I was going to start this post with the typical grandiose third person description of what and who this is. Trace and Circle, a Natural Language Interaction (NLI) AI consultancy built on 30 years of corporate enterprise deployment history, delivering efficient cost saving chatbot systems to the largest, household names, blah blah blah… It all started to sound so stuffy and bloated and that’s the whole point. My name is Pete, me and my good friend Theo have worked building and deploying those systems for the last 15 years each and we’ve learnt so much and got to the point where we realised we could have way more fun building and delivering systems that aren’t stuffy or bloated, that use ground breaking innovative techniques and technologies without sacrificing reliability or performance. We want to help people understand a tech sector that often seems overwhelmingly complex and difficult to measure.

Admittedly, it is difficult to get excited about call centre deflection, especially when the metric of success is how many real life call centre agents are replaced by a system we built. I’m sure we could debate about the inevitable march of technology and progress or the grander topics of workforce value, automation and capitalist ethics, in fact, I’ll probably write a post on that in the future, but right now I’m genuinely excited about the possibilities of what we have created and it’s application outside of the call centre. Theo’s kind of a genius, it’s been a pleasure to work with him for nearly a decade, to watch and help him bring to life our latest project. An advanced dialog management system of our own design, a modular neural net and rule based intent classification framework, proactive product recommendation and complex behaviour queueing with an easy to understand visual construction of complex action sequences and content creation. The list goes on with the sexy features Theo has skillfully developed into a deceptively simple and easy process of building and deploying our chatbots. I’ll create some posts and videos breaking down the features and functionalities of our platform soon. In short, we can offer advanced intelligent, easy to use conversational AI to SME businesses that can interact with customers, resolve issues, complete activities, deliver deeply personalised recommendations and improve user engagement in sectors that really haven’t had a use for chatbot technology to date. 

Since the mid 2010s there’s been a new surge of interest and innovation in the sector, from the steady maturation of machine learning and mass adoption of voice assistance via Siri, Alexa and Google. The core technologies have been around since the 80’s, and a fledgling scene of linguistic AI startups and scholars (I’ll go into the fascinating history of the industry in another post) but it’s taken a solid 40 years for this technology to really become reliable, adopted and accessible. Now it’s surprising when we come across a large corporation that hasn’t engaged in automation for at least internal support, the battle is now for AI companies to provide streamlined experiences, further reduce call deflections and integrate with all manner of 3rd party apps. It has, in my opinion, driven the tech forward and opened a lot of sceptics’ eyes to the possibilities of deploying chatbots. This new raft of features, functionalities and most importantly, accessibility is what’s changing the industry the most. These reliable, hugely cost saving systems are no longer just in the budget remit of the large multinationals. The advances in tech, with machine learning frameworks, further reduce the labour costs and difficulties in getting an accurate and performant chatbot deployed. One of the main technology platforms we employ at Trace and Circle is RASA, an open source Natural Language Understanding framework, that offers secure, versatile and reliable methods for us to extract intent and sentiment from user’s inputs. We studied and researched what we liked and didn’t about the RASA product as a whole, then built on top of this to enable the easiest approaches of content creation and advanced dialog management. We designed a system to visualise the complex flows and behaviours a chatbot needs to provide a feature full experience to users, whether it’s data retrieval, issue completion, product recommendation, or a deeper, more complex dialog.

After 4 years I was offered a job at Artificial Solutions in Spain. 3 years later, my wife and I both took career breaks to travel Asia for a year. On returning, the US division of Creative Virtual offered me a position, which brings us up to now. Of course the fear is there, to be personally responsible and make the decisions on the kind of company Trace and Circle will be, the projects we will take on and businesses we will support. It’s a privileged position to be in, if a scary one. 

The main reason I've committed myself to this endeavour is Theo, a good friend and educator to me for many years. I'm constantly impressed by his capability in difficult and complex topics. Ideas that either of us have proposed, developed, trialled and tested. I’m determined to show the world what incredible things, genuinely pioneering features we have built and how businesses and users can benefit from them. I think Trace and Circle has a unique and pioneering technology platform and a clear and value driven application. I really am hyped to see where we can go with this.

It was during this time I met Pete and we became great friends and bonded over our mutual love of technology, fast cars and constantly winding each other up. We even went on a road trip across California where I almost drowned in a river (possibly a touch over dramatic) but that’s a story for another day.

I love working with Python, Linux, Docker, Neo4j, Rasa and Spacy. My first piece of code was written at the ripe old age of 27 and it was a messy, nested loop, 15 or so layers deep that still gently haunts me to this day. But it worked! It saved me time and it saved the company money. Encouraged, I started to experiment with different ways to abstract chatbot content and behaviour, either using simple spreadsheet representations or via directed graphs and flowcharts. I was hooked and could see loads of ways to save money and magnify human effort.

And it wasn’t just the technical possibilities that got me excited. I had a team of people around me to make things happen, from genius Python and AI wizards, to inspirational leaders and expert project managers. Pete was one of these legends, always bringing new ideas to the table and always making me question my own (annoyingly at times, but required). We had an amazing team and we have been delivering successful chatbots ever since.

So why did I want to start Trace and Circle? I’ve spent most of the last 15 years building customer service chatbots for large companies, and although I enjoyed and value this experience, I started to feel I needed an outlet to do something a bit different. My hope is to bring chatbot technology to new applications across entertainment, marketing, sales, recommendations, online dating, or any use case involving complex dialogue and actions. I also feel I can help make the process of building these systems more efficient, hopefully meaning we can make our chatbots available to SMEs and smaller brands with more unique applications.

It’s early days but we hope to validate the approach with our own chatbot, Trace, who’s designed to guide users through our services and answer broader questions around chatbots and AI. Eventually we hope this to be a useful resource for anyone building a chatbot and wanting to understand more, as well as a showcase for our product features.

So that pretty much sums up my journey. I’m pretty excited about the future and hope you will give us a shout if you’re thinking about building a chatbot. We’re up for chatting!

Cheers!