For all the boasts about generative artificial intelligence revolutionizing the world, the technology has yet to significantly transform administrative work. Workers are dabbling in chatbots for tasks like composing emails, and companies are launching countless experiments, but office work hasn’t seen a major AI reboot.
Perhaps that’s just because we haven’t yet given chatbots like Google’s Gemini and OpenAI’s ChatGPT the right tools for the job; They are generally restricted to receiving and spitting out text through a chat interface. Things could get more interesting in the business environment as AI companies begin to deploy so-called “AI agents,” which can act by operating other software on a computer or over the Internet.
anthropican OpenAI competitor, Announced a major new product today that attempts to prove the thesis that tooling is necessary for the next leap in AI utility. The startup allows developers to direct their chatbot Claude to access external services and software to perform more useful tasks. Claude can, for example, use a calculator to solve the kinds of mathematical problems that bother large language models; be required to access a database containing customer information; or be forced to make use of other programs on a user’s computer when it is useful.
I’ve written before about how important AI agents that can act can be, both for the push to make AI more useful and for the quest to create smarter machines. Using Claude’s tools is a small step towards the goal of developing these more useful AI helpers that are being released into the world right now.
Anthropic has been working with several companies to help them create Claude-based helpers for their workers. online tutoring company Study searchFor example, it has developed a way for Claude to use different features of its platform to modify the user interface and content of the syllabus displayed to a student.
Other companies are also entering the Stone Age of AI. Google demonstrated a handful of AI agent prototypes at its I/O developer conference earlier this month, among many other new AI elements. One of the agents was designed to handle returns for online purchases, searching for the receipt in a person’s Gmail account, completing the return form, and scheduling package pickup.
Google has not yet released its return robot for mass use, and other companies are also acting cautiously. This is probably partly because getting AI agents to behave is difficult. LLMs do not always correctly identify what they are being asked to achieve and may make incorrect assumptions that break the chain of steps required to successfully complete a task.
Restricting early AI agents to a particular task or role in a company’s workflow can prove to be a clever way to make the technology useful. Just as physical robots are typically deployed in carefully controlled environments that minimize the chances of them making mistakes, keeping AI agents under control could reduce the chance of mishaps.
Even those early use cases could prove extremely lucrative. Some large companies already automate common office tasks using what is known as robotic process automation or RPA. It often involves recording the actions of human workers on screen and breaking them down into steps that the software can repeat. AI agents based on the extensive capabilities of LLMs could make it possible to automate much more work. IDC, an analyst firm, says that the RPA market It’s already worth $29 billion, but expects an injection of AI to more than double that figure to around $65 billion by 2027.
