ChatGPT, Copilot and other AI automation tools are rapidly arriving – but how can your business leverage them effectively? Read on for guidance.
It's been less than a year since OpenAI launched ChatGPT, and already its presence is ubiquitous. After setting the record for fastest growing consumer application of all time in February, research showed that nearly half of all companies were using it. Alarmingly, a quarter even reported job displacements as a consequence of its adoption.
At one time, the possibility of next-gen AI taking over office work and pushing humans out was remote - the stuff of speculative fiction and overly excited tech journalists. But now it is a reality, and with Microsoft set to release Copilot any day now, the virtual assistant built on top of OpenAI's technology is posed to revolutionize the way we do business forever.
Hype abounds, and so do questions. Where do tools like ChatGPT and Copilot belong in your tech stack, how can they benefit your business, and what steps can you take to prepare for a future of AI-driven automation?
It's hard to give an answer that will remain true tomorrow: AI strides are happening so fast that a large group of tech influencers have asked researchers to please slow it down until we can all catch up – but a moratorium is not forthcoming.
In times like these, organizations must lean on seasoned expertise to help them navigate uncertain disruptions in the world of business tech. But in the meantime, we will give some answers that are true today, beginning with an overview of two tools: ChatGPT and Copilot.
While ChatGPT hit the Internet like a bolt of lightning when it launched in November of 2022, the technology behind it is not new: OpenAI has been building and refining Large Language Models (LLMs) since 2018, from GPT 1 to GPT 3.5.
To create ChatGPT, OpenAI simply modified an existing LLM for a more conversational style of input and stuck that technology behind a user friendly interface (with certain tweaks to prevent problematic outputs).
But why is the technology so effective? Today's LLMs are based on emerging machine learning (ML) models that are trained on very large corpuses of text. With the help of advancements like attention mechanisms and long short-term memory, these models have gotten better and better at predicting what word follows another in a sequence of words.
Thanks to this simple function, they have also gotten better at writing everything from blog posts and emails to poetry and love letters. With language being integral to everything we do in the context of business and office work, apps like ChatGPT exhibit emergent functionality that can be bundled into new apps and use cases.
For instance...
Announced in March of this year, Microsoft's Copilot promises to integrate GPT-powered assistance into its suite of Office 365 apps, including Word, PowerPoint, Excel, and the Power Platform. Moreover, it augments individual sessions with collaboration tools like Business Chat, and a unified database for emails, documents, meetings and conversations that will make the assistant more useful.
For now Copilot is still in its beta testing phase - but if initial reactions are anything to go off of, it's fair to say that it will be a one-of-a-kind product: an ultra-smart personal assistant accessible to everyone in an organization using Office 365 products. With a single prompt, users will be able to:
Even if Copilot falls short of these goals, just one of them could make a major difference for businesses and their employees. But how much of a difference?
We often hear about new tech breakthroughs, and many of them are not truly newsworthy. According to Gartner's Hype Cycle, most tech advances go through a period of inflated expectations, a period of disillusionment and a period of stabilization.
AI is no exception - in fact, AI goes through these cycles so often that they have special names: AI spring (inflated expectations) and AI winter (a period of disillusionment). Today, we're in the middle of an AI spring that arguably began about 10 years ago when generative adversarial networks (GANs) were perfected by Ian Goodfellow.
But we can't assume our AI spring is unfounded - and even a conservative assessment suggests there's more fire than smoke in the hype train's engine right now. Business benefits from ChatGPT alone have been numerous:
While there’s every reason to regard media coverage of ChatGPT, Microsoft Copilot and other emerging use cases for AI automation with skepticism, touted benefits are not merely pie-in-the-sky projections: they are grounded in the concrete truth of what’s already happening.
Ultimately, like shadow IT, employees will bring ChatGPT, Copilot and other AI applications into the workplace (CodeWhisperer, Bard, Bing…take your pick) whether they are authorized to or not.
Most companies won’t mind it – but taking an intentional approach will help them to leverage the benefits more effectively. It will also help them to spot opportunities in emerging AI use cases, such as code writing, product design, cybersecurity, and ML-models fine-tuned for their business.