Check out the Low-Code/No-Code Summit on-demand sessions to learn how to successfully innovate and achieve efficiencies by enhancing and scaling citizen developers. Watch now.
We love stories with dramatic breakthroughs and wordy endings: Lone Inventor solves technical challenge, saves the day, the ending. These are the recurring tropes that surround new technologies.
Unfortunately, these tropes can be misleading when we’re actually in the midst of a technological revolution. It is the prototyping that gets too much attention rather than the complex and incremental refinement that truly delivers an innovative solution. Take penicillin. Discovered in 1928, the drug didn’t actually save lives until it was mass-produced 15 years later.
The story is funny that way. We love our stories and myths about defining moments, but many times the reality is different. What actually happens, those long periods of refinement, make the stories a lot less exciting.
This is where we are currently in the artificial intelligence (AI) and machine learning (ML) space. Right now, we are seeing the excitement of innovation. There have been amazing prototypes and demos of new AI language models, like GPT-3 and DALL-E 2.
smart security summit
Learn about the critical role of AI and ML in cybersecurity and industry-specific case studies on December 8. Sign up for your free pass today.
Regardless of their impact, these types of long-language models have yet to revolutionize industries, including customer service, where the impact of AI is especially promising, regardless of the general business cases.
AI for the customer experience: Why haven’t bots had more impact?
News about new prototypes and tech demos often focus on the model’s performance in the “best case”: what does it look like on the golden road, when everything is running smoothly? This is often the first evidence that disruptive technology is coming. But counter-intuitively, for many problems, we should be much more interested in “worst-case” performance. Often the lower expectations of what a model is going to do are much more important than the higher ones.
Let’s look at this in the context of AI. A customer service bot that sometimes does not give answers to customers, but never gives them misleading answers, it’s probably better than a bot that always answers but sometimes gets it wrong. This is crucial in many business contexts.
That’s not to say the potential is limited. An ideal state for AI customer service bots would be to answer many customer questions, those that do not need human intervention or nuanced understanding, “free-form” and correctly, 100% of the time. This is rare now, but there are disruptive applications, techniques, and embeds being built towards this, 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 implementers. Fortunately, the market has matured in the last 3-5 years to get us to this point. We no longer face an immature bot landscape, with just Google DialogFlow, IBM Watson, and Amazon Lex: good NLP bots, but very difficult for non-developers to use. It is the ease of use that will make AI and ML an adoptable and impactful product.
The future of bots isn’t some flashy new use case for AI
One of the most important things I’ve learned from watching companies implement bots is that most don’t do the right implementations. Most companies build a bot, have it try to answer customer questions, and watch it fail. This is because there is often a big difference between having a customer service representative do their job and articulating it correctly than something else, an automated system, can do it too. Typically, we see companies have to iterate to achieve the accuracy and quality of bot experience they initially expect.
Because of this, it is crucial that companies do not become dependent on scarce developer resources as part of their iteration cycle. Such dependency often leads to not being able to iterate to the actual standard the company wanted, leaving them with a shoddy bot that undermines credibility.
This is the main component of that complex, incremental refinement that doesn’t make exciting stories, but instead offers a truly innovative solution: bots should be easy to build, iterate, and deploy, independently, even for those without engineering training. or development.
This is important not only for ease of use. There is another consideration at play. When it comes to bots answering customer support questions, our internal research shows that we’re dealing with a Pareto 80/20 dynamic: good information bots are already close to 80% of their final destination. Instead of trying to squeeze out that last 10-15% of informational queries, the industry’s focus must now shift to figuring out how to apply this same technology to resolve non-informational queries.
Democratizing action with no-code/low-code tools
For example, in some business cases, it is not enough to just give information; a action must also be taken (ie reschedule an appointment, cancel a reservation, or update an address or credit card number). Our internal research showed that the percentage of support conversations that require action to be taken reached a median of approximately 30% for businesses.
It should be easier for companies to configure their bots to perform these actions. This is somewhat related to the no-code/low-code movement: since developers are scarce and expensive, there is disproportionate value in allowing the teams most responsible for owning the bot implementation to iterate without dependencies. This is the next big thing for trading bots.
AI in the customer experience: from prototypes to opportunities
There is a lot of focus on prototyping new and upcoming technology, and right now, there are exciting new developments that will make technology like AI, bots, and ML, along with the customer experience, even better. However, the clear and present opportunity is for companies to continue to improve and iterate using technology that is already established, to use new product features to integrate this technology into their operations so that they can achieve the business impact that is already available.
We should spend 80% of our attention implementing what we already have and only 20% of our time prototyping.
Fergal Reid is Head of Machine Learning at Intercom.
Welcome to the VentureBeat community!
DataDecisionMakers is where experts, including data technicians, can share data-related insights and innovation.
If you want to read about cutting-edge ideas and up-to-date information, best practices, and the future of data and data technology, join us at DataDecisionMakers.
You might even consider contributing an article of your own!
Read more from DataDecisionMakers