Customer Chatbots: The Good, The Bad, and the Ugly

Chatbots have changed how businesses handle customer service - sometimes for the better, sometimes not.

They can offer quick responses, scale routine enquiries, and stay available outside office hours.

But their value depends on how they are designed, secured, and handed over to people when the conversation gets difficult.

When routing and escalation are clear, they remove repetitive work. When they are not, customers end up repeating themselves or sharing sensitive details in the wrong channel.

Below is a practical look at the benefits, limits, and risks of chatbots in customer service, and when other channels may be a better fit.

The Good

At their best, chatbots help teams answer predictable questions quickly without tying up agents on every ticket.

That can reduce wait times and free people for complaints, exceptions, and conversations that need judgement.

24/7 Availability

One obvious advantage is availability. Chatbots do not need shift rotas to cover evenings, weekends, or peak surges.

Whether it is mid-afternoon or the early hours, they can respond to simple queries while human teams are offline.

According to HubSpot's roundup of chatbot research, 64% of users say 24-hour access is the best feature of chatbot support.

Round-the-clock access only helps if the bot can resolve the query or route it cleanly. Out-of-hours chats that collect personal data still need the same security discipline as daytime service.

"A chatbot can answer at 2 AM, but the firm still owns what happens to the data typed into that window. Availability is not a substitute for access control."

Paul Holland, Founder and CEO, Beyond Encryption (Mailock)

The next gains usually come from deflecting predictable work, not from stretching bots into conversations they were never designed to handle.

Efficient Management of Routine Tasks

For repetitive, predictable queries - order status, password resets, delivery updates - chatbots are often a good fit.

Some systems improve over time as they learn from past interactions, although quality still depends on training data and clear escalation rules.

Some industry estimates suggest chatbots can handle a large share of routine service questions, although the exact figure varies by sector and implementation. Potential.com cites figures up to 80% in its overview of AI chatbot use cases.

Volume deflection is only useful if the bot resolves the query correctly. Misrouted tickets can cost more than they save.

Cost-Effective Solution

By deflecting simple enquiries, chatbots can reduce pressure on contact centres and lower the cost per resolved ticket.

Talkative's overview of customer service chatbots cites studies suggesting organisations can cut support costs by up to 30% when bots are deployed selectively alongside human teams.

"If a bot collects account details or documents, the security model has to be explicit: what is stored, who can access it, and how it is deleted when the case closes."

Michael Wakefield, CTO, Beyond Encryption (Mailock)

Speed matters most when the answer is simple. The test is whether the customer leaves the chat with the issue actually resolved.

Quick Response Times

Most people dislike waiting in a queue for straightforward answers. Chatbots can reply almost instantly to common questions and keep the conversation moving.

A Chain Store Age survey summary reported that 62% of consumers would rather use a chatbot immediately than wait more than 15 minutes for a live agent.

Data Collection and Analysis

Chatbots also generate interaction data: common failure points, repeat questions, and drop-off moments in a journey.

Used carefully, that can help teams improve content, routing, and product fixes rather than simply deflecting volume.

Gabba's overview of chatbots and customer experience cites research suggesting 55% of businesses using chatbots see stronger lead quality, although results depend heavily on sector and setup.

"The useful part is what the business does with the pattern: fix broken journeys instead of sending every frustrated customer back through the same script."

Adam Byford, COO, Beyond Encryption (Mailock)

Those upside cases only hold when the bot stays inside its lane. The problems start when the conversation needs judgement the system does not have.

The Bad

Chatbots are not a fit for every conversation. When the query is emotional, ambiguous, or high stakes, limitations show quickly.

Customers notice fast when a bot cannot understand context or keeps offering the same unhelpful option.

Limited Understanding

Bots handle scripted flows well. They struggle more with nuanced language, sarcasm, or multi-part problems that do not match their training.

Many still rely on narrow intent models, which can produce answers that feel robotic or miss the point entirely.

Survey data collated by FinancesOnline includes findings that 59% of customers feel chatbots often misunderstand the nuances of human communication.

Understanding language is only part of the problem. Customers also expect appropriate tone when something has gone wrong.

Lack of Empathy

Even advanced AI cannot replicate genuine empathy in a complaint, bereavement, or cancellation conversation.

When emotions run high, customers usually want a person who can listen, acknowledge the issue, and adapt - not a fixed script.

Talkative's summary of chatbot limitations cites research suggesting 60% of consumers believe humans understand their needs better than bots.

Technical Glitches

Like any software, chatbots can freeze, loop, or fail mid-conversation - often at the worst moment for the customer.

Talkative's overview references survey findings that 43% of customers have experienced technical problems when using bots.

Language and Cultural Barriers

Natural language processing has improved, but multilingual and culturally specific service is still hard to deploy well.

For global organisations, that remains a real constraint. FastBots reports that only 29% of businesses have successfully rolled out multilingual chatbots.

Operational limits are frustrating. Security and governance failures can be far more serious for the business and the customer.

The Ugly

Beyond day-to-day friction, chatbots raise security, liability, and governance questions that boards and compliance teams cannot ignore.

Privacy and Security Risks

When chatbots collect names, account details, or documents, that data becomes an attractive target for attackers.

Weak retention, logging, or vendor controls can lead to breaches, reputational damage, and regulatory scrutiny.

The NCSC has warned about risks linked to large language models (LLMs) and AI-driven chat interfaces, including data leakage and insecure integrations.

"Customers need to know what channel they are using, what data it collects, and how long it is kept. A bot that feels helpful but stores too much is still a data-protection problem."

Emily Plummer, Marketing Director, Beyond Encryption (Mailock)

Security failures are not the only reputational risk. Incorrect or misleading bot advice can create liability as well.

Misinformation and Liability Issues

Sometimes a bot gives the wrong answer, or advice that creates legal or commercial exposure.

The Air Canada chatbot case - where a customer relied on incorrect bereavement fare information - is a well-known example of liability following automated guidance.

Ethical Concerns

As bots become more capable, transparency matters more. Customers should know when they are speaking to automation.

Biased training data can also produce unfair outcomes, which creates both ethical and reputational risk for the brand.

Job Displacement

Automation changes roles as well as tasks. Contact centres may shrink in some areas while growing in others, such as complex case handling or bot supervision.

In its 2020 Future of Jobs report, the World Economic Forum estimated that automation could displace 85 million jobs globally by 2025, while creating new roles elsewhere. Timelines and sector impact continue to shift as adoption evolves.

None of that means abandoning automation. It means choosing the right channel for each type of customer need.

Potential Alternatives

Most service teams need a mix of channels. Chatbots work best as one part of a model that includes humans and, where appropriate, asynchronous secure communication.

Practical complements to chatbot-only support include:

Human Agents

A hybrid model lets bots handle simple queries and escalates exceptions to trained agents with full context.

That keeps efficiency high while giving customers a clear route to a person when the bot runs out of road.

Other channels suit customers who do not want a live chat at all, or who need a richer conversation than text allows.

Self-Service Portals

Some customers prefer to find answers themselves. FAQs, status pages, and community forums can resolve issues without a live chat at all.

Good self-service content also trains bots and agents with consistent answers.

Video Calls

For visual products or high-consideration purchases, live video can add clarity that text chat cannot.

Retail Customer Experience describes how retailer DFS used virtual consultations to lift customer satisfaction above 90%, outperforming phone service in its comparison.

Secure Email

For sensitive matters - billing disputes, identity documents, legal correspondence - secure email is often a better fit than a real-time bot.

Customers can review and reply in their own time, while the sender keeps AES-256 encryption, recipient authentication, and message tracking on the thread.

Mailock is built for regulated teams that need to keep email as the delivery route while adding those controls and audit trails of what was sent and accessed.

Getting the mix right means matching channel to risk, urgency, and sensitivity rather than routing every enquiry through a bot by default.

Striking the Right Balance

Chatbots are here to stay, but they are not a complete service strategy on their own.

The strongest setups use bots for predictable, low-risk work and reserve people - plus the right secure channel - for complexity, emotion, and regulated data.

Encrypted email, video, and self-service each have a place. The aim is a journey that feels fast when it can be, and appropriately careful when it must be.

FAQs

Where Do Chatbots Work Well in Customer Service?

They can help with routine tasks, simple triage, quick answers, and out-of-hours support where customers need speed more than judgement.

Where Can Chatbots Damage Trust?

They struggle when queries are emotionally sensitive, ambiguous, or need empathy, escalation, and context.

What Should Regulated Firms Consider Before Deploying Chatbots?

Make sure human support is available, data handling is controlled, and customers know when they are dealing with automation.

References

18 Customer Service Chatbot Examples, HubSpot, 2023

Cut Costs, Boost Service: The 24/7 Power of AI Chatbots, Potential, 2024

Customer Service Chatbots: Everything You Need to Know, Talkative, 2022

Limitations of Chatbots, Talkative, 2024

Survey: Customer Expectations for Chatbots Are Rising, Chain Store Age, 2022

Chatbots: Friend or Foe? How They're Reshaping Customer Experience, Gabba, 2024

86 Critical Chatbot Statistics: 2021 Data Analysis & Market Share, FinancesOnline, 2021

How AI Chatbots Can Break Down Language Barriers, FastBots, 2024

ChatGPT and Large Language Models: What's the Risk?, NCSC, 2023

Air Canada Chatbot Case Highlights AI Liability Risks, Pinsent Masons, 2024

AI Overload: The Good, the Bad and the Ugly Impact on Customer Experience, Retail Customer Experience, 2024

Recession and Automation Changes Our Future of Work, But There Are Jobs Coming, Report Says, World Economic Forum, 2020

Reviewed by

Sam Kendall, 30.05.26

This content is for general information only and is not legal advice.