A Systems View Across Time and Space
Type of Service bots | Description | Limitation |
---|---|---|
NLP or linguistic service bots are rule-based and use if/then logic to generate conversational flows that deliver the fine-tuned control and agility omitted in machine-learning service bots These are the most common bots encountered by the public through live chat, an e-commerce website, or Facebook messenger More advanced service bots are multi-language | Linguistic service bots have a highly labour-intensive approach which can be rigid and slow to develop, since language conditions review the order of words, synonyms and common phrases, ensuring that questions with the same connotation get the same answer Capabilities include interactive, frequently asked questions, delivering specific scheduled communications, slot filling, making reservations, purchasing catalogue items, updating customer profiles, or other basic transactional competencies, such as taking pizza orders Virgin Trains in the UK uses service bots with NLP to automate customer refunds by reading customers’ emails, reducing daily processing time and manual labour by 85 per cent | Interactions with these bots are specific and structured, during which automated tests check the bots’ quality and consistency but cannot correct any bot misinterpretations, since they require humans to modify the conditions |
AI service bots use machine learning that is more sophisticated, interactive and personalised than rule-based service bots, since they are more conversational, data-driven and predictive | Algorithms mimic human cognitive functioning allowing service bots to adapt and handle non-standard cases by observing humans resolve problems (such as system errors, unpredicted system behaviour, or changing forms) AI software can sense, reason and act with structured and unstructured data, performing tasks normally associated with human intelligence, such as decision-making, visual/pattern recognition, speech recognition and translation between languages Gradually with the power of data, they become contextually aware and leverage NLU to personalise a user’s experience through predictive intelligence | AI service bots work best if tasks are well matched to their capabilities, since they require a large amount of training data and human training specialists to perform even basic tasks Any malfunctions are difficult to fix, let alone optimise and improve |