Chatbots have been automating everyday customer service tasks for years: they lighten the workload for employees and help businesses save time and money. The emergence of ChatGPT, however, has widened the scope of the technology, and its potential for streamlining and maximizing business processes, beyond anything we’ve seen before. The new chatbots can even write e-mails and debug computer code. ChatGPT’s instant popularity has prompted well-known app factories to launch a flood of similar artificial intelligence tools, but the benefits it offers for a specific range of communication and analysis tasks also come with limitations in a business setting.
ChatGPT is a chatbot built on top of GPT, a generalist conversational AI created by the company OpenAI. Built using neural networks and other algorithms, GPT software is capable of natural-sounding dialogue on any topic. Its main purpose is to create content, but it can also be used as a search or query tool, as a machine translation source, or as an AI assistant.
ChatGPT technology "changes the way you interact with the machine. Since the 1990s, the paradigm has been based on menus and icons," says Kiron Ravindran, an IT professor at IE University, referring to the icons in the folders we use every day to save or open files. "But from now on the algorithm will follow direct written or spoken commands from users. This means personalized communication has finally arrived."
What is ChatGPT for?
ChatGPT was designed to create text and answers in response to user prompts. Once the AI is given guidelines and topics to be covered, it accesses information and creates content automatically, using a style and tone that convincingly emulates a human.
This feature makes ChatGPT more powerful than traditional chatbots in handling customer service queries and tracking consumer data at the same time, as it processes data faster and more accurately. The new functionalities will have a real impact on the call center industry. Esteve Almirall, a professor at Esade's department of operations, innovation and data science, adds that this technology will have implications for apps such as Stripe, a credit card payment processor that interacts with customers and merchants to detect duplicate payments and resolve any existing payment processing issues.
Almirall points out that the greatest transformative effect of ChatGPT will be seen by the software industry. Any user with limited programming knowledge will be able to interact with the tool to write computer code. ChatGPT may provide incorrect answers, but, even so, it democratizes tasks previously reserved to trained specialists. "Estimates suggest that productivity in the software industry will rise by at least 50%, with a clear reduction in bugs and errors. It seems obvious that in future software programming will be chatbot-assisted. The new programming environments will probably rely on ChatGPT to automate tasks such as unit testing,” explains the Esade expert.
ChatGPT also creates content. This makes it a useful SEO copywriting tool in the marketing industry: you can edit content in a range of different formats and craft a single, unified business message for posts on social media or blogs. ChatGPT is hugely versatile for professionals who need to write articles, poems, short stories, tutorials or speeches. It saves time spent on thinking up, structuring and writing material. ChatGPT can also be used to compose personalized emails to customers, putting forward simple but attractive value propositions based on the information provided about the recipient.
These features will lead to a real sea-change in content production, education and legal practice, according to Álex Rayón, vice principal of international relations and digital transformation at the University of Deusto: “Lawyers make a living from reading a vast corpus of case-law, from which they draw out the key points so as to reach a conclusion. This mirrors the value chain that ChatGPT is designed to address: process and summarize textual content, blend it together and generate an output.”
In addition, ChatGPT provides back-office functions that will have a cross-cutting impact on many businesses. The tool offers solutions for entering any form of data, so it can be used as a personal assistant, e-mail organizer or meeting manager. Microsoft has in fact added ChatGPT to its Bing search engine (you can use GPT-4 for free) and already has a beta version of Teams that summarizes meetings in real time and ex post.
The limitations of ChatGPT
ChatGPT operates with publicly available data only: that’s the first limitation. AI generates inaccurate output if the underlying information is unavailable, or out of date, or obscure, or overly generic. "Genuine business solutions emerge when the business creates its own specific GPT using its own private data and knowledge," explains Álex Rayón.
Proprietary data is of vital importance to companies. A report by data security consultancy Cyberhaven revealed that 11% of what company employees copy and paste into ChatGPT is confidential. The average firm leaks sensitive information to ChatGPT hundreds of times a week. Businesses are taking steps to remedy this problem. Many organizations have already begun to restrict ChatGPT use on a discretionary basis among employees. Amazon, Microsoft and Wal-Mart have all issued warnings to their staff.
The limits of enterprise implementation of GPT are also inherent in some industries. "Adoption will be swift in firms with clear business models, such as legal practice, consulting, or software development. Adoption will be slower where the aim is not just to enhance a capability but to replace it altogether, such as customer service—all the more so in sectors with low or no competition, such as the public sector," explains Esade expert Esteve Almirall. However, he points out that pilot projects are underway at local government bodies in Japan and Iceland.
Generative technology itself involves tough questions, says IE University professor Ravindran: "Not knowing whether the output is true or false can have disastrous consequences." This can happen with text, and also with images, where GPT technology is taking its first steps.
Alex Rayón mentions several ways in which using this technology for image creation could involve ethical and legal issues. An AI can sometimes reinforce biases as a result of the statistical generalization inherent in its algorithm. The ability to generate fake images could create a need for a legal duty to disclose that an image is AI-generated, so that humans can distinguish true from false. In addition, AI image generation requires a rethink of creativity and intellectual property, according to the Deusto expert: “Mass use of images made by humans to generate a new image using AI implies an appropriation. So the new cultural framework that emerges from that may be artificial rather than real."
GPT technology providers
The enterprise ecosystem that is adopting this technology includes OpenAI, which owns ChatGPT, Whisper and Dall-E (text, voice and image generators, respectively), and competitors such as Stability AI and Anthropic. Industry giants such as Microsoft, Amazon, Google, Meta and Apple, and several Chinese companies, have also released or are expected to release their own GPT versions leveraging the fact that the technology is open source.
The ecosystem is a hybrid space made up of household names and new challengers bidding to gain ground with a technology that will be at the center of global business disruption.
Expert Álex Rayón predicts that fragmentation will create friction in the market: “There will be a clash between challengers and incumbents. But I sense that the challengers can move forward faster because they aren’t reined in by the inertia of the past. They aren’t listed on the stock market, nor are they committed to an income target they need to hit."