Smith.ai Offers Perspective on AI Communications Today, Tomorrow

We recently posted Part 1 of a Q&A with Maddy Martin, head of growth and education at Smith.ai Virtual Receptionists, who will be speaking at The Future of Work Expo early next year.

Here’s Part 2 of that conversation.

How does AI come into play in the Smith.ai solution?
Martin: AI allows our receptionists to focus only on relevant calls, so their time is spent with new leads and clients – interactions that impact the business and are not a waste of time, money, and other limited resources.

AI allows our receptionists to focus on providing excellent customer service and effective lead qualification, to be friendly, less frustrated because they’re not dealing with spammers, and to generally more satisfied with their work. It reduces basic decision making thanks to smart prompts in or system, so they can focus on being their best selves and delivering great phone and chat assistance. It’s a better experience for the receptionists, which results in a better experience for the clients of the small businesses we serve.

AI also broadly allows individuals working on AI-assisted software to increase their technical skills and be more marketable as individuals in the workforce. It’s educational and sharpens their skills. In the context of our receptionists, specifically, it allows them to refine their interpersonal skills because they’re not bogged down by as many routine, administrative tasks or grunt work. The more basic work is automated, the more our receptionists can operate at a higher level. It’s like Maslow’s needs hierarchy – when the basic needs are met, higher functioning can be unlocked. The more our AI assists our receptionists, the better they’re able to perform, which is in turn better for them and our clients.

From human resources and business management perspectives, these efficiency and quality gains in outsourced workforces result in lower staffing costs without incurring a reduction in work quality. Outsourced services like ours can mean that instead of hiring in-house receptionist or other member of support or administrative staff, and paying benefits and other costs associated with that staff person, businesses can contract the work out and save money. It also means savings can be put toward more specialized and highly skilled workers who are more essential to hire in-house and help the business maintain or achieve a greater competitive advantage.

High-quality outsourced workers are also infinitely more flexible and scalable. And less often discussed, they also make higher quality workers accessible to businesses operating in rural or remote areas where applicant pools are not as robust.

Some sources argue that many offerings in the marketplace that claim to employ AI are actually not AI. How does Smith.ai define AI?
Martin: Advances in deep learning and availability of vast amount of computational power have provided major breakthroughs in AI from facial recognition (Apple’s Face ID) to real-time language translation and natural language understanding. We are witnessing the very first inning of applying AI at massive scale (instead of in research labs) even though it’s still in the early days. There are lots of applications that already take advantage of AI, and the systems are getting better with more data and training over time.

Because of this, we believe the best way is to provide the combined intelligence of both humans and machines to tag, train, and provide real-time feedback for hand off. We have not yet reached the stage where a chatbot can completely take over. As most people have seen, these chatbots without human involvement can be very frustrating and dissolve the customer experience.

Our view is by combining human intelligence with AI, we can offer a personalized human touch, improved by machine intelligence. For example, we block spammers by checking against 20 million known spammer phone numbers using machine learning and multiple sources. That’s far better than filtering spam by hand.

How would you describe the maturity and potential of artificial intelligence-based solutions today?
Martin: Today’s AI still require lots of training, tuning, and hands-on human correction. AI works really well for deterministic tasks and is mostly great for FAQ or simple Q&A. When it comes to ambiguous human interactions that require contextual information, we are still not there yet.

Having said that, it’s a matter of time until AI systems can capture more data, combine more sources, and build more sophisticated models to capture information on state and context.

What’s next for AI?
Martin: We are going to see AI take on more complex decision making (as opposed to simple Q&A) based on past history, context, and human guidance.

There will be more autonomy (e.g., self-driving cars) and AI that is better at being both proactive and reactive.

We will see a significant increase in productivity, which will result in people being able to spend more time (now freed up thanks to AI efficiencies) building personal connections and running better businesses.

Edited by Maurice Nagle