Artificial intelligence (AI) is entering a new phase as businesses move beyond chatbots towards systems capable of carrying out complex tasks with limited human intervention. Known as agentic AI, these systems are designed not only to generate responses but also to plan, make decisions and execute actions across multiple software applications. A small but growing number of Bangladeshi companies are beginning to develop products based on the technology.
Unlike conventional AI assistants, which typically answer questions or generate content in response to prompts, agentic AI is designed to work towards a defined objective. It can break a goal into smaller tasks, determine the sequence of actions required and interact with external software through application programming interfaces (APIs) to complete those tasks.
For example, while a traditional AI assistant might recommend available flights, an agentic AI system could search for suitable options, compare fares, complete the booking after receiving user approval and send the confirmation automatically.
The technology is also attracting growing interest from businesses worldwide. Global research firm Gartner forecasts that by 2028, agentic AI will make autonomous decisions in around 15% of day-to-day work decisions, compared with virtually none in 2024. The research firm also expects one-third of enterprise software applications to include agentic AI by that year.
AI takes on customer conversations
Although adoption remains at an early stage in Bangladesh, local technology companies have begun developing products that apply agentic AI to business operations, customer engagement and video intelligence.
One example is FastCom AI, a platform designed to manage customer interactions from start to finish rather than simply responding to individual questions. According to the company, it combines conversational AI, business knowledge retrieval and task execution, allowing businesses to answer enquiries, process requests and continue conversations across multiple communication channels.
Yaseen Nur, founder and chief executive of Ascend AI, the parent company of FastCom AI, said the company’s agentic AI platform is designed to automate customer support, sales and follow-up interactions across messaging platforms. Nur said the system can answer routine customer enquiries, retrieve business information and carry out tasks such as updating records without requiring constant human intervention.
“Our product has enabled AI to handle most routine customer enquiries, allowing businesses to provide customer support around the clock,” Nur said. “Users often perceive conversations as natural because responses are delivered quickly and with contextual awareness.”
Citing one deployment at an e-commerce business that sells through social media, Nur said the platform reduced the number of staff managing customer messages from 20 to two. “This reduced costs for the business while enabling faster responses to customers’ product enquiries,” he said.
According to Nur, the platform supports customer interactions in Bangla, Banglish and regional dialects, including voice messages, across WhatsApp, Messenger, Instagram direct messages and web chat while maintaining conversation context. Beyond responding to enquiries, it can place orders, confirm addresses, book delivery slots, and interpret product photos and voice notes.
Turning CCTV into an AI agent
Another example is EnlytiX, an agentic AI-powered video intelligence platform developed by Dhaka-based AI company Sigmind Limited. The platform is designed to transform conventional CCTV infrastructure from passive video recording systems into AI-powered tools capable of interpreting events, responding to natural language instructions and carrying out security-related workflows automatically.
Unlike traditional surveillance systems that require operators to manually review recorded footage, EnlytiX uses AI models to analyse live video streams, identify anomalies and answer questions about incidents through text or voice commands.
Abu Anas Shuvom, founder and CEO of Sigmind said, “EnlytiX acts as an autonomous, multi-camera security agent that natively understands the physics, geometry, and semantics of the physical world. Instead of writing complex software code or drawing rigid pixel lines, enterprise operators can simply tell the agent what they want to care about in plain natural language, and the system orchestrates and acts immediately.”
According to Shuvom, the platform supports interactions in more than 140 languages, including Bangla and English, and can simultaneously analyse video feeds from multiple cameras. The platform reduces bandwidth and storage requirements by transmitting only changes detected within a scene instead of continuous raw video.
Citing a pilot deployment at a garments factory in Bangladesh, Shuvom said EnlytiX was used to monitor environmental health and safety compliance, identify production-line issues and generate automated alerts. The platform was integrated with the factory’s existing IP cameras and network video recorder (NVR) infrastructure, allowing it to be deployed without replacing existing surveillance equipment.
While commercial adoption in Bangladesh remains limited, the emergence of locally developed products suggests companies are beginning to explore applications beyond conventional chatbots. Whether wider adoption follows will depend not only on technical capability but also on how organisations address governance, accountability and trust. Because these systems can make decisions and perform actions with limited human intervention, organisations deploying them must establish clear operational boundaries, maintain transparency and protect users’ personal data, particularly in applications involving financial transactions, healthcare, public services and other high-risk sectors.