Chatbots and AI-powered virtual agents can help contact centres manage growing volumes of customer enquiries without requiring every interaction to involve a human adviser. But successful automation is about more than the number of conversations handled without human intervention.
A chatbot might answer a question, provide a link or conclude a conversation without ever resolving the customer’s problem. The customer may then try another channel, repeat the enquiry or abandon the process altogether.
From the contact centre’s perspective, the interaction may appear to have been successfully contained.
From the customer’s perspective, very little has been achieved.
As organisations invest in increasingly sophisticated self-service technology, that distinction is becoming critical.
The objective should not be to keep more customers inside the chatbot. It should be to help more customers reach a satisfactory resolution.
Understand the Difference Between Containment and Resolution
Containment and resolution are related but different measures.
Containment generally describes interactions completed without escalation to a human agent. Resolution measures whether the customer’s underlying need was successfully addressed.
A customer asking for delivery information might receive a tracking link and leave the conversation. That could represent a successful outcome.
Another might receive the same link but still be unable to establish where their parcel is. If they subsequently telephone the contact centre, the original chatbot interaction has not necessarily reduced customer effort.
Zoom’s February 2026 Virtual Agent 3.0 announcement specifically emphasised moving beyond containment towards end-to-end customer resolution, including better handling of complex enquiries and smoother transfers to human advisers.
For contact centre leaders, the practical implication is straightforward.
Don’t measure only whether the chatbot completed the conversation.
Measure whether the customer completed the task.
Identify Where Self-Service Is Failing
Improving chatbot performance begins with understanding the conversations that do not work.
Contact centre teams should review enquiries where customers repeatedly rephrase questions, abandon the interaction, request an agent or return through another channel.
These behaviours can reveal several different problems.
The chatbot may not understand the customer’s language. The relevant information may be missing from its knowledge base. It might understand the request but lack access to the system needed to complete it.
Alternatively, the enquiry may simply be unsuitable for automation.
Capacity provides AI-enabled customer support, knowledge management and workflow automation capabilities that can help organisations connect customer requests with relevant information and processes.
However, the technology itself is only part of the improvement process.
Managers need to understand why particular journeys fail and decide what should change.
Improve Intent Recognition
Customers do not always describe their problems using the terminology an organisation expects.
Someone asking about a delayed order might say their parcel has disappeared, delivery hasn’t happened or they cannot find their purchase.
A chatbot should recognise that these different expressions may refer to the same underlying need.
Modern conversational AI can help interpret less structured customer language, but performance still needs to be tested against real enquiries.
Contact centre teams should regularly examine conversations where the system failed to identify the customer’s intent or selected an inappropriate response.
Those examples can inform changes to intent definitions, instructions, knowledge content and conversational flows.
The most valuable training examples may be the questions the chatbot misunderstood rather than the ones it answered correctly.
Keep Customer Knowledge Accurate
A virtual agent can only provide reliable information if the knowledge supporting its responses is accurate and current.
Outdated delivery policies, conflicting refund instructions or incomplete product information can undermine customer trust quickly.
Knowledge ownership should therefore be clearly established.
Who updates information when a policy changes? How are outdated articles removed? What happens when the chatbot cannot find an authoritative answer?
Capacity combines AI-powered customer support with knowledge-management functionality, while Zoom Virtual Agent supports integrations with knowledge sources and business systems.
These capabilities can help make relevant information available during customer interactions.
But organisations still need a process for reviewing that information.
A more capable AI model does not make an inaccurate policy correct.
Move Beyond Answers to Completed Actions
Many customer enquiries involve a task rather than a request for information.
A customer may want to change an appointment, update account details, report a missing order or cancel a subscription.
Providing instructions can be helpful, but completing the task directly may create a better experience where appropriate.
This requires self-service technology to interact with CRM, order-management or other business systems.
Zoom Virtual Agent, for example, supports connected workflows and actions across customer-service systems, allowing suitable enquiries to progress beyond conversational responses.
Wavenet and Kerv provide contact centre and AI integration capabilities relevant to connecting automated interactions with wider business processes.
However, increased automation also requires appropriate controls.
Organisations need to establish which actions can be completed automatically, how customers are authenticated and when human approval is necessary.
A chatbot should never confuse being able to answer a request with being authorised to carry it out.
Make Escalation Part of the Customer Journey
Some contact centres treat escalation as a failure of self-service.
That can be counterproductive.
A chatbot that recognises a complex or sensitive enquiry and promptly transfers the customer to the right adviser may deliver a better outcome than one that continues attempting to resolve an issue beyond its capabilities.
The important question is whether escalation occurs at the appropriate time.
HGS combines human customer-service operations with AI and automation, illustrating how the two can work together.
Ciptex provides cloud contact centre and digital engagement technology, including applications where automated triage helps direct customers towards appropriate support.
Contact centre teams should establish clear escalation criteria.
These might include repeated failed attempts, requests for human assistance, complex complaints, sensitive circumstances or situations where an automated decision could create material risk.
A successful handover is not evidence that automation failed. It may be evidence that the system recognised its limits.
Stop Making Customers Repeat Themselves
Few things undermine a digital customer-service experience more quickly than being transferred to an agent and asked to explain everything again.
An effective handover should preserve relevant information from the automated interaction.
That might include the customer’s original enquiry, any identity verification already completed, the information provided, actions attempted and the reason the matter remains unresolved.
Zoom Contact Center supports AI-generated handover summaries, allowing advisers to receive context from preceding virtual-agent interactions.
Gnatta provides customer engagement technology designed to manage conversations and workflows across digital channels, while IP Integration supports cloud contact centre technology and system integration.
For buyers, the important test is whether the receiving adviser can continue the conversation rather than restart it.
A transfer should feel like progress.
Route Customers to the Right Expertise
A seamless handover involves more than transferring conversation history.
The customer also needs to reach somebody capable of resolving the problem.
Contact centres should examine whether routing can use information gathered during self-service to identify the appropriate team or adviser.
For example, a complex billing dispute may require specialist support rather than a general customer-service queue.
A technical enquiry may need somebody with access to particular systems or product knowledge.
Digital contact centre providers such as Kerv, Wavenet and IP Integration offer technology relevant to connecting customer interactions with routing and workflow capabilities.
The objective should be to use information collected during automation to make human assistance more effective.
Otherwise, customers may experience a technically successful transfer followed by another series of internal handovers.
Give Agents Better Information
Human advisers should benefit from the same improvements to knowledge and system integration that support self-service.
If customers receive one answer from a chatbot and a different answer from an adviser, confidence in both channels can deteriorate.
Organisations should therefore consider whether automated and human support draw from consistent, approved information sources.
AI-powered agent assistance can also help employees retrieve relevant knowledge and understand previous interactions.
Zoom, Capacity and HGS offer capabilities relevant to combining automation with human customer support.
The objective is to create a consistent service environment in which AI and human agents support one another.
Self-service should not operate as a separate, disconnected department.
Use Human Conversations to Improve Automation
Every escalation creates a potential learning opportunity.
If a human adviser resolves an enquiry that a chatbot could not handle, the organisation should examine what made the difference.
Did the adviser use information that was missing from the knowledge base? Access a system unavailable to the chatbot? Recognise a phrase the automation misunderstood?
Or did the situation genuinely require human judgement?
Zoom’s Virtual Agent 3.0 introduced capabilities designed to identify knowledge gaps and use validated outcomes from human-agent interactions to improve subsequent automated service.
This illustrates an important development in contact centre technology.
Human advisers should not simply inherit the problems automation cannot solve.
Their experience can help determine which problems automation should be able to solve next time.
Don’t Automate Every Enquiry
The availability of more sophisticated AI does not mean every customer interaction should be automated.
Some situations involve vulnerability, emotional distress, unusual circumstances or decisions with significant consequences.
Others may technically be possible to automate but create unnecessary friction for particular customers.
The UK Government’s Service Manual advises that digital services should provide appropriate assistance for people who lack the skills, confidence or access needed to complete tasks independently.
Although written primarily for public services, that principle provides a useful benchmark for wider customer experience.
Customers should not be forced through an unsuitable automated journey simply because the organisation wants to reduce contact volumes.
The best self-service strategy includes knowing when not to use self-service.
Review Accessibility and Customer Choice
Digital support should be usable by customers with different needs and abilities.
Chat interfaces need to be accessible, including appropriate support for keyboard navigation, screen readers and clear language.
Organisations should also consider whether customers can easily switch to another form of assistance.
The Web Content Accessibility Guidelines provide established principles for making digital interfaces more accessible.
Where a customer needs additional support, the transfer process should not create an unreasonable barrier.
Improving chatbot resolution rates should not involve making human support more difficult to reach.
That might improve containment figures while producing a worse customer experience.
Protect Customer Information
Self-service systems can process personal information, account details and records of customer interactions.
Where AI is involved, contact centres should understand how information is accessed, processed, retained and shared with technology providers.
The Information Commissioner’s Office provides guidance on applying data-protection principles to AI systems, including transparency, accountability and appropriate technical controls.
These requirements become especially important where a chatbot can retrieve account information or carry out transactions.
Customers may need to be authenticated before sensitive information is disclosed or particular actions are completed.
However, security controls should remain proportionate.
A customer asking a general question should not necessarily be required to provide unnecessary personal information.
Measure the Complete Customer Journey
Improving chatbot performance requires measures that distinguish customer outcomes from operational activity.
Useful indicators include confirmed self-service resolution, repeat contact, abandonment, escalation, customer satisfaction and the time required to reach a satisfactory outcome.
Organisations should also review why customers escalate and whether the receiving human agent can resolve the enquiry efficiently.
For example, a rising escalation rate might initially appear negative.
But if customers are reaching suitable advisers sooner, repeating themselves less often and achieving better outcomes, the overall experience may be improving.
Conversely, a higher containment rate may conceal customers abandoning unsuccessful conversations.
The most useful measurement framework should consider the journey from the customer’s original need through to its eventual resolution.
Questions to Ask Self-Service and Chat Providers
Contact centre leaders should consider:
- How does the platform distinguish containment from genuine resolution?
- Can failed or abandoned conversations be identified?
- How are misunderstood customer intents reviewed?
- Can knowledge gaps be detected and prioritised?
- How is approved information kept current?
- Can the chatbot complete actions in connected business systems?
- What authentication and authorisation controls are available?
- When and how can customers request human assistance?
- What information passes to the receiving agent?
- Can customers be routed to specialist teams?
- Can human-agent outcomes inform future chatbot improvements?
- How are sensitive enquiries and vulnerable customers handled?
- What accessibility provisions are supported?
- How are customer conversations and personal data protected?
- Can reporting follow the complete journey across automated and human interactions?
Frequently Asked Questions
What is a chatbot resolution rate?
A chatbot resolution rate measures the proportion of interactions in which a customer’s enquiry or task is successfully completed through automated support. The exact definition depends on how the organisation verifies resolution.
What is the difference between chatbot containment and resolution?
Containment generally measures interactions handled without human escalation. Resolution measures whether the customer’s underlying problem was satisfactorily addressed. An interaction can be contained without being resolved.
How can chatbot resolution rates be improved?
Organisations can improve results by analysing failed journeys, refining intent recognition, maintaining accurate knowledge, integrating business systems and making human escalation more effective.
When should a chatbot transfer a customer to an agent?
Escalation may be appropriate when the system cannot resolve the enquiry, the customer requests assistance or the situation requires specialist knowledge, human judgement or additional safeguards.
How can contact centres measure successful handovers?
Useful measures include repeat explanations, time to resolution, transfers between teams, customer satisfaction and whether the receiving adviser has the information needed to continue the conversation.
Product & Services Guide
The Contact Centre & Customer Services Summit brings together senior contact centre decision-makers and relevant solution providers through pre-arranged one-to-one meetings, offering an opportunity to explore web self-service, AI chatbots, live chat, customer experience technology and operational support. Participating suppliers include:-
Capacity
AI-powered customer-support platform offering virtual agents, knowledge management, helpdesk automation and workflow capabilities. Its solutions support automated answers, connected customer-service processes and human-agent assistance.
Website: https://capacity.com/
Ciptex
Customer engagement technology specialist delivering cloud contact centre, digital communications and automation solutions. Its capabilities include chat and customer-routing technology relevant to coordinating self-service and human support.
Website: https://ciptex.com/
ellio
Customer experience services provider combining operational support, advisory expertise and technology. Its capabilities are relevant to organisations seeking to improve service processes and manage complex customer interactions.
Website: https://www.elliocx.com/
Gnatta
Customer engagement technology provider supporting digital conversations, workflow management and contact centre operations across communication channels.
Website: https://gnatta.com/
HGS
Customer experience and business process management provider combining human support operations with conversational AI, automation and agent-assist technology.
Website: https://hgs.com/
IP Integration Ltd
Cloud contact centre and customer experience technology specialist providing communications solutions, systems integration and support for connected service workflows.
Website: https://ipintegration.com/
Kerv
Technology and customer experience provider offering cloud contact centre services, AI chatbots, integration and consultancy, including capabilities supporting Genesys Cloud environments.
Website: https://kerv.com/
Magellan
Outsourced customer-service provider offering live chat and broader contact centre support. Its services can help organisations provide human assistance alongside automated digital channels.
Website: https://www.magellan-solutions.com/
Wavenet
Managed technology and communications provider offering cloud contact centre solutions, AI-enabled self-service, automation and integrated customer experience capabilities.
Website: https://www.wavenet.co.uk/
Zoom
Communications and customer experience technology provider offering Zoom Contact Center and Zoom Virtual Agent. Its capabilities include conversational AI, workflow automation, live-agent escalation and contextual handover summaries.
Website: https://www.zoom.com/
Resolution Is the Real Measure of Automation
Self-service technology can help contact centres operate more efficiently, but its success should not be judged by how many customers avoid speaking to an adviser.
The more useful objective is to help customers resolve straightforward enquiries independently while ensuring more complex situations receive appropriate human support.
A practical improvement cycle is: understand → resolve → recognise limits → hand over → learn
Understand what customers need.
Resolve suitable enquiries through accurate information and connected workflows.
Recognise when automation cannot provide the appropriate outcome.
Transfer the customer without losing context.
Then learn from the interaction to make the next journey better.
The best chatbot isn’t the one that keeps the most customers away from agents. It’s the one that helps customers reach the right resolution with the least unnecessary effort.
The Contact Centre & Customer Services Summit brings together senior contact centre decision-makers and relevant solution providers through pre-arranged one-to-one meetings, offering an opportunity to explore web self-service, AI chatbots, live chat, customer experience technology and operational support.
Related Reading
This is the second article in our October Web Self-Service & Chat series.
Our first article, Web Self-Service & Live Chat: What Contact Centre Teams Should Compare, explored how buyers should assess knowledge bases, virtual agents, live chat, integrations, human escalation, security and measurement.
This follow-up focuses on improving operational performance after deployment, with practical guidance on reducing failed chatbot journeys, increasing genuine resolution and making human handovers more effective.
Sources
- Zoom – Virtual Agent 3.0: https://news.zoom.com/zoom-introduces-zoom-virtual-agent-3-0/
- Zoom – Managing Support Channels in Virtual Agent: https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0058249
- Zoom – Contact Center Agentic AI Expert Assist: https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0085180
- GOV.UK – Assisted Digital Support: An Introduction: https://www.gov.uk/service-manual/assisted-digital/
- GOV.UK – Encouraging People to Use Your Service Online: https://www.gov.uk/service-manual/helping-people-to-use-your-service/encouraging-people-to-use-your-digital-service
- ICO – Guidance on AI and Data Protection: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/
- W3C – Web Content Accessibility Guidelines: https://www.w3.org/WAI/standards-guidelines/wcag/
Image credit: https://unsplash.com/photos/blue-and-green-logo-guessing-game-s7AH5W-59m0
