The Third Wave Of AI For Chat With Peter Voss
Business process automation is slowly becoming a reality as the third wave of AI breaks into the business scene. Will this development help bring success? Our guest for today, Peter Voss, the founder of Aigo.ai, hopes to answer that question for our host, Chad Baumeister. Peter talks about his early forays into the software business and the idea that eventually became his next-generation chatbot. They discuss what makes a chatbot AI tick and how an AI makes the job easier. Chad and Peter also discuss the business future of AI and other related technologies.
Listen to the podcast here:
The Third Wave Of AI For Chat With Peter Voss
I've got a special guest with me, Peter Voss, talking about something that a lot of companies are finally getting around to, which is up-leveling their chat experience on their websites. Peter has been at this for many years and launched the company Aigo.ai here in the last several years. We're excited to have you on the show, Peter. Welcome.
Thanks for having me.
It's great to have you. I always like to look at the new virtual world behind what's going on there. What's the plaque and the article on the wall behind you?
A few years ago, I took my first big company public. This is the actual listing plaque and articles associated with that. That was quite a buzz.
That is something that not too many people in the world can say they've done. Congratulations. I like our readers to understand what makes you tick. The best way to do that is to go back to when you're younger. Some of your first memories as a kid. What are those? What was your passion when you walk out of the house in the morning? What would you love to do? We like to tie that back to why are you doing what you're doing now.
That fits in with my experience. I got an electronics experimental kit as a present for my birthday as a kid. I started tinkering around with that. I always had an interest in electronics. When I started my career, I ended up working as an Electronics Engineer and designing stuff. I started my own electronics company. That was super exciting but then I fell in love with software. I could see the potential of software was so much greater than even what you can do with hardware, how easily you can configure things and make them smarter, and how you can use software to improve the human condition, generally. I saw the potential early on. My hardware company turned into a software company. I developed a comprehensive ERP software package.
All the chatbots out there right now don't have a brain. They can’t learn, think, and reason. That’s the big difference.
We grew very rapidly. That’s the company I went public. We went from the garage to 400 people and did an IPO. That was super exciting. That's what got me excited about the software. When I exited that company, I had enough time and resources to say, “What's the next big thing I want to do?” What struck me is as proud as I was of my software, all software are still pretty dumb. It doesn't have any common sense. It can't learn interactively. If the program didn't think of something, it'll crash or give you some error message. I embarked on this project of understanding what intelligence is and how can we make software intelligent. That’s what's been my passion for the last several years. It’s figuring out how to build intelligent software.
This ultimately culminated in Aigo.ai, my current company, where we are offering a chatbot with a brain. All the chatbots out there right now don't have a brain. They can’t learn, think and reason. That’s the big difference. This is the second generation of our software. When I started in 2001, I spent several years researching intelligence and what intelligence entails both from even philosophy, epistemology, theory of knowledge, cognitive psychology, how do children learn, and IQ tests measure. In 2001, I coined the term Artificial General Intelligence, AGI, together with two other people. We wrote a book on it to bring the field of AI back to building thinking machines. After many years of R&D, I launched my first commercial product in 2008 in the call center space. That is an IVR with a brain. That was the first generation of our technology. Now, Aigo is the second generation.
I remember meeting with the CEO of a Bay Area company several years ago. He had purchased code. I remember he said he mortgaged his house to buy the code from an Israeli firm because instead of reading the first sentence only, it had to read an entire paragraph. The Israeli listening intelligence from the government provided that form of intelligent ability to read more than just the sentence. I'm assuming that your code-base has the ability to read more than how much does it cost. How does the actual black box work? Can it understand entire paragraphs?
A key requirement is that you have short-term memory. You remember what was said 2 to 3 sentences ago, but you also want long-term memory that you remember what you said yesterday or a year ago to have a useful experience. In particular, you need to be able to remember what's happened in a conversation and to be able to use that information. You need to be able to understand the whole paragraphs. In our case, it's more of the to and fro of the conversation. You need to remember and utilize what was said earlier in the conversation. The way we go about it without going into too much detail is our system is a chatbot with the brain. What the brain does is it has a knowledge graph that has a lot of background knowledge about the topic that you do. If it's a sales assistant, then it would have the background knowledge about sales and how you do sales.
As you have the conversation, that knowledge graph has expanded the new information you learn as you're having the conversation. That knowledge graph is also used to make sense of what the person is saying. You don't want to just understand three words. You want to be able to understand the way humans talk, which could be long sentences with clauses and subclauses and if, then, but. Our system can understand much more complex conversations. As it understands, it integrates it into the knowledge graph. It becomes part of the knowledge that's available to continue the conversation and help the customer.
Is it with just that customer or is the knowledge graph shared across other customers as the system learns?
Let me give you an example here. If you had 1,000 salespeople using this intelligent chatbot and one salesperson has a particular way of doing things and he teaches his Aigo, “This is how I want to do sales.” You don't want that to apply to all of your salespeople. It may be bad practice. You want that to be on a QA loop. A useful way of explaining it is our brain is in three layers. The inner layer is what is common to all companies and all conversations. That's common-sense knowledge, common sense conversational ability and so on. We then have an additional layer that is the corporate information that is shared by that corporation using the Aigo. What their business rules are, it would have an API to their backend services.
Using a salesperson as an example, that would be the layer that has the integration to Salesforce or whatever you're using and the particular business rules for that company. The third layer, the outer layer is for each individual using the system. The individual salesperson might also want to teach Aigo about their family, “Remind me to pick up the kids on the way home.” You wouldn't want to share that. Both in terms of the privacy issue of the individual user and the fact that they may be teaching bad practices, you want to QA layer. That information can be used and put into the middle layer that can be shared between all of the users, but you want management or QA to sign off on that.
I'm getting a feeling there's a human in the loop on the smart chatbot. Where does the human play? Is this purely automated?
What I'm talking about is fully automated. We are nowhere near human-level intelligence. For practical applications like customer support or sales, Aigo will handle whatever Aigo can handle. To give you an example of a large retail customer that we have, Aigo will start to take the sale, but then the customer may ask something like buying flowers or how do they maintain them. Aigo may not have that information. We may not have the APIs or we may not have yet been implemented. There will be limitations about what you can do. You can hand it over to an agent and the agent gets a transcript of what has already happened, who the customer is, it’s already verified and this is how far we got in the sales process but here's this question. Can you answer that question? The human in the loop can answer that question and potentially hand it back to Aigo to complete the sale. Take the credit card information, delivery details and complete the sale. You can have the tango between Aigo and a live agent.
It makes me think of a partner that we have that we're deploying that would fit in a retail environment like this for post-sale. Imagine they buy a shirt and it's ready to ship. Immediately, that can talk to another AI that then sends a personalized video from this woman that says, “Thank you so much for buying. I'm packing your shirt right now. This little necklace and bag would look so good with it. Click below if you'd like me to add it to your order.” She's looking you in the eye. She even says your name because the video merge in through a thing called stitching, “Peter, thank you so much for ordering. I'm putting it in the box right now.” The level of sophistication around this stuff is getting pretty advanced.
Even the advertiser is becoming that good. You don't even need necessarily a human in the video. The uncanny valley problem potentially. It's getting pretty good. You could have custom videos made like that as well.
The piece that comes up when we start to get into that gray zone. It's coming. Let's face it. The question that I always often ask is around ethics. Who gets to decide what's said, how it's said and who says it. There are so many of those decisions. Do you see a need for an ethics committee at companies? How do ethics get interjected into these kinds of AI topics?
AI is not about replacing humans. It is about helping them.
I don't see that any different from other business decisions that you need to make. On a spectrum, either you're more ethical or less ethical. Sales is one of the areas that has a pretty bad reputation for ethics. Anything goes as long as you close the deal. Better companies do have ethics and say, “We won’t step over that line. We won't do this. We won't do that.” There's feedback from users. What will users accept? What will they complain about? What will they say? “They upsold me but I'm happy with what I got. I regretted it and then I returned it and they took it back. No problem.” It's an interaction between the company or customers you have and finding that balance and how hard you want to push. I don't see AI ethics being at all a separate issue.
It’s just ethics. I like that perspective. I ran a team of eighteen people in Manila who did manual chat. The thing I was most frustrated about is I'd go there for eight days. I'd implement something and then you'd come back and magically disappears with the new group of people that came in and took over those seats. The company I worked with used chat to sell small 1 and 2 suite deals. I talked to another company that sold a $375,000 suite to an NFL playoff game or it was the Super Bowl. It can go from small, medium to big-ticket sales. What are your thoughts on where does chat work the best and where it doesn't work?
I don't know of any other chat technology that has a brain. If we talk about what is out there at the moment, where chat works well is if it's like “What's my account balance?” thing, or maybe FAQ's where if you have a good database and a good search engine. You can type in a question and it can come back with the answers, or “What's the status of my order?” and you can give the order number. It's very straightforward one-off type things. Chat can be good at that but at the moment, as soon as you get into a more complex conversation, a to and fro, you need a brain. You need intelligence to be able to handle that. Otherwise, it's frustrating. That's where we are right now with chatbots. We want to change all that and lead in the next generation of chatbots.
Thinking about your sales cycle with your prospects, what are you seeing in 2021 related to AI in the sales motion? If you take chat out for a second, what other neat tools are you seeing out there that have AI built into them?
Our company is still relatively small. We're 25 people right now. We have the strategy of targeting a small number of large customers. We don't have a large sales team. It's my cofounder and me that do most of the selling. We contact companies at a high level and do that personalized sale. In terms of what we're seeing out there and what AI is used for are recommendations and targeted advertising. These are the big things where big data machine learning can help. If you have high volumes of sales, you get a lot of feedback on what works and what doesn't work. That’s what's driving the whole deep machine learning revolution. That's why billions of dollars are going into it because it's making trillions of dollars to companies like Amazon, Netflix and Google. You don't have to get it right for every customer. In fact, if you get it 51% right, you can already make money.
You might be annoying 49% of the customers where you get it wrong, which we are all annoyed with. If I look at what ads are served to me, I usually get ads for stuff that I just bought. I bought the thing, so I don't need an ad or stuff that's totally irrelevant to me. That's where it's currently working. The more sophisticated things might be like Einstein and Salesforce where it has some knowledge engine and recommendation engine that can recommend what the best product combinations are. That's where we're seeing AI right now.
We acquired a piece of code that's pretty sophisticated. Let’s say there's a list of 300 retail type companies that have a checkout cart online or whatever your target market is, then we plug into LinkedIn to see who you know that know those buyers. It could be a first connection or a second. It's a large data set. In eight minutes, it runs and comes back and says, “You have 6,700 pathways into those 300 companies.” Instead of manually copying and pasting, you send the email to the influencer, “I’ve been doing a little research on your chat tool. What if you gave your chat tool a brain?” It's amazing how people want to help other people. By adding in big data and analytics, how effective that can be?
What do you think is the future of AI in sales? It seems like chat is the four-letter word of 2021 so far. Everybody has a CRM. A lot of people have LinkedIn Navigator. There are basic core fundamentals of a company. It seems like chat has started to become a core, but smart chat and chat with a brain will be the next level. Take me past that. What do you think is the thing after that that's going to be mandatory?
What we’re finding as we're talking to these large companies is almost all of them have tried to implement chatbots or have implemented chatbots. They've tried all sorts of different technologies and they've either abandoned it or are not having good results. Adding intelligence to the chat is crucial. It's almost not like what's the next thing beyond that. It's so huge. Imagine, you have a chatbot that becomes good enough to replace the human operator. Everybody has a problem in the call center. Call center is not a lifetime career for most people. You have constant retraining, quality problems and scaling problems. You have peak periods, holiday sales and how do you staff up for it. Your products change all the time, how do you train people with different business rules. It's a nightmare.
As chat becomes more intelligent, not only can it handle more of the conversations, but it can also monitor the live agent and help them train them on the job. The agent becomes smart enough to help them, but then it can probably handle a conversation itself anywhere. That’s the end game from an automation point of view. Most chat companies push the whole automation-human partnership thing. We're not replacing humans. We are helping them. Part of that is because it's politically much easier to sell that we're not going to replace any of your staff even though the accountants will say, “Can we replace your staff?”
I remember when they first came to market and we’re talking about emailing, social and all these different technologies. A few people that I respect in the industry said, "Chad, are you going to pay taxes for the $60,000 a year costs that you would have had to spend?" It's like, "When Excel spreadsheets came along and you moved off of a piece of paper, that was a $60,000 a year job too. They are paying taxes on their own."
The tech companies are paying taxes for the technology that they sell. In a way, it's a more important and more revolutionary outcome of intelligent chat. That is hyper-personalization. If you essentially have a concierge service, the customer is happy, they are repeat customer, you get greeted by name, and you remember what the previous transaction is. It’s something that humans can't do unless they have a sophisticated CRM and are plugged into it. I've not ever experienced that in any company that I've dealt with. Whereas with intelligent chat with long-term memory, as soon as you have the 2nd and 3rd interaction with a customer, you know more about them. Let’s say I want to buy some chocolates for my niece. You know her niece’s name and address. You know she likes chocolates. It might be for her birthday. A year later, “Do you want to send your niece some chocolates again?” That’s hyper-personalization.
You need intelligence to be able to handle complex conversations.
As you talk about, there's an interesting technology that I'd recommend you look at. It's called Codebreaker Technologies. CodebreakerTech.com is their website. They can look at a LinkedIn profile and they can tell you which one of four buckets that person fits within, blueprint, action, nurture or knowledge, BANK. It's your BANK code. If you're an action first, then you would fray. It's like DISC but a more advanced version. If you could quickly and easily scan a LinkedIn, you're in the chat and it says, “Is this you on LinkedIn?” “Yes.” It pulls that in. The system could learn what code of personality are you. If you're an action driver, I'm going to give you a different sentence than if you're a nurture or a blueprint. To me, that's where the level of sophistication under the hood of AI gets where it would be hard. I've been learning this stuff for several years at different sales classes and yet I still talk in the same way that I normally talk to most people.
That’s personalization but what I'm talking about is even taking that further because you still are putting people into buckets. That's a good starting point. If you don't know much about the person, they're that kind of person or this kind of person, but you want to hyper-personalize to the individual. That's hyper-personalization to a billion people that every single person is treated as an individual. That's possible with this technology but also think about it as having a personal assistant as a salesperson. Salespeople all hate using Salesforce. I haven't yet met somebody who likes it. For whatever reason, they just want to go out and sell. They don't want to share their information.
If you have a personal assistant that's an interface to Salesforce, before your sales call you can say, “Aigo, tell me about my next appointment.” Aigo tells you what the previous conversation was. You can say, “Does he have any hobbies? What are his kids' names?” Whatever you may already know about them. Aigo can tell you about that. You go on your sales call, you finish the sales call, and then you go and say to Aigo, “Remind me next Tuesday to follow up on this. Send the customer brochure X and let my boss know what's going on.” Having that personal assistant for each individual salesperson and your calendar reminders. Your spouse will say, “Remind me to pick up the kids on the way home.”
My parents need to be reminded of going to see my son because they missed a lunch meeting.
That's a huge application as well. Not within sales but as an elder companion helping keep contact between your parents and your children, medication, holidays and all sorts of things.
We've been talking with Peter Voss, Founder and CEO of Aigo.ai. If you want to get ahold of him, go to the website. I'm pretty sure you'll talk with a very smart, ingenious chatbot. Go experience it for yourself. Ultimately, I'm sure you'll be introduced to Peter. Great talk with you. Thanks for joining the show, Peter. I appreciate you being here.
About Peter Voss
Serial Entrepreneur, Engineer, Inventor and a Pioneer in Artificial Intelligence. Coined the term ‘AGI’ (Artificial General Intelligence) with fellow luminaries in the space.
Started in electronics engineering, then fell in love with software. First major success was developing a comprehensive ERP package and taking that company from Zero to 400-person IPO in seven years.
Fueled by the fragile nature of software, I embarked on a journey 15+ years ago studying what intelligence is, how it develops in humans, and the current state of AI. This research culminated in the creation of our natural language intelligence engine that can think, learn, and reason -- and adapt to, and grow with the user.
Currently I’m focused on commercializing the second generation of our AGI-based ‘Conversational AI’ technology called 'Aigo' (say: I-go).
Aigo.ai is the most advanced natural language interaction platform available. It is implemented using a brain-like cognitive architecture – also known as ‘The Third Wave of AI’.
This approach puts Aigo at ‘Light Years’ ahead of chatbots and other so-called ‘Personal Assistants’, and puts the forefront of the 'Conversational AI is the new UI' trend.