Gaurav Kachhawa is the CPO of Gupshup, which builds advanced conversational messaging platforms. He knows AI both through his work and working with clients as they deploy it. He gave us a look at what he expects in the very near future with artificial intelligence. (Interview edited for length and clarity.)
Q: What do you see happening right now with business and AI?
A: Every decade there’s an inflection point in technology. We were working in AI way before this ChatGPT phenomenon, but it has accelerated everything. It has been an inflection point and people have realized that this technology has reached a point where you could have meaningful human-like conversations and drive efficiency.
Every time there’s a new technology, you go through a phase of reality check and optimism. This time, what has happened is the base experience itself has jumped up expectations several notches in a very short time frame. In terms of customer expectations, they have been educated through their own interactions with ChatGPT, and they say, “How do I deploy it in business?”
For example, if I’m a car marketplace, I would say, “Instead of deploying, all these different, tools, can I have a chatbot where people can just talk and, book an appointment and make that frictionless process?” If it’s a bank or an insurance company, they’re saying, “I have sales staff everywhere. How do I train them? How do I make sure that the bot can answer the question and they can then use that instead of internal searches?”
So, from employee enablement to support people enablement to customer engagement, there are a whole lot of use cases coming up.
Of course, they don’t realize that there are things like hallucinations, that there are challenges in it.
And so one of the things we realized is the consumer-grade conversational AI here, but there’s a layer above it called business-conditional layer, which puts guardrails, puts all the IP protection, puts security. So that’s what we are in. How do you make this tame this technology for business use?
Dig deeper: How AI enhances multimedia content creation
Q: In what ways is AI different from other technologies you’ve seen adopted?
A: The first thing is, one of the big trends here is that we have closed-source models and open-source models. And that’s the thing about how you drive competition in an evolving space to bring down costs.
If there was only ChatGPT and we had the paid option in closed source for Enterprise, we’d say, okay, I love it, but it costs a lot.
But in a world now where you have a lot of open source models, the cost of actually accessing the software is reasonable. You can host it in your environment, wherever you want. You can have different sizes of it. So that’s a very powerful phenomenon.
Dig deeper: How to do an AI implementation for your marketing team
We have a bigger control of this technology and how to evolve it with open source. That will continue to lower the cost, which means we’ll be surprised where AI will be used. It’ll be on your devices, on your phone. It’ll be a different form factor, but lowering the cost base will allow a lot of adoption everywhere.
And that’s one thing.
The second thing is it’s not just about bigger, it’s about better. New kinds of architectures are coming up. Instead of a horizontal AI general-purpose search, it’ll be a vertical search. You apply AI to a very narrow topic and go deep, this vertical AI is gold, in the sense that it is going to make our lives easier. Imagine you’re getting a medical scan and you get a better predictive capability of what’s ailing you and what are the treatments. Those are the kinds of vertical AI applications we think are going to be big.
Those are not necessarily going to come from Microsoft or Google because they don’t have those deep vertical things. That’s why industry players who have vertical context, like the hospitals and others who can try these things, will take this further for the betterment of, you know, overall society.
Q: What are you seeing in terms of the adoption of AI by companies you work with?
A: First, they all jumped on it, and they thought it’s going to be great. But then they deploy and they realize it’s not doing what we want. Like, I don’t want my insurance bot answering a question with how to make a martini recipe or how to get a pizza.
So, we’re doing a lot of tooling on top of these systems to give so businesses, are able to say this is the kind of boundaries that I want for the bot. This allows them to play with it. They can say how much they want the data coming from their repository versus public internet versus, confidence level. There are a bunch of tools we have given to them that allow them to quickly check these different variations.
This is similar to how search evolved. For a while, it was very hard to manage quality search, but chatbot quality is moving very quickly.
Search worked as bringing in documents that were hidden in the nooks and crannies, but there were still links, you still had to click on links. So this phenomenon that you can ask a question and it will go find that information for you, wherever it is sitting, is just amazing. It’s going to change our behavior. Right? I’m not going to go to Google anymore. I’m just going to ask ChatGPT — especially on the phone, I’ll speak, and I’ll get the answers.
So the bar is at a point where the user interface is evolving. The interface is the most important part to humans. We still want to talk on the phone, we’re not going to change that, but instead of me looking for search and blue links, I’m going to look for an answer. It’s a massive shift.
Get MarTech! Daily. Free. In your inbox.
