Sam Oliver, Founder of OpenFi- Interview Series

Sam Oliver is a tech entrepreneur, property investor, and author who launched his first startup at just 20 years old. His latest venture, OpenFi, equips large companies with conversational AI on WhatsApp to onboard and nurture customer relationships.

Can you explain why you believe the term “chatbot” is inadequate for describing modern conversational AI tools like OpenFi?

A chatbot is a preset rigid conversation, often a large flow chart with multiple choice or equivalent responses. The static nature of the conversation doesn't create any emotional or personal connection. If a question or piece of data doesn't fit in the static box the conversation becomes broken. How many times have we been the frustrated user pleading to speak to a human representative who we can have an actual human conversation with?

This is where modern conversational AI tools like OpenFi come in. Firstly there is no need for the rigid structure, gone are the multiple choice questions bringing back bad memories from school. Instead there is a genuine conversation that feels like there is a human helping. So not only do customers gain the benefit of a friendly, engaging and personal feel via the conversational AI, but the business is also gaining the ability to turn human flow conversation into required and optional data. 

Calling conversational AI a chatbot is like referring to an iPhone as a landline. They’re just not even in the same category. 

How does OpenFi differentiate itself from traditional chatbots in terms of functionality and capabilities?

The most important differentiation is the user experience. Because chatbots are extremely limited in their prescribed conversations, they’re prone to breaking down when asked anything outside of those fields. They also tend to make mistakes like asking for the same information multiple times or responding to a question with multiple-choice answers that are not even relevant to the original question. 

Conversational AI is a completely different experience. You really believe you are talking with a human and a very capable one at that. We refer to our conversational AI as Superhuman. The natural language format with a style that is both personal and on brand is something chatbots are not capable of. From those natural language conversations, OpenFi can extract data and store it according to client preferences, often within a more traditional database field structure or CRM. 

What common misconceptions do business leaders have about integrating conversational AI into their workflows?

The biggest misconception is equating conversational AI with a chatbot that isn't “on brand” or has a poor user experience. They often put chatbots in the same bucket as a telephone IVR experience. When business leaders use our conversational AI as a user for the first time they quickly realise that is not the case. 

Can you share some specific ways OpenFi has improved lead generation and customer nurturing in the financial services industry?

We recently completed a hugely successful trial for a mortgage broker via our SocialBoostAi product. We generated new leads via social media just like a normal agency, but instead of the leads going through to a busy sales team, they were handled by our superhuman AI that can immediately reply and through their conversations pre-qualify, educate and ultimately escalate ripe leads to the sales team. The results were staggering with conversion from initial lead to finalized mortgage increased by 3.2x! 

This highlights the difference in user experience between a traditional form fill and triggering a live, two-way WhatsApp conversation. The ability of our conversational AI to offer informed and educational content is a big differentiator that helps qualify the user, but also engages them, pushing them through the sales funnel.

How does OpenFi address the challenges of informational inaccuracy and inflexibility that are often associated with traditional chatbots?

We develop industry and client models from bespoke knowledge bases. This ensures that the conversational AI will only respond from a source of truth. It's like an actual human – when acting in an informational or educational capacity,he/she can only answer from the knowledge within his/her brain. We've replicated this structure but rather than needing years of education and product training, we can set it up in a matter of days 

We've spoken about the inflexibility of chatbots and how our conversational AI has a natural flow which isn't rigid. A typical human conversation doesn't have a preset flow, so neither does our conversational AI. Yes, there is often certain information we want to extract but the order and manner of how we collect it isn't set in stone.

How does OpenFi use AI, NLP, and LLMs to provide a seamless and engaging customer experience across platforms like WhatsApp, live chat, and email?

Currently OpenFi is in beta, and we’re only communicating on one channel, which right now is WhatsApp. Additional channels like live chat, email, LinkedIn and socials will come later this year. 

Large Language Models are a tool that use Natural Language Processing to provide the simulation of human intelligence. OpenFi’s platform is enabled by ChatGPT4o’s LLM and NLP capabilities.

Currently many customers' experiences can vary drastically from business to business. Some customers will send enquiries and never receive a reply while others don’t hear back in days. The first objective of OpenFi is to provide instant responses to customers 24/7, which is possible when using conversational AI agents powered by a LLM

The second failure of many customer experiences is the inability of systems to provide specific information. Chatbots and live chat facilities often run on simple scripts and logic trees that can not handle edge cases or complex enquiries. OpenFi provides consistency of knowledge. We train the AI on a company's entire website and all of their product, sales and onboarding documents. This creates a knowledge base inside the AI that can answer detailed questions about the company’s products and services, pricing, onboarding steps, returns process etc. Not only will the AI always give consistent and accurate information, it is able to do it in any language. 

The third superpower of OpenFi is consistency of action. The majority of sales conversations stop because the sales person does not consistently follow up, nurturing and educating the prospect to the point of sale. OpenFi’s AI agents will never be the last one to message back, the AI is trained to ensure the conversation is always moving forward. 

OpenFi also syncs all data to SalesForce, meaning that the single source of truth remains up to date and has accurate information stored. 

What are some common mistakes companies make when designing and implementing conversational AI tools, and how can they avoid them?

The biggest mistake we made when starting was prompt structure. In order to retain information when using the GPT API, we needed to submit the original prompt and all subsequent answers at every step of the conversation. This led to a large token count that grew exponentially as the conversation increased, making conversational AI cost prohibitive. It took a lot of extra research to discover a new feature in GPT that we could utilise in order to solve this problem in conjunction with some custom development work.

How does OpenFi customize its AI solutions to fit the specific needs of different verticals and client data?

We develop client- and industry-specific models rather than utilising generic and potentially inaccurate models. We focus on developing bespoke knowledge bases which short circuits the learning process while also providing the model with a source of truth. 

What role do human sales teams play in the lead nurturing process facilitated by OpenFi, and how does the handoff from AI to human agents work?

This is up to our clients. We set up our Superhumans to work on the task that our clients want. Many want the lead to be qualified and then for an appointment to be booked with a human. In reality, our Superhumans can work end-to-end or anything in between and hand off to a human whenever or wherever our clients want. This handoff can be in the form of an email, a calendar booking, CRM insertion or even a call being triggered. 

What are some future developments or enhancements we can expect from OpenFi in the field of conversational AI for financial services?

The alpha version of our core product SalesTalkAI is live. It automates the SDR (Sales Development Representative) role, to pre-qualify leads and book them in for appointments. You can check it out by visiting the www.openfi.tech website and clicking on the WhatsApp button in the bottom right side of the page. Your feedback would be appreciated! The more brutal the better.