a layer of human approval.

This gave us confidence that the AI’s suggestions could still be fine-tuned by the user to ensure they aligned with brand standards and customer expectations. The human-in-the-loop approach allowed us to take advantage of the AI's efficiency while maintaining full control over the final output.

Live

Testing and iteration were at the core of our approach to this project. From the outset, we adopted an iterative mindset, knowing that the initial integration of AI would require continuous refinement to meet the specific needs of our users and brands.

We began by testing the AI’s ability to generate relevant and accurate responses to various types of customer reviews—ranging from simple comments to complex, multi-point feedback.

In the early stages, we found that the AI occasionally missed key nuances or didn’t fully capture the tone we wanted. This led us to conduct frequent user testing with internal teams, allowing us to gather insights on where the AI fell short or exceeded expectations.

During testing, some of the "spiciest" issues and back-and-forth discussions revolved around balancing the AI’s capabilities with brand tone and authenticity.

One major challenge was that, while the AI could generate responses quickly, it sometimes missed the subtleties of tone—especially in emotionally charged reviews. The AI might produce a response that was factually correct but felt too robotic or lacked empathy.

This led to several rounds of tweaking the system, adjusting how the AI interpreted sentiment, and creating guidelines for more sensitive topics.

Another key issue was ensuring that the AI didn’t offer generic responses to detailed, multi-point reviews. Early testing showed that the AI occasionally defaulted to overly broad or pre-canned answers, which risked frustrating customers rather than addressing their concerns. To fix this, we had to refine how the AI prioritised different parts of a review, so it could deliver more specific and thoughtful responses.

One key area where data is driving improvement is the analysis of response accuracy and tone. By tracking how often users make edits to AI suggestions before approval, we’re able to pinpoint specific patterns—whether the AI is consistently missing certain nuances or struggling with particular types of reviews. This data allows us to fine-tune the AI’s ability to better understand sentiment, prioritise key points in reviews, and generate responses that require fewer user adjustments.

Looking back over the whole project, one of the most personally interesting aspects for us was seeing how AI and human creativity could come together to solve a real-world problem. It was fascinating to work with an existing AI tool and tailor it to meet the nuanced needs of customer service.

Finding the right balance between leveraging the AI’s efficiency and maintaining the human touch was both a technical and philosophical challenge that kept us engaged throughout.

We also found the iterative process of refining the AI’s tone and responses particularly compelling. Watching how user feedback and data helped shape the product in real-time made it clear how much of an impact even small tweaks could have.

Each iteration brought the tool closer to delivering responses that felt genuinely on-brand and empathetic and seeing that evolution firsthand was incredibly rewarding.

This project has had a significant impact on our clients’ businesses, improving both efficiency and customer engagement. AI-assisted responses helped the teams manage high volumes of reviews more quickly and consistently, ensuring timely, thoughtful replies. As a result, customer satisfaction improved across most of our clients, and their brand's visibility in local search results also increased.

In addition to freeing up time for the different teams by reducing manual review management, the review analysis feature provided valuable insights into customer feedback, enabling our clients to identify trends and make operational improvements.

Overall, the solution not only solved the original problem but also delivered strategic benefits for our clients.