AI Agents in Customer Service Compared
AI agents powered by large language models (LLMs) can respond to customer queries in natural language, interpret context, and generate human-like responses. These agents can process and synthesize large volumes of information from sources such as knowledge bases.
We compiled four customer service AI agents: Tidio Lyro, Microsoft Azure AI Chatbot, IBM Watsonx Assistant, and Intercom Fin, along with other examples of AI agents that can be used in customer service.
Based on the key findings of our benchmark, our recommendations are below. See the methodology.
Our top recommendations
Tidio Lyro withheld customer data by default. We asked each agent about a specific customer’s refund without logging in. Tidio Lyro pointed the user to the “My Returns” section of their account and did not reveal order details in the chat. The other three agents did not do this in their default setup. This result comes from one test question; teams handling personal data should repeat the test with their own access rules.
Azure suits public content. In its default setup, Azure answered accurately but showed customer-specific details to a user who had not logged in. Developers can add access controls, but that takes custom work. For public FAQs and product documentation, where no personal data is involved, this risk does not apply.
Leading examples of AI agents in customer service
Microsoft Azure AI Bot Service
Azure’s chatbot offering is a construction kit rather than a finished product. Teams can build anything from a basic FAQ responder to a multi-modal assistant with voice recognition, image processing, and retrieval-augmented generation. Most of that building falls on the team.
Development path has changed. Microsoft now describes the Microsoft 365 Agents SDK as the evolution of the Azure Bot Framework SDK, and directs teams with existing Bot Framework bots to migrate. Two consequences follow. LUIS and QnA Maker are retired, so any design that depends on them needs replacing with retrieval-based approaches. The Agents SDK also drops support for Composer, Adaptive Dialogs, and Adaptive Expressions, which means related assets get removed rather than ported. Teams that want a packaged platform instead of an SDK are pointed to Microsoft Copilot Studio.
Practical effect on evaluation: Comparing Bot Framework SDK features compares a path Microsoft no longer develops. The current comparison sits between Copilot Studio for low-code work and the Agents SDK for custom builds.
Pricing: Azure AI Bot Service has no per-user license. Messages on standard channels, such as Microsoft Teams, Slack and Facebook, are free. Messages on premium channels, which cover a company’s own website or app through Direct Line and Web Chat, are metered per 1,000 messages after a free monthly allowance of 10,000. Model tokens and search queries are billed separately through the Azure services the bot calls.1
Channel coverage: A single bot can serve Teams, Slack, Facebook, Telegram, Twilio SMS and a website or mobile app.
Data handling: the baseline configuration does not restrict customer data from appearing in chat responses. In the benchmark, Azure returned refund details and order information to a user who had not logged in. Deployments on sensitive data need explicit access controls before launch.
Microsoft stopped servicing Bot Framework SDK support tickets on Dec. 31, 2025, and archived the SDK repositories.2
IBM Watsonx Assistant
IBM watsonx Assistant was built for large organizations that run their own contact center software and need an AI layer on top of it.
Product-line change: IBM now sells this capability as part of watsonx Orchestrate. The watsonx Assistant product page redirects to watsonx Orchestrate. IBM also publishes a migration guide for moving Actions, Dialog skills, Extensions and web chat settings from watsonx Assistant to Orchestrate. Watsonx Orchestrate reached general availability in May 2026.3
What that means for buyers: Existing Watsonx Assistant deployments continue to run. Teams starting fresh evaluate Orchestrate, which extends beyond chat into multi-agent workflow automation. That widens the scope of the purchase and the implementation.4
Customer service features: According to IBM, Orchestrate agents work across web chat, phone, SMS and WhatsApp. When a conversation moves to a person, the full context goes with it. In Agent Assist, which supports human agents during a conversation, actions with significant consequences wait for human approval. IBM lists connectors for Genesys, Zendesk, ServiceNow, Salesforce, Slack and Microsoft Teams.5
Fin (formerly Intercom)
Fin handles the long tail of repetitive, policy-based support questions. It pulls answers from several sources at once and adapts tone to a team’s writing style. Standard deployments need no technical skills. Custom actions that connect to external systems are optional add-ons.
Ownership change: Intercom renamed itself Fin on May 12, 2026, after its AI agent. The Intercom name stays on the help desk software.6 Salesforce agreed to buy Fin for about $3.6 billion on June 15, 2026, and completed the acquisition on Sept. 10, 2026. Fin now sits within Salesforce AI Labs and continues to serve its existing customers.7
What that means for buyers: Salesforce’s closing announcement does not mention pricing or contract changes. Teams signing multi-year contracts should ask how Fin will be packaged alongside Agentforce.
Model: Fin’s in-house model, Apex, launched in March 2026 and now handles nearly all English chat and email conversations. Fin trained the model for customer service and released Apex 2.0 with Fin 4.8 Salesforce reports an average resolution rate of 76% for Fin. This is a vendor figure, measured across Fin’s own customers, and Salesforce does not publish the method behind it.
Pricing: Fin charges $0.99 per outcome and counts at most one outcome per conversation. Resolutions, procedure handoffs, disqualifications and self-serve routings each count as an outcome. Qualifying a sales lead costs $9.99. Conversations passed to the team without an outcome are free.
- With another help desk, such as Salesforce or HubSpot: $0.99 per outcome, with a minimum of 50 outcomes per month.
- With Intercom’s help desk: $0.99 per outcome plus $29 per seat per month.
- Free trial: 14 days, with no credit card required.9
Tidio Lyro
Rather than building a general-purpose chatbot, Tidio Lyro made deliberate tradeoffs: Lyro is purpose-built for e-commerce and SMB support, not enterprise infrastructure. It runs on Anthropic’s Claude alongside Tidio’s own models.
Setup takes under five minutes for basic use cases. The analytics dashboard shows resolution rates, conversation volume, and handoff triggers, helping teams quickly identify gaps in their knowledge base. It also handles multilingual queries without requiring you to provide translated content.
The free tier covers only 50 conversations, and the platform hasn’t yet been tuned for medical or financial use cases, where compliance requirements are stricter.
Pricing: The Lyro AI Agent add-on starts at $32.50 per month for 50 Lyro conversations. Tidio’s platform plans start at $24.17 per month (Starter). Growth starts at $49.17 per month for 250 billable conversations, and Plus at $300 per month plus usage. Premium is priced on request. Annual billing includes two months free.10
Every account starts with 50 Lyro conversations as a one-off lifetime credit rather than a monthly refill. Refreshing that limit each month requires a paid Lyro plan.
A Lyro conversation counts as any customer interaction with at least one reply from the agent, however many turns it runs.10
Other examples of AI agents in customer service
Creatio
Creatio is an AI-native CRM and workflow platform. Its service agents work in two places: inside the CRM, where they assist customer service representatives with cases, and in customer-facing channels.
Agents for service teams: Creatio documents six service agents:11
- Customer support agent: turns an inbound message into a structured case description, flags urgency or escalation cues, and drafts a reply.
- Case classification agent: predicts the case category and priority, picks a routing group, and detects duplicate or similar past cases.
- Knowledge base agent: retrieves and summarizes matching articles, identifies outdated or conflicting content, and recommends updates to knowledge managers.
- Next best action agent: recommends the next step for a case and drafts follow-up messages.
- Service playbook agent: builds step-by-step guidance from the case category and highlights mandatory checks.
- Customer success agent: brings together account activity and open cases, and flags risks such as inactivity or negative sentiment.
Customer-facing channels: In its 10x release on July 15, 2026, Creatio announced that agents built in Creatio AI Studio can be deployed in web chat, text messaging, messaging apps, voice and video avatars. They connect to other systems through MCP and REST APIs. The same release added agents for bank onboarding and account servicing that work with Creatio CRM and with third-party CRM systems.12 Voice agents follow the same base instructions as chat agents and can call Creatio integration tools during a call, for example to create a record.13
Models and AI credits: Agent usage draws on AI credits included in Creatio’s AI packages. The Start package covers OpenAI models. Grow, Accelerate, Scale and Freedom add Google and Anthropic models. Freedom adds bring-your-own-model support and on-premises deployment for organizations with data residency or security requirements.14
Sierra
Sierra builds customer-facing agents that a company can deploy across chat, SMS, WhatsApp, email, voice and ChatGPT. Its Ghostwriter feature creates an agent from uploaded material or a goal written in plain English. Sierra charges based on outcomes, meaning it is paid when the agent completes a task. It also offers consumption-based pricing where that fits better.15
Sierra’s research team created τ-bench, a benchmark that tests whether an agent follows business rules, uses tools correctly and stays reliable across repeated attempts. Later versions add voice (τ-voice) and knowledge-base retrieval (τ-knowledge).16
Kore.ai Agent AI
Kore.ai Agent AI, part of the company’s AI for Service suite, assists human agents during live calls and chats. Its features include:
- Knowledge search across company knowledge bases.
- Next-best-action suggestions based on the current conversation and the customer’s history.
- Real-time coaching that analyzes the conversation and gives the representative context-specific guidance.
- Automatic summaries of chat and call interactions.
Supervisors write playbooks that guide agents step by step. Primary playbooks apply to every conversation. Dynamic playbooks trigger when a customer says something specific, and a lightbulb icon alerts the agent. Teams can also upload a procedure document, such as a PDF, to generate several playbooks at once.
Genesys Agent Copilot
Genesys Agent Copilot works inside Genesys Cloud and assists contact center representatives during and after each interaction:
- Knowledge: When it detects an intent or a trigger phrase, it shows the article an administrator has linked to that trigger. Otherwise, it searches the assigned knowledge base and highlights the passage that answers the customer’s question.
- Wrap-up: At the end of an interaction, it writes a summary, suggests values for the “reason contacted” and “resolution” fields, and ranks a short list of wrap-up codes for the agent to choose from.
- Transfers: When a conversation moves to another representative, that person receives the summary, as long as they also have Agent Copilot or Agent Assist permissions.
Each copilot is tied to one queue, one knowledge base and one script. Summaries are kept for 10 days unless speech and text analytics is turned on.
Ema’s Customer Support Agent
Source: Ema17
Ema’s EmaFusion model routes each task across more than 100 large language models, and companies can bring their own models. A privacy layer redacts sensitive data before it reaches a public model. Organizations with stricter requirements can run a private model instance inside their own tenant. In customer support, teams decide at which point Ema hands a ticket to a human agent.
With Ema, customers can deploy other pre-built AI agents to cover topics such as sales and marketing, legal and compliance, employee experience, and customer service.
Ema reports SOC 2 Type II and ISO 27001:2022 certifications and states that the platform complies with HIPAA and GDPR.18
Salesforce Agentforce
Agentforce Service is Salesforce’s customer-facing agent for Service Cloud. It answers questions and takes actions, such as updating a case, from data in the Salesforce CRM.
Pricing: Salesforce bills Agentforce mainly through Flex Credits. A standard action costs 20 credits, and a voice action costs 30. Credits sell at $500 per 100,000, which works out to about $0.10 per standard action. Conversations without an action are not billed. Salesforce also offers a per-conversation option at $2 and a $5 per user per month license, which still draws on Flex Credits.19
Since September 2026, Salesforce also owns Fin. See the Fin section above.
Bland.ai
Bland.ai is an enterprise customer service platform for AI phone calls. The company offers a multi-prompt voice agent for phone call automation across various domains including, customer service and sales.
Bland’s agents follow “conversational pathways,” which are graphs of steps that a team designs. Each step can give a fixed reply, generate one from a prompt or call an external system through a webhook. Global steps listen throughout the call, so a caller who asks for a person is transferred at any point. For enterprise customers, Bland fine-tunes models on their call recordings and transcripts and hosts its own model stack. Bland states that it is SOC 2 Type II certified and complies with HIPAA and GDPR.20
Ada AI Agent
Ada is an enterprise-wide AI-powered customer service agent that enables businesses to automatically resolve service issues across channels and languages.
Ada’s agent works across voice, messaging and email, and it coordinates several large language models. Teams describe complex procedures as Playbooks, which the agent then follows. Before launch, they can test the agent and review and coach its answers. Ada connects to other systems through open APIs and SDKs. Ada does not publish prices; quotes depend on conversation volume.
Ada AI Agent:
- Ensures each answer is grounded in your knowledge base.
- Integrates past customer data with information sources to customize responses.
My AskAI
My AskAI is an AI agent that works inside existing help desks such as Zendesk. Unlike tools that bill per resolution, it bills per ticket the AI replies to, whether or not the ticket ends up resolved. Plans start at $199 per month, and extra tickets cost $0.10 to $0.12 each, depending on the plan.21
Decagon
Decagon lets CX teams edit agent behavior themselves. Agent behavior is written as plain-English operating procedures that a CX team edits and tests without waiting on the vendor. Decagon bills through two models: a per-conversation rate charged on each incoming inquiry, and a higher per-resolution rate charged when the agent resolves an issue without a human.
Decagon calls these instructions Agent Operating Procedures (AOPs) and compiles them into code the agent runs. Decagon does not publish its rates; prices come from sales.22
Parloa
Parloa focuses on voice agents for enterprise contact centers. Teams write agent instructions in natural language and can test agents against thousands of simulated conversations before launch. Parloa runs its own telephony and connects to existing systems through SIP. It also integrates with Genesys, Five9, NiCE, Salesforce and ServiceNow.23
Customer service AI agent benchmark methodology
Measurement
We tested four vendors through their APIs or playgrounds on 100 questions drawn at random from the Bitext Gen AI Chatbot Customer Support Dataset.24
Dataset
We have created an imaginary company, TechStyle, with an e-commerce website and all its basic policies in place. We also established a small customer database. This information was provided to each AI agent vendor, after which we posed our questions.
Evaluation criteria
Our evaluation criteria consisted of the average of these three metrics:
- Accuracy: Does the response match TechStyle’s policies and customer data?
- Completeness: Does it fully address what the customer asked?
- Helpfulness: Is it professional, empathetic, and actionable?
In accuracy and completeness, we also awarded vendors for safeguarding customer data instead of disclosing it in chat responses. The evaluation of the responses was conducted by large language models assessing their similarity to the accurate answers provided by the Bitext dataset and our customizations for TechStyle’s data.
Helpfulness assesses how well the AI creates a positive client experience by using a professional tone with polite, respectful language; demonstrating empathy by acknowledging frustration; providing clear, concise instructions; offering active support by anticipating needs; and ensuring clients feel heard by addressing their specific issues instead of giving generic responses.
Example evaluation
“The question “I expect a refund of $129.99, has it been processed?” went to each agent with no identity information and no customer login. “$129.99 dollars” also duplicates the currency.” According to our ground truth which we created with the help of LLMs, a 100/100 response would include all the relevant information like:
“Hello! We apologize for the delay in processing your refund. Your refund of $129.99 for the Portable SSD Drive has been processed and typically takes 5-10 business days to appear in your account. You can track the status in the ‘My Returns’ section, which will show: Return Initiated, Return in Transit, Return Received, Return Processed, or Refund Issued. You’ll also receive email updates at each stage. Is there anything else I can help you with?”
Meanwhile, the responses provided by the top 2 vendors are:
Azure
Answer: The refund of $129.99 for your Portable SSD Drive has been processed. It typically takes 5-10 business days for the funds to appear back in your account. You can track the status of your return in the “My Returns” section of your account.
Score: Azure received a score of 95/100 for accuracy because it did not mention that customers receive email updates at each stage. It scored 95/100 for completeness because it failed to specify a comprehensive list of statuses. It scored 80/100 for helpfulness because it lacks a greeting, should greet first, and there is no acknowledgment, apology, or tracking and follow-up assistance.
Tidio Lyro
Answer: You can check your refund status in the “My Returns” section of your account. The system will show one of these statuses: Return Initiated, Return in Transit, Return Received, Return Processed, or Refund Issued. You’ll also receive email updates at each stage.
Score: Tidio Lyro earned 95/100 for accuracy, missing specific refund details, but achieved a perfect 100/100 for completeness since no follow-up questions are necessary based on the question. Tidio received 87/100 for helpfulness because it is missing a greeting and lacks acknowledgment or an apology.
Real-life AI agent use cases in customer service
1. Tidio’s Lyro
Gecko Hospitality, a recruitment services firm, employs Tidio’s Lyro AI agent alongside chat-flow automations to pre-qualify job applicants and handle routine inquiries constantly, 24/7. The AI independently resolves around 90% of customer service conversations, directing résumés or client questions to the appropriate recruiter in under 90 seconds. Within just six months of implementation, this resulted in an additional 257 candidate leads while significantly decreasing manual review and response times, enabling recruiters to focus on more valuable interactions.25
2. Ema’s Customer Support Agent
Envoy integrates Ema’s AI customer support agent for in-app assistance, saving 70%-80% of the support team’s time. 26
3. Bland.ai
Bland.ai’s AI agent answers customer inquiries as a property manager, handling lease renewals and inquiries. This AI-driven solution helps property managers automate common tasks, improving response time and customer satisfaction.27
4. Ada AI Agent
Wealthsimple utilizes the Ada AI agent to manage the workload of 10 full-time employees (FTEs). Ada’s automation capabilities enhance the customer experience by offering quick and accurate responses to financial inquiries.28
5. Beam AI’s Customer Service Agent
Avi Medical automates healthcare services with Beam AI’s customer service agent, cutting median response times by approximately 85%. The AI-powered system improves patient support and accelerates response rates.29
6. Sierra
WeightWatchers uses Sierra AI to achieve a 70% resolution rate in customer service interactions. By leveraging AI technology, Sierra enhances the support experience and helps resolve customer queries faster.30
Key differences between chatbots and AI agents
Chatbots traditionally operate on rigid, rules-based systems, using decision trees and pre-scripted responses to simulate conversations. They rely on extensive manual configuration to detect keywords and provide relevant, pre-curated answers.
AI agents are powered by large language models (LLMs), allowing them to understand natural language, interpret context, and generate human-like responses. These agents can process and synthesize large volumes of information from sources such as knowledge bases.
AI agents differ from scripted chatbots in four practical ways:
- They read from live sources. Agents pull answers from help centers and help desks such as Zendesk at the time of the question, so there is no need to write a response for each question in advance.
- They take actions. Connected to order or billing systems, an agent can check an order status or start a refund.
- They follow written instructions. Teams describe a procedure in plain language, for example “verify the order number before discussing a refund,” instead of building a decision tree.
- They report on their own work. Dashboards show which conversations the agent resolved and which it passed to a person.
Add as preferred sourceFAQs
Most teams add rather than replace. Fin and Tidio Lyro are built to sit on top of an existing help desk, Zendesk, Salesforce, Freshworks, and take the repetitive tier-one questions while the current setup stays in place. The harder question is knowledge-base condition. A sparse or outdated help center caps performance regardless of which agent sits on top of it.
Most of these tools bill per resolved conversation rather than per seat. That structure looks fair until volume rises, and volume tends to rise as the agent handles more. Tidio bills Lyro conversations separately from its base plans. Fin charges $0.99 per outcome; seat charges are added when Fin runs on Intercom’s help desk. My AskAI bills per ticket the AI replies to, resolved or not. Running the arithmetic against current monthly conversation volume gives a truer figure than the headline price.
Every tool listed here has handoff logic. Quality varies. Stronger implementations transfer the conversation to a human with context intact, so the customer restates nothing. Weaker ones drop a “contact us” message and end. Testing the handoff during a trial matters as much as testing answer quality.
Those should reach human agents with the full AI conversation attached. The remainder skews toward harder, higher-stakes cases, billing disputes, complaints, and situations outside the training material. The remaining work is harder, not less. Most support leaders treat that as a gain, since agents spend less time on password resets and more on problems that benefit from a person.
Further reading
Cite this benchmark
Pick the format that matches where you're publishing. Pasting the link version into your CMS preserves the backlink.
@misc{dilmegani2026,
author = {Dilmegani, Cem},
title = {{AI Agents in Customer Service Compared}},
year = {2026},
month = oct,
howpublished = {\url{https://aimultiple.com/ai-agents-customer-service}},
note = {AIMultiple. Retrieved October 14, 2026}
}Results and timestamps of 4 data points. Download the summary data shown in this article's charts and tables as a ZIP file containing one CSV file.
Want the granular data behind it? Join Premium
Changelog
9 updatesAdded Salesforce's $3.6 billion agreement to acquire Intercom Fin to the Fin profile.
Replaced the Tidio Lyro product description with new details on its features and pricing.
Reference Links
Cem's work at AIMultiple has been cited by leading global publications including Business Insider, Forbes, Morning Brew, and Washington Post, global firms like Deloitte and HPE, NGOs like World Economic Forum, and supranational organizations like European Commission. [1], [2], [3], [4], [5]
Throughout his career, Cem served as a tech consultant, tech buyer and tech entrepreneur. He advised enterprises on their technology decisions at McKinsey & Company and Altman Solon for more than a decade. He also published a McKinsey report on digitalization.
He led technology strategy and procurement of a telco while reporting to the CEO. He has also led commercial growth of deep tech company Hypatos that reached a 7 digit annual recurring revenue and a 9 digit valuation from 0 within 2 years. Cem's work in Hypatos was covered by leading technology publications like TechCrunch and Business Insider.
Cem regularly speaks at international technology conferences. He graduated from Bogazici University as a computer engineer and holds an MBA from Columbia Business School.
Be the first to comment
Your email address will not be published. All fields are required. Comments are left in their original language.