While many predict ChatGPT and other public-use chatbots will soon become regular on-the-job tools, there is growing support for private chatbots, exclusive to company employees. On June 8, Meta (META) previewed an artificial intelligence tool called Metamate to its workforce. Metamate is trained on internal company data to answer employee questions, produce meeting summaries, write code and debug features. The productivity assistant is reportedly rolling out to a small group of Meta employees at this time.
There is already a demand for companies to provide consumer-facing chatbots, including ones that can perform customer service tasks and aid in e-commerce purchases. For example, Wells Fargo is launching a “virtual assistant” that can complete tasks based on prompts from the user, like transferring money and searching for purchases from a specific merchant. With this technology in hand, it’s likely the bank is developing a chatbot for internal use as well, similar to the Metamate, said Alberto Rossi, director of Georgetown University’s AI, Analytics and Future of Work Initiative. The group researches how AI will change the social and economic fabric of the workforce.
“I wouldn’t be surprised if, by the end of the year, all the major companies have this tool,” Rossi told Observer.
Meta has the in-house capabilities to develop its own chatbot through its Reality Labs division, but other companies don’t have the same technology and teams at their disposal. They would have to pay a third party to build and train a chatbot on data they provide. Companies like Meta, Google (GOOGL) and Microsoft (MSFT) could earn revenue by offering this service. The Wells Fargo assistant runs on Google Cloud’s AI.
Big Tech will likely dominate this market in the future, but it is still early, Rossi said. There are also a handful of software companies providing customer service chatbots that would have the technology to offer private company chatbots, including Thankful, Intercom and Genesys.
What purpose could an internal company chatbot serve?
In addition to the secretarial and programming functions Meta lists as use cases for Metamate, companies could train chatbots on human resources functions and integrate them into their intranets, Rossi told Observer. Intranets are private company networks for sharing information, and despite being around for more than two decades, they aren’t optimized at many workplaces. Complaints include being difficult to use, not being up to date, lacking personalization and having ineffective search functions.
Chatbots knowledgeable about a company’s healthcare and 401k options could answer employee questions or help them switch plans, Rossi said. They could also provide a search function for company documents. While the current iteration of chatbots, including ChatGPT, aren’t updated with real-time data, Rossi expects they will be in the future.
The human resources applications seem like only phase one of this technology, said Kirk W. McLaren, founder and CEO of Foresight CFO, a business consulting firm. Chatbots could perform data analysis for employees—seeing patterns a human might miss and reducing errors in the analysis—he said. Rather than using a public model like ChatGPT for this, a private chatbot could offer additional context from the internal data that trained it. It could also complete accounting tasks and draft legal documents, McLaren told Observer. Employees might feel safer putting financial numbers or sensitive legal information into a chatbot owned by the company, rather than a public one. The use of AI won’t mean employees do less work, but they would have more time and energy to tap into the creative and collaborative side of work that AI can’t necessarily replicate, he said.
Rossi doesn’t expect private chatbots to be an expensive investment for companies, but it depends on how many software companies are offering the service, he said. The more companies that provide the technology, the less expensive it will generally be, he said.
Could people have individualized chatbots?
The technology goes further than company use, according to Rossi. “We are headed towards hyper-personalization, and it’s not just that companies will use chatbots. Even smaller entities are going to have their own personalized chatbot, even individuals,” he told Observer.
Chatbots for individual use could be trained with someone’s social media accounts, purchasing information, food preferences, location data and more, he said. They could offer more precise answers to prompts than ChatGPT does, he said. For example, in deciding where to go to dinner and what to order, a personal chatbot could factor in what cuisine someone likes, what they’ve recently eaten, food allergies and budget. It could signal what dishes on a menu have the combined ingredients someone prefers, based on data from their previous dining experiences, he said.
“This would require someone to give up a ton of personal information and basically trust these companies with your life, but the truth is you’re already doing it,” Rossi said. Social media apps collect data to determine a user’s interests and purchasing preferences, and financial management apps see all of a user’s expenses, including where they travel and shop. “They have a pretty clear picture of what your habits are.”