Skip to content
EN
Demo
Platform
CallsQueues and routingIVR (phone menu)Smart transferAuto dialerBrowser-based softphoneClick-to-call
Messaging and customer dataUnified inboxForm builderCustomer portalCustomer success platform
ManagementHistorical reportingReal-time monitoringShift scheduling
Deployment and securityIn our cloudOn your serversHybrid deploymentData security
Platform →
Solutions
IndustriesRetailReal estateBanking and insuranceTravelHealthcareEducation
Customer serviceFAQ self-serviceOrder status without an agentAfter-hours coveragePeak days without overloadConversation quality assurance
SalesLead qualificationAppointment bookingRepeat salesLoyalty program
Outbound communicationPayment and overdue remindersNotifications and confirmationsPost-interaction survey
Who it's forBy company sizeBy role
Solutions →
AI agents
For customersVoice AI consultantAI consultant in chat
For employeesAI assistant for calls and salesAssistant for work questionsNew hire onboarding
TechnologyKnowledge baseSpeech analyticsAI connectorModelsData security
ServicesAI agent trainingMachine learning
AI agents →
Integrations
CRM systemsSalesforceHubSpotDynamics CRMSAP CXZoho CRMPipedrive
ERP systemsSAPNetSuiteBusiness CentralSageOdoo
Help desk and team chatZendeskFreshdeskServiceNowSlackMattermost
Messaging and socialWhatsAppViberInstagramFacebookSMSLive chat
Telephony, API, and AIAsteriskAMI, AGI, and ARIKamailioFreeSWITCHAPI and protocolsMCP connectors
Integrations →
Pricing
Company
ResourcesBlogCalculatorsComparisonsLibraryGlossary
ServicesOpen SourceImplementationDevelopmentSupportTraining
ContactsContact usAboutFacts and figures
Company →
Language: English
Demo

AI agents

AI agent training for contact centers

An off-the-shelf model doesn't know your prices, your policies, or how your team resolves customer issues. Without training, an AI agent speaks in generalities, gets details wrong, and hands off to a person more often. CONTACT.WORK trains AI agents as a service. Afterward, the AI consultant resolves more inquiries without human help, and the AI assistant suggests replies to employees that follow the way your company does things. We test every release against your experts' answers, and nothing worse than the previous version goes live.

Get a quotePricing
2training methodsWith or without fine-tuning
Reference answersfrom your expertsBased on real inquiries
Everyrelease is testedNothing worse goes live
New modelsbenchmarked as they come outThe best-performing one stays

Tools

What goes into training

You decide where the AI agent answers by itself and where it brings in an employee.

Answer breakdownSee which document the AI agent found, what data it pulled from the CRM, and why it answered the way it did.
Errors after launchAn expert reviews every mistake the AI agent makes in production, and each case is added to the reference answers.
Company vocabularyThe AI agent understands your plan names, internal abbreviations, and terms as well as your employees do.
Voice checksFor voice AI agents, we check that prices, addresses, names, and numbers are pronounced correctly.
Conversation rulesWe set how the AI agent greets customers, what it asks first, and what it never says – the way your team does it.
Voice and textVoice and chat AI consultants, as well as employee AI assistants, are trained on a single set of reference answers.

Methods

Two ways to train an AI agent

We choose the method based on the task. The first is usually enough; the second is for when the model needs to adopt your style or run entirely inside your company.

Knowledge base, rules, and reference answers

The model stays the same, and the AI agent gets everything it needs to answer: the knowledge base, step-by-step procedures for common situations, and CRM data. Your best employees pick the correct answers to real inquiries, and we measure quality against them.

  • Knowledge base built from documents
  • Step-by-step procedures
  • Reference answers from experts
Method 1 of 2

The methods work together: a fine-tuned model also uses the knowledge base and goes through the same reference-answer testing.

How it works

How we verify that the AI agent got better

Reference answers

Expert answers set the quality bar

For each inquiry, your top specialist marks the correct answer, or notes that the AI agent should hand the question to an employee. The set of reference answers grows after every review.

  • Correct answer
  • When to hand off to a person
  • The set keeps growing

Release testing

We release only improvements

Before release, the AI agent runs through every reference question, and we compare the results with the previous version: what got fixed, what broke, and why. If important answers got worse, the release is held back and the errors go into review.

  • Every reference answer in every test
  • Comparison with the previous release
  • Worse means no release

Costs

Model costs

Training uses both paid and free models: free ones where possible, paid ones where higher quality is needed. That way, training doesn't eat up your model budget.

  • Free where possible
  • Paid where needed
  • Predictable costs
Less load on your linesA trained AI agent handles routine questions, and only complex inquiries reach your staff.
No customers lost to mistakesA wrong price or the wrong policy in an answer can cost you a customer. Training makes these errors much rarer.
Updates without regressionsAfter an update, the AI agent doesn't start failing where it used to answer correctly. We check this before every release.
First version in a dayOnce your materials are gathered, the first version on a cloud model is up in a day and improves with every release.

Launch

How AI agent training works

  1. 01GoalsWe pick the topics where the AI agent will take the most work off your staff, and the knowledge sources to use.
  2. 02DataWe gather documents, chats, and calls, and select what is relevant.
  3. 03Reference answersExperts choose the correct answers, which become the test set.
  4. 04First versionWe connect the knowledge base, set conversation rules, choose a model, and test the AI agent against the reference answers.
  5. 05RefinementWe review errors, prepare new releases, and fine-tune the model if needed.
  6. 06LaunchThe AI agent goes live with customers, and new inquiries become reference answers.

Requirements

What training requires

Data
Documents, procedures, chat history, or call recordings
Expert
Your top expert on these inquiry topics
Model
Cloud or self-hosted – your choice
Server
A self-hosted model and fine-tuning require a GPU
Access
The CRM and systems the AI agent pulls data from
Cost
Training and fine-tuning are billed in support plan hours

Preparation

What to prepare for training

What should the AI agent handle?Your most frequent questions, where the benefit will be most noticeable.
Where does your knowledge live?The latest versions of procedures, price lists, and instructions.
Do you have conversation history?Chats and calls covering at least a few months.
Who will be the expert?The employee who knows the inquiry topics best.
What data is restricted?What cannot be sent to a cloud model, and why.
Where will the model run?A cloud model or your own hardware with a GPU.

Q&A

Frequently asked questions

Why train an AI agent if the model is already smart?

The model knows language and general facts, but not your prices, plans, or how you handle returns and complaints. An untrained AI agent gives answers that are "generally right," but customers need an exact answer based on your company's rules – otherwise they call back or leave.

How do we measure the impact of training?

We compare before and after: the share of inquiries the AI agent resolves without help, how many conversations it hands to employees, and how many customers contact you again.

How is training an AI agent different from fine-tuning a model?

Training leaves the model unchanged: the AI agent gets documents, rules, and examples, and quality is measured against reference answers. Fine-tuning changes the model itself so it absorbs your style and terminology. The first is usually enough; the second is for narrow tasks and small models installed on your servers.

How much data do you need?

To start, documents and labeled inquiries on your most frequent topics are enough. Fine-tuning needs significantly more conversations – we will tell you exactly how many after reviewing your data.

Who marks the correct answers?

Your expert – the employee colleagues go to with tough questions. We prepare the inquiries and explain what to mark, and we review disputed cases together.

How long does training take?

With a cloud model, the first version of the AI agent can go live in a day once the materials are gathered. Fine-tuning takes longer: time goes into preparing conversations, the training itself, and testing.

Will our data end up in someone else's model?

If your data cannot leave your company, we train and run the AI agent on an open-source model inside your network, and conversations stay with you. We choose a cloud model only if it fits your data storage policies.

Does the AI agent keep learning after launch?

Inquiries where it made a mistake or handed off to a person go to your expert and are added to the reference answers, and we test the next release against the expanded set.

What if a better model comes out?

Every three months, we test new models against your reference answers. If one performs better, we move the AI agent to it after the same testing.

When does the AI agent learn about policy changes?

The AI agent starts applying a changed policy after retraining. Small changes take minutes, large ones a few hours.

Who owns the fine-tuned model and reference answers?

The customer who paid for the training hours: the fine-tuned model, the reference answers, and the AI agent's materials belong to them.

Do you train AI agents on other platforms?

No. We train only AI agents on our own platform, where preparation, release testing, and customer conversations all happen in one system.

How much does AI agent training cost?

Training and fine-tuning are billed in support plan hours, and we give you an estimate before work starts. After launch, the AI agent is billed by usage: voice per minute and text per conversation. On your own servers, minutes are not billed.

Pricing →

Glossary

Terms on this page

Model fine-tuning
An existing model is trained further on company materials so it speaks the company's language. Learn more →
Cloud deployment (cloud)
The platform runs on the vendor's servers, and agents connect over the internet. Learn more →
Knowledge base
All answers, policies, and prices in one place – for agents and AI. Learn more →
CRM – customer relationship management
Where a company keeps its customers, deals, and full communication history. Learn more →
On-premises deployment
The system runs on the company's own servers, not in the vendor's cloud. Learn more →
Personal data
Anything that points to a specific person: a name, a number, a voice. Learn more →
Call recording
Calls are saved so they can be replayed and disputes settled. Learn more →

More

More on AI in the contact center

Machine learning More connects and more deals from the same numbers – proven on your data. Learn more Voice AI consultant Like talking to an experienced rep – with no hold time, day or night. Learn more AI consultant in chat Customers write whenever it suits them and get an answer right away – not the next morning. Learn more AI assistant for calls and sales In-call suggestions, customer summaries, and a daily call plan. Learn more

Blog

Fresh from the blog

All articles →
Services · 7 min read Custom development of your contact center platform for your processes When contact center settings are not enough: integrations, modules, AI workflows. How to tell development from configuration and estimate the payback. Read the article → Management · 9 min read How many agents does your contact center need? Erlang C in plain English Calculate contact center staffing with Erlang C: workload, service level, shrinkage, and headcount by shift – all in one worked example. Read the article → Integrations · 7 min read CRM and phone system integration: what it does and how it works How to connect your phone system to your CRM: calls and recordings in the record, screen pop, click-to-call. What to decide first and how to check it works. Read the article →

Updated: