Nov. 2, 2025 /Mpelembe Media/ — AI agents are primarily used in specialised applications designed to act alongside humans within specific fields and manage complex business processes, often forming the basis of “artificial specialised intelligence” (ASI).
The specific uses for AI agents and the applications that rely on them include:
- Customer Service: Companies like Sierra utilise AI agents designed to act alongside humans in customer service roles.
- Legal Services (Law Firms): AI is provided to law firms (for instance, by Harvey). Their tasks go beyond simple requests, focusing on complex assignments such as helping to co-ordinate several law firms involved in a mega-merger, rather than just generating standard paperwork like non-disclosure agreements.
- Medical Services (Doctors): AI is provided to doctors (for example, by OpenEvidence).
- Software Development: AI agents underpin applications like Cursor (by Anysphere), which helps software developers write code.
App developers rely on these agents to tackle complex business processes that are not generalisable and are often described as “the most boring, mundane thing…hidden in the back of a company [that is] slow, expensive and consequential”.
A key function of these agents, as argued by the app developers, is to continually acquire specialised data the longer they remain operational, which consequently improves their performance and reliability for the customer. This is considered essential for retaining a competitive edge over general AI models.
One new revenue model that AI app developers are experimenting with is to charge customers according to outcomes, rather than usage.
This technique is being adopted to help developers offset their rising marginal costs. Unlike previous waves of software development, AI app developers find that the bigger they grow and the more they use Large Language Models (LLMs), the higher their operational costs become. By charging based on the results or value delivered (“outcomes”) rather than simply the volume of queries or time spent using the model (“usage”), they aim to create a more sustainable business model.
For instance, Harvey, which provides AI to law firms, can afford to opt for the largest models because law firms are willing to pay for perfect accuracy in their high-stakes tasks.