Automation Temptation #1
You may have seen some of Real Estate Foresight's China City Reports, like this one on Guangzhou. An interesting case as Guangzhou is now a top performer on house price growth momentum by CREIS as well as NBS measures; but do you know that we automate the production of the reports with our software? Technically, powered by REF's sister venture Robotic Online Intelligence (ROI) where we focus on automation of research.
That's only the starting point of our automation efforts.
In this and the following two short notes, I will share what we have been up to in bringing robots and AI to research to save precious human time, noting that the technology itself is not tied to the property sector or China, but it was natural for us to apply it to China property first.
Such applications tend to be labelled as Robotic Process Automation (RPA), with a definition from Gartner: "Robotic process automation (RPA) tools perform "if, then, else" statements on structured data, typically using a combination of user interface (UI) interactions or by connecting to APIs to drive client servers, mainframes or HTML code. An RPA tool operates by mapping a process in the RPA tool language for the software "robot" to follow, with runtime allocated to execute the script by a control dashboard."
It's a large sector, where our focus is on the niche of automation of market research and intelligence in the investment context.
Note #1: 20x in Production
You can think of this first level as the automated 'PDF/PPT manufacturing' from a range of CSV / XLSX files or databases, with the text logic overlay with many rules and exceptions, which in the end yield a report.
We use that for the weekly data updates, the city reports, as well as the monthly 100-page PPT decks for clients' reports to their clients.
Our wow moment happened with the production of the REF China City Reports. Some time back, it used to take around 12 hours for 12 reports, even with the help of typical templates and data tools.
But with the new approach, we compressed that production time to 2 hours for 40 reports, with the same human effort and including the human checks, but with less room for any human error.
That's going from 60 mins to 3 mins per a report on average. Well, roughly, as while we are data-obsessed, we don't actually count these minutes and we don't do that at scale.
60 to 3 is 20x but contrary to some impressions we may leave when we show how this level of automation works, it is not about some AI going out there and magically getting all the data and writing text around that... Even GPT-3 is still far away from anything close to that. This is about the pure efficiency of production, with human oversight and corrections, using licensed and proprietary data. No predictions, just saving time. And you don't need technical skills to run such production.
The key to making it work is the human (that kind of 'neural network') expertise in designing the logic of the reports or analysis, where automation only kicks in after the broader structure and parameters have been defined. That's the essence of our original approach.
Customization is needed but with sufficient levels of abstraction in our code, we can deploy it in any domain/scenario that meets the criteria.
Technically, a Python code runs at the back, with inputs from multiple data sources in a defined format. At the moment, we use it extensively for our reports as well as for some clients.
And that was the origins of our automation journey. Then we have taken it much further and I will share more in the next weeks:
Note #2 Market Intelligence: Model, Search, Filter, Publish. Repeat.
Note #3 Robo-Analyst Jr.
Robert Ciemniak is the Founder-CEO of Robotic Online Intelligence Ltd (ROI) and Real Estate Foresight (REF).