Software Automation is an interesting category. It's essentially a bunch of 1’s and 0’s being told how to act like a turbo-charged human. Crazy, right?
But how do you turbo-charge for the most impact? What things should you teach software to do? And more importantly... HOW should they do it? This is a challenge we face daily and I wanted to give you a peek behind the curtain of what drives this Lucy AI, so you can understand the impact it has on your business.
When you boil it down, our software development is largely driven by two things:
We marry the two together to help us make decisions whenever we make changes to Lucy, be it new features or adjustments to old ones.
What do you mean by "data"?
Great question. Let me hide in the bushes with binoculars and put on a creepy coat to explain it to you. By data, I mean the actions that real humans take when presented with scenarios while using Lucy. It might be that Lucy came back with a pricing error and gave the Customer Service Rep (CSR) a couple of options. We record that action. Maybe Lucy suggested a particular product because the original document had the wrong product code on it. We record that too. What I'm really trying to say is, we record everything. Each action taken by humans (and other machines too) gives us a better understanding of the situation. This data goes into a big bucket of 1’s and 0’s for us to refer back to when embarking on changes.
Tell me about the Customer Requests
Ok. So when our customers request new features or changes (via https://ideas.letlucy.com) we evaluate them in a couple of different ways. Here’s a basic walk-through of the process we use to create a new/changed feature.
Step 1 for us is really about understanding the problem. More often than not we're presented with ideas for the solution (e.g. I would like a big bag to help me carry my desktop computer from place to place) rather than the actual problem (e.g. It's difficult to easily move my computer). The only way we can offer a better solution (a laptop computer), is if we start with a true understanding of the issue at hand.
Step 2 is then about understanding who else has that problem. If it's affecting a lot of people then it's easy for us to set the priority on it. This is where that awesome data above starts to come into play. We can generally look at the patterns of other organisations to see if they align with the requirement from Step 1, so we can put our arms around the problem. If the data isn't telling enough of the story then we start picking up the phone or walking into offices to see it first-hand. It's amazing the kind of insight you get by just spending a day entering sales orders with a team of CSR's.
Step 3 - Now that we understand the problem and who has it... we’ve gotta work out how we fix it. Up again comes our amazing data to help us look for any patterns. Perhaps people are already solving the problem manually, and we just need to help Lucy to understand what she should be doing to automate this particular problem. We saw this when it came to price acceptance. Customers of our customers were always getting pricing wrong and we saw a pattern in what CSR's were doing... so we taught Lucy about it and now she automatically accepts price variances based on business rules and logic. If it's not already being done manually and we're starting from scratch then we generally round table ideas and try to work out as a team the best way to solve the problem so it suits everyone.... which can be tricky because so many businesses skin cats in different ways.
Step 4 - Implement, Test, Tune. Once we've fixed a problem we don't just assume it's fixed. We always keep our eyes on the data and make sure that the results match up with our hypothesis. Lucy is a growing, learning AI. It's not the kind of software that you build once and it's done. Lucy is constantly teaching us the nuances of purchase orders, and in turn, we teach her how to handle them. This loop will continue forever…. Or at least until Lucy becomes sentient and takes over the world.
But Sales Order Automation is simple... right?
On the surface, sure. If we had to build a system that worked perfectly for one single company, we could probably all pack up and go home now. The intricacies come with understanding the different ways in which people run their business, providing automation software that can be customised to suit the environment, and still maintaining ease and simplicity. You don't need an IT degree (or any degree for that matter) to use Lucy and we intend on keeping it that way.
So keep those ideas and feature requests coming, and we’ll look after the 1’s and 0’s.