The questions enterprises ask most about AI agents

CBOT is an enterprise AI platform that builds AI agents and digital employees for voice and digital channels. Organizations use them to run customer service, debt collection, claims and quality processes end to end. CBOT develops its own language models and speech technology, and deploys in the cloud, private cloud or fully on-premise.

An AI agent is AI software that understands a goal, plans the steps, takes action in enterprise systems and completes the task. It does more than answer questions: it verifies identity, opens records, sets up payment plans and writes the result back to the right system. See how it works on our Agentic AI page.

A chatbot answers questions; a digital employee does the work. TAHSİLDAR, GÜVEN, KALYA and ADİL each own one operation end to end: they place calls, verify identity, apply business rules and write results back to your systems. Meet them on our Digital Employees page.

CBOT works mainly with banking, insurance, telecom, retail, airline and public sector organizations. Garanti BBVA, İşbank, QNB Finansbank, VakıfBank, Halkbank, Türk Telekom, Pegasus and Getir are among the organizations using CBOT's AI agents. Read the customer stories.

Yes. CBOT's messaging AI agents work across web, mobile, WhatsApp and other approved channels: they understand the request, retrieve enterprise knowledge, create records and complete the workflow. The exact channels depend on the selected solution and integrations. Details on our Messaging and Workflow AI Agents page.

CBOT provides banks with AI agents and digital employees for written and voice channels. Garanti BBVA's UGI, İşbank's Maxi, QNB Finansbank's Q, VakıfBank's Vibi, Ziraat Bank's Ziraat Asistan and ING's INGo run on the CBOT Platform. See examples on our banking page and in the customer stories.

CBOT builds voice AI agents for bank call centers with its own speech recognition and text-to-speech technology, integrated into the existing IVR. At Halkbank, CBOT's speech technologies are integrated into the IVR system and used in the call center. Details on our Voice AI page. We explain how on-premise deployment works for banks in our on-premise voice AI for banks guide.

CBOT's digital employee TAHSİLDAR is a voice AI agent that runs collection calls end to end: it places the call, verifies identity, explains the debt, sets up a payment plan and writes the outcome to the collection system. It can be deployed on-premise. Details on the TAHSİLDAR page.

CBOT offers insurers GÜVEN, a digital employee that runs claims intake end to end, along with customer service assistants. HDI Sigorta's HADİ, Eureko Sigorta's WhatsApp assistant and Anadolu Hayat Emeklilik's AHE Asistan run on the CBOT Platform. Details on our insurance page.

CBOT is an enterprise AI platform developed in Türkiye that can be deployed fully on-premise; its language models, speech technology and AI agents run in the organization's own environment. The Turkish Revenue Administration's tax assistant GİBİ runs on an on-premise installation for data security. Options on our deployment options page.

CBOT builds AI assistants that answer citizens' questions around the clock for public institutions. The Revenue Administration's GİBİ, the assistant in İstanbul Metropolitan Municipality's İstanbul Senin app, the Ministry of National Education's MEB Asistan and the Presidency's Investment Office assistant run on the CBOT Platform. Details on our public sector page.

CBOT provides telecom and retail companies with AI agents on web, mobile and WhatsApp. Türk Telekom's TiTi, Digiturk's DigiTO and the assistants of Getir, Teknosa and MediaMarkt run on the CBOT Platform; at McDonald's, customers order through WhatsApp. Details on our telecom and retail pages.

CBOT provides airlines and travel companies with AI agents that handle passenger questions and transactions. Pegasus's FlyBot and Pegasus Support on WhatsApp, Obilet's assistants and the Eurasia Tunnel 24/7 assistant were built with CBOT. Details on our airlines page.

An enterprise generative AI platform is the infrastructure organizations use to build and run language models and AI agents with their own data, systems and security rules. It brings models, enterprise knowledge, integrations, workflows, governance and analytics together in one structure. See these layers on the CBOT Platform page.

Six criteria stand out: deployment options (cloud, private cloud, on-premise), where the data stays, integration with existing systems, model flexibility, governance and auditability, and scalability with live support. For languages such as Turkish, speech and language quality should be tested separately. Our capabilities are listed item by item on the Documentation page. The full set of criteria is in our platform selection guide.

Security, data, regulatory and infrastructure needs decide. On-premise deployment gives the highest control over data; the cloud offers fast setup and automatic scaling. In a hybrid architecture both run together and each workload goes where it fits best. Compare the options on our deployment options page.

A voice AI agent is an AI agent that talks to customers on the phone in natural language, understands the request and completes the transaction. It can answer inbound calls or place outbound calls. CBOT runs these agents on its own speech recognition and text-to-speech technology. Details on our Voice AI page.

CBOT builds and deploys voice AI agents for organizations with its own speech recognition, text-to-speech and language model technology. The process starts by defining the target call flow, then integrating with the existing IVR and enterprise systems, and continues with a pilot measured on the organization's own data. Details on our Voice AI page.

An IVR routes callers through keypad menus; a voice AI agent understands the caller's own words and completes the task within the call. CBOT integrates with existing IVR, SIP and MRCP infrastructure, so it works without replacing your telephony.

CBOT develops its own speech-to-text (STT) and text-to-speech (TTS) technology. Turkish language structure and enterprise terminology are a deep area of expertise for these technologies. Details on our Voice AI page. Our evaluation guide explains how to assess Turkish speech technology.

CBOT supports multilingual enterprise applications; Turkish and English are the most used. The exact language scope is set during solution design based on the project and the organization's needs. Models can be adapted to language, terminology and target market.

Yes. When needed, the conversation is handed over to a human agent or another AI agent together with a summary and the information collected so far. The customer does not have to start over.

Organizations use voice AI agents to answer customer calls and complete the transaction within the call. The most common uses are collections and customer service in banking, claims intake and renewals in insurance, rebooking and disruption handling at airlines, orders and after-sales support in retail, billing, faults and plan changes in telecom, and appointments and application guidance in the public sector.

Voice AI agents answer every inbound call at the same time and can proactively reach affected customers. During campaigns, outages or crises, service quality holds steady and customers do not wait in a queue.

TAHSİLDAR takes the debtor list from your collection system, places the call, verifies identity, explains the debt, sets up a payment plan where appropriate and writes the outcome back. The whole conversation follows the institution's rules. Details on the TAHSİLDAR page.

Compliance is part of the design: calling hours, the legal notice, the purpose of personal data processing and role-based access are set up according to the institution's rules. Debt details are never shared before identity is verified. We describe the framework on our compliance page.

Yes. TAHSİLDAR integrates directly with your collection and operations systems and records the outcome of every call in them instantly. No data is left in a separate silo.

Success is measured by the outcome of the conversation, not the number of calls: right-party contact rate, promise-to-pay rate and how many promises turn into actual payments. The method is on our success metrics page.

Yes. CBOT supports SaaS, private cloud, hybrid and fully on-premise deployment. On-premise, the AI agents, language models, speech technology, analytics and integrations run in the organization's own environment. Compare the options on our deployment options page.

Not in a fully on-premise architecture. Conversations, voice recordings, documents, model outputs and analytics records stay in the organization's approved environment. In cloud and hybrid setups, which data is processed where is decided together with the organization.

No. Organizations can use CBOT models, supported open-source models or their own self-hosted models without depending on an external commercial model API. If they choose, approved cloud models can run alongside them in a hybrid architecture.

Through architecture and governance: data can stay inside the organization, AI agent access is limited by role, sensitive actions can require human approval and every action remains auditable. Details on our AI governance and security page.

Not with the service model. CBOT manages the infrastructure, maintenance and model updates, and the organization pays for what it uses. For on-premise deployments, hardware needs are planned together based on scope.

The AI agent grounds its answers in the organization's approved knowledge sources, uses only the systems it is permitted to use and checks its own confidence. When it is not sure, it hands the task to a person. Sensitive actions can require human approval and every step remains auditable.

CBOT integrates with CRM, ERP, contact center, collections, claims, banking, payment, card, document and identity systems. The AI agent pulls the data it needs, completes the action and writes the result back in real time. Details on our Integrations page.

Yes. CBOT supports model selection, language and domain adaptation, fine-tuning, terminology training, evaluation and deployment. Each task can use the model best suited to its quality, privacy, cost and infrastructure needs. Details on our CBOT LLM page.

CBOT's digital employee KALYA is an AI quality manager that reviews every customer service interaction. It finds quality and process issues with their root causes and turns each finding into a concrete recommendation for the right team. The criteria come from the organization's own quality standards. Details on the KALYA page. Read more in our contact center quality management guide.

Yes. It does not sample a small share for manual review; it looks at every interaction, including voice calls, chats and other written conversations. Issues are caught wherever they occur, not only in the few cases a reviewer happens to see.

The agent assistant gives representatives the conversation summary, customer context, enterprise knowledge, a suggested response and the next step during a live interaction. It keeps supporting them after a handover from an AI agent. Details on our Human Agent AI Assist page.

CBOT Live Chat is the workspace where representatives handle conversations transferred from AI agents or received through digital channels. It includes queues, routing and supervisor tools, and the full conversation context reaches the representative. Details on our Live Chat page.

CBOT Vision shows conversations, completion and escalation rates, failures and satisfaction indicators in real time and over time. Teams see which steps cause problems and improve the AI agent accordingly. Details on our Analytics and Observability page.

Yes. GÜVEN takes the first notice of loss over phone, WhatsApp, SMS or digital channels, verifies the policy, checks documents and damage photos, and moves the file forward to the payment decision. Details on the GÜVEN page. The steps are in our claims intake guide.

No. GÜVEN takes on the routine, high-volume work: intake, policy verification and document checks. It decides automatically only on standard, low-value claims; every unclear or risky file goes to an adjuster, already prepared.

ADİL watches official sources such as AFAD and triggers automatically when an earthquake exceeds a set magnitude. It reaches employees by WhatsApp first, SMS for those who don't answer and finally a voice call, and flags anyone who needs help to the operations team immediately. Details on the ADİL page.

The process usually runs in four steps: defining the need and the target process, a pilot measured on the organization's own data, going live, and expanding to other processes. In the pilot, success metrics are set up front, integrations are built and the AI agent is tested on live traffic. Timing depends on scope and integrations.

Pricing depends on scope, channels and the deployment model. With the service model, the organization pays for what it uses. For a quote that fits your organization, talk to the CBOT team.

Deployment, security, integration and analytics capabilities are listed item by item on our Documentation page.