How AI iteration can improve the customer experience

We love stories of dramatic breakthroughs and neat endings: the lone inventor solves the technical challenge, saves the day, the end. These are the recurring tropes around new technologies.

Unfortunately, these tropes can be misleading when we are in the midst of a technological revolution. It’s the prototypes that get too much attention rather than the complex, step-by-step refinement that truly delivers a breakthrough solution. Take penicillin. The drug was discovered in 1928 and didn’t really save lives until it was mass produced 15 years later.

History is funny that way. We love our stories and myths about breakthrough moments, but often the reality is different. What really happens – those often long periods of sophistication – make for far less thrilling stories.

This is where we are currently in terms of artificial intelligence (AI) and machine learning (ML). Right now we are seeing the excitement of innovation. There have been great prototypes and demos of new AI language models, such as GPT-3 and DALL-E 2.

Regardless of the impact they’ve made, these kinds of big language models haven’t yet revolutionized industries — including customer support, where the impact of AI is particularly promising, let alone general business cases.

AI for customer experience: Why haven’t bots had more impact?

The news about new prototypes and tech demos often focuses on the model’s “best case” performance: how does it look on the golden path, when everything works perfectly? This is often the first evidence that disruptive technology is coming. But counterintuitively, for many problems we should be much more interested in the worst case performance. Often the lowest expectations of what a model will do are much more important than the highest.

Let’s look at this in the context of AI. A customer support bot that sometimes gives customers no answers, but never gives them misleading, is probably better than a bot that always answers but is sometimes wrong. This is crucial in many business contexts.

That is not to say that the potential is limited. An ideal state for AI customer support bots would be to answer many customer questions – which don’t require human intervention or nuanced understanding – “free form”, and correctly, 100% of the time. This is now rare, but there are disruptive applications, techniques, and embeddings that build on it, even in the current generation of support bots.

But to get there, we need easy-to-use tools to get a bot up and running, even for less technical performers. Fortunately, the market has matured over the past 3 to 5 years to bring us to this point. We are no longer dealing with an immature bot landscape, with only Google DialogFlow, IBM Watson and Amazon Lex – good NLP bots, but very difficult to use for non-developers. It is the ease of use that will make AI and ML an adoptable and impactful product.

The future of bots isn’t some new flashy use case for AI

One of the biggest things I’ve learned from seeing companies deploying bots is that most don’t get the deployments right. Most companies build a bot, have it try to answer customer questions, and see it fail. That’s because there’s often a big difference between a customer support rep doing their job and wording it correctly enough that something else — an automated system — can do it too. We typically see companies having to iterate to achieve the accuracy and quality of the bot experience they initially expect.

That’s why it’s critical that companies don’t rely on scarce developer resources as part of their iteration loop. Such a dependency often leads to not being able to replicate the actual standard the company wanted, leaving it with a poor quality bot that undermines credibility.

This is the most important part of that complex, step-by-step refinement that doesn’t produce exciting stories, but delivers a real, groundbreaking solution: bots should be easy to build, iterate, and deploy – independently, even by those not trained in engineering or development.

This is not only important for ease of use. There is another consideration. When it comes to bots answering customer questions, our internal research shows that we are dealing with a Pareto 80/20 dynamic: Good informational bots are already about 80% where they will ever go. Instead of trying to squeeze out that last 10 to 15% of informational questions, the industry’s focus should now shift to discovering how the same technology can be applied to solve the non-informational questions.

Democratize action with no-code/low-code tools

For example, in some business cases it is not enough just to provide information; a action must also be taken (i.e. reschedule an appointment, cancel a booking, or update an address or credit card number). Our internal research found that the percentage of support calls that require an action to be taken reaches a median of around 30% for companies.

It should become easier for companies to actually set up their bots to perform these actions. This is somewhat related to the no-code/low-code movement: since developers are scarce and expensive, enabling the teams most responsible for owning the bot implementation to iterate without dependencies is disproportionately valuable. This is the next big step for business bots.

AI in customer experience: from prototypes to opportunities

There is a lot of focus on the prototypes of new and emerging technology, and right now there are new and exciting developments that will make technology such as AI, bots and ML, along with the customer experience, even better. However, the clear and current opportunity for companies is to continue to improve and iterate using the technology that is already established – to use new product features to integrate this technology into their operations so that they can leverage the business impact already available. realize.

We should spend 80% of our time implementing what we already have and only 20% of our time on the prototypes.

Fergal Reid is Head of Machine Learning at Intercom.