Business leaders wrapping their heads around the impressive capabilities of ChatGPT are still wondering how LLMs can impact customer service.
Understanding the business implications of ChatGPT and Large Language Models (LLMs)
“ChatGPT represents an urgent strategic priority for us right now." I’ve gotten dozens of messages like this in the past seven days.
Why? Well, just last Wednesday, the time it takes to summarize a meeting, write a press release, code a script, or do literally thousands of other tasks plummeted from many hours to a handful of seconds. With the release of ChatGPT, the type and amount of work humans are expected to perform has fundamentally changed, transforming the world of business along with it.
First, some background. ChatGPT is a conversational application built on top of one of OpenAI’s large language models, GPT-3.5. ChatGPT demonstrates just how impactful applications built on top of large language models (LLMs) can be. As a result, the profound implications that LLMs have for businesses, particularly within the customer service domain, is becoming more and more evident.
Imagine a future where the cost of assigning every one of your individual customers a dedicated support person is essentially zero. Where a trusted member of your team is available every moment, in every channel, in every language to resolve any question the customer asks. What would be the impact to your business of offering this kind of VIP experience to every customer? How would speaking with every one of your customers every day impact how you make decisions? And how would being this available to your customers change the way you think about your competitive landscape?
This future is now much closer than you think because of applications that leverage LLMs. So if you’re not already figuring out what LLMs mean for your business, you’re likely behind.
Here’s what you can do immediately:
- Educate yourself and your team about what LLMs are. Here is a useful primer to share with your team.
- Identify where in your business LLMs can have the greatest impact. Create an #LLM channel in your company and have colleagues share their learnings in the open.
- Assess your existing vendors on their LLM competency and how LLMs are shaping their roadmap.
Four guiding principles to lead your organization through LLM change
Expect a lot of noise and hype as you craft your long term strategy. To cut through the noise as you lead your organization through this period of change, look to these four guiding principles to ensure you’re using LLM applications in a way that’s actually impactful to your business:
- Accuracy. LLMs have no understanding of factuality, and therefore can generate inconsistent answers to the same question. It’s critical that as you apply LLMs, you’re sure their output is accurate. Oftentimes generations seem correct, which can make inaccuracies particularly nefarious. While this might be useful in a creative context for consumers, in a business context, if your store hours or insurance policy is misreported, you’re in trouble. And while inaccurate predictions are bad enough, undetected bias is a related challenge that can have even more lasting effects on the trustworthiness of your AI and therefore your brand reputation.
- Explainability. It’s very challenging to interpret the decision made by an LLM and understand how it arrived at a prediction, which is why models are often referred to as a black box. But being able to peek inside this box so you can explain a given result is essential to both diagnosing errors and increasing trust and adoption of your AI system.
- Safety. Safety is a foundational expectation that requires due consideration. You must have confidence that your AI won’t communicate anything offensive, treat someone unethically, or put your customers or your business at risk. While technology will soon solve most of the obvious safety concerns, it’s the issues that fall in an ethical gray area where most companies will realize they’ve failed to adequately invest in safety. Companies who include AI ethics representation as part of how they continuously improve their AI will have fewer safety issues and therefore less risk associated with their AI strategy.
- Continuous Improvement. Once you have confidence that your AI is capable of powering a truthful, explainable and safe experience, the single most important factor that will determine the overall success of your strategy is the rate at which you can improve your AI over time. This requires the ability to rapidly detect errors and new automation opportunities, teach your AI how to behave, and measure the resulting impact. The speed at which you move through this loop is a function of both your technology and how your team is organized. Companies that centralize this “AI Continuous Improvement (AICI)” function in an automation team or center of excellence will realize more value faster. Meanwhile companies that rely solely on engineering to improve model performance will take longer to realize transformative impact. In all cases, your AICI velocity will become just as important to your business as the time it takes for you to go from a job posting to a productive employee.
While these principles apply to all domains, your business will need a version of an LLM that is specialized for your domain and customer base. At Ada, we’re focused on empowering businesses to automatically resolve customer service inquiries across all channels and languages with the least amount of effort. Over the last 6 years, companies have automated billions of customer service interactions using Ada. More than a year ago, we started applying LLMs in service of our mission. Our customers are using LLMs in production with Ada today and our teams are more focused than ever on ensuring that the customer service conversations we automatically resolve are truthful, explainable, safe and, most importantly, drive continuous improvement.
Mike Murchison is CEO of Ada, a proud University of Toronto graduate, marathon runner, and Raptors fan. Recognized by Forbes 30 Under 30 and EY’s Entrepreneur of the Year program, Mike is also a Fellow at Creative Destruction Lab and volunteer for VentureKids, a program for Canada’s underserved youth.