10 Questions with Ken Soh, the Creator of TagUI, an Open Source Robotic Process Automation Software
Recently, we had the privilege of interviewing Ken Soh, the creator of TagUI, an open source RPA (Robotic Process Automation) software, notwithstanding his crazy schedule. For those who wish to learn more about TagUI, you may wish to refer to our previous post The Democratization of Robotic Process Automation — A First Look at TagUI.
In this interview, Ken shared his insightful thoughts on a wide range of topics related to RPA. What is admirable is Ken’s candid sharing on how the TagUI project was initially conceived, and some of his hopes and expectations for the future. As RPA gets increasingly institutionalized, it is certainly refreshing to hear the human story behind this powerful and flexible RPA software.
We hope you enjoy this interview as much as we do.
Without further ado …
[CFB Bots] You founded a company called Tebel.Automation that developed TagUI, an open source RPA software. Could you please explain to us what TagUI is?
It can also control desktop applications through visual recognition and has OCR (Optical Character Recognition) built-in through Sikuli’s integration. Machine learning models are easily used through APIs (Application Programming Interfaces) or native Python / R integration. Yes — you can write Python code directly in TagUI scripts to do inferencing based on the deep-learning model you have built.
I’m now maintaining TagUI as part of my work at AI Singapore. I’ll be frank and share with you that if AI Singapore had not hired me in Dec 2017 to continue developing and maintaining TagUI, it would have gradually died off. Popular open-source tools are typically maintained by large organizations as their user base is too big to be supported on a part-time basis by an open-source contributor.
[CFB Bots] You started back in 2016 when there were already a number of commercially available RPA tools. What was the problem statement that you were hoping to address with TagUI back then? Are there any specific market segments that you were looking to capture?
[Ken] Back then, it was more of a selfish reason and personal motivation than trying to address a market need. I was writing software at DBS Technology & Operations to test trading and investment applications at scale. I resigned and spent a year to develop an open-source version of the then relatively unknown RPA software.
It was partly to learn something new by actually developing an automation software which I want to use myself, partly to create something which can be a free alternative to the expensive commercial RPA software, and partly to assess viability of relocating to Europe with my wife.
[CFB Bots] Prior to the founding of Tebel.Automation, you have had a very successful career, holding technical and leadership positions in various multinational corporations. What was the trigger for this career switch? Also, why did you decide to make TagUI open source, instead of trying to commercialize it, bearing in mind that RPA is one of the emerging technologies rapidly gaining mindshare?
[Ken] The trigger was mainly to assess relocation viability to Europe. The only way to do that was to resign, pack my bag and stay for a year in Europe with my wife to get a feel of the European culture and to dedicate time on validating our assumptions. I also had a hunch that automating processes in production environments would be more challenging versus in staging and test environments when I was in DBS. Around that time, the Singapore government announced its focus on automation. The time was ripe to act!
Open-source is a great way to distribute software and get user feedback fast — which will in turn help me iterate and improve the software fast. I wasn’t thinking of the business potential or profit as the end goal of the TagUI project. So there wasn’t a timeline towards commercialization, but rapid iterations towards a digital process automation tool that I would like to use myself. NSFW — I had a wishful thinking that creating a free automation software can gradually help to increase jobless rates, which can in turn be the fuel towards a utopian society where UBI (Universal Basic Income) is accepted and perhaps the only way forward.
[CFB Bots] There has been a lot of talk about the democratization of RPA. In this regard, leading market players like UiPath and WorkFusion has launched the UiPath Community Edition and WorkFusion RPA Express respectively. Where do you see TagUI fitting into this picture?
[Ken] The global RPA market is big enough for different RPA software providers. There are cool features that UiPath Community Edition, WorkFusion RPA Express and AI Singapore TagUI RPA have, which I encourage RPA software users to explore and use one that best fits their needs.
[CFB Bots] Obviously, one of the key attractions of TagUI is that it is open source, i.e. there is no need for users to pay annual subscription fees for the licenses (as is the norm in this industry). Are there any other reasons why one who choose TagUI over the other options currently available in the market?
[Ken] There are areas where each of those tools does better. For TagUI, being open-source, users would have clear visibility of each line of code behind the tool. That is a plus where security is mission-critical. Vulnerabilities get reported and fixed fast. Also, this hackability gives TagUI users a lot of power to shape their TagUI versions to work exactly the way they want.
There are more comparisons on TagUI GitHub page, amongst its strengths would be rapid development, deployment and maintenance. The scripts can be written in 20+ human languages and integration with machine-learning models directly through API / Python / R is a breeze. For the right users, using TagUI can be really fun!
[CFB Bots] You mentioned that TagUI is now maintained by AI Singapore, a Singapore government-funded initiative to build local artificial intelligence capabilities. With this new arrangement, what changes can we be expecting?
[Ken] Due to the nature of the domain AI Singapore is in, there is focus on adding AI and machine-learning capabilities to TagUI. For example, users can now write Python or R code directly in their TagUI automation scripts to call their deep-learning models for inferencing. They can also easily consume machine-learning APIs from the tech giants.
Secondly, while TagUI has been open-source and has large number of users globally, this is the first time where there is a concentration of TagUI users in the same geographical location. This, coupled with AI Singapore’s reach, can yield interesting user feedback and insights which will make TagUI a more versatile and battle-tested automation tool.
[CFB Bots] Within this nascent industry, there is a discernible shift away from RPA and towards Intelligent Automation. What is TagUI’s roadmap with regards to artificial intelligence or machine learning capabilities?
[Ken] At the moment, users can use TagUI to consume generic AI / ML APIs from tech giants. If users have developed a specific deep-learning model to address their business use scenario, they can directly integrate with their model whether it is implemented in Python, R, or command line packages.
The holy grail for RPA would be letting RPA software watch videos of business processes and automatically generating automation scripts from the videos. This is a challenging problem which is currently being tackled by commercial RPA software. This is something I would like to build into TagUI when the time is ripe.
[CFB Bots] According to HfS, the global market for RPA software and services is forecasted to grow at a compound annual growth rate of 36% over the five years ending 2021. How do you think the market will evolve over the coming years as the industry matures? Any predictions or forecasts that you can share with us?
[Ken] There is likely to be consolidation of RPA software providers, and perhaps service providers. This can be a good thing as there will be more efficiency and economics of scale, and will generally reduce price-to-entry for users planning to embark on their RPA journey.
Nobody can accurately predict the future and it is futile to try to. Nevertheless, with shorter technology adoption and distribution cycles, probably in a few years RPA may reach a saturation point and something else other than RPA or AI will be the next hot thing.
I personally think that will be in biotechnology and biomedical fields, especially in specializations that intersect with transhumanism — where technology is used to natively modify biological limitations such as ageing and terminal illnesses, in additional to adding superhuman capabilities to a tiny group of ‘elites’ in the world.
[CFB Bots] As RPA gains mainstream adoption, we are increasingly seeing cases of job redesign — users who used to perform the mundane, repetitive transactions are now taking on the role of RPA developers. As a seasoned developer, what advice do you have for this group of people, many of whom are from non-IT backgrounds?
[Ken] I just want to say — be open to learning new skills, unlearning old skills and repeat. If you are feeling the heat, know that in the tech sector, engineers are also feeling the heat in an ultra-competitive field that increasingly leads to winner-takes-all outcomes by just a few people. As technology adoption and distribution cycles become shorter, most importantly, do what you are happy to do and passionate about. The money and rewards will eventually be evened out across different sectors as supply and demand adjusts themselves to find equilibriums.
For an example of how being in tech has its own challenges, I’m typing emails at 2:27am on a Sunday night. This is typical of how late I work, sometimes until 3–4am, before I get up at 7+ to conduct a workshop or head to office. The thing is, I enjoy the work that I do, so that’s fine. But would someone from non-IT background rather switch places with me? The person likely wouldn’t, unless he or she has a genuine interest in some of these stuff and not because this sector is hot. Like everyone else, it is just too tough to go on without a real passion in doing the work that every one of us do.
[CFB Bots] What are your visions for TagUI? Can you give us a sneak preview of the upcoming enhancements that are in store for TagUI?
[Ken] My ultimate vision for TagUI, as with commercial RPA software, is that RPA software becomes so helpful and easy to use, that every laptop has some form of RPA software running on it. Imagine Microsoft Word, Powerpoint, Excel and so on. What used to be software that specialists use becomes software that any computer user can make use of in their daily work — to make life easier for themselves and create value for the people around them.
Yes, next up is native integration with Windows applications besides only interacting with them using visual recognition and automation. This is likely to be done by integrating with open-source software such as AHK (AutoHotkey) and Roro (RoroScript). The maintainers of AHK, Roro and TagUI know each other. TagUI’s strength in fast development and deployment, especially for web applications is more or less maximized. Its next phase would be tackling its weakness by introducing enterprise features such as auditing, dashboard and reporting.
As a personal side project, I hope to get some personal time to integrate TagUI with hardware robotics. I’ve already bought a pretty versatile 6-axis robotic arm and OpenMV camera for computer vision. I can’t wait to find time to work on integrating them with TagUI, so that process automation does not restrict itself within the digital domain, but can interact with physical environments at the same time. The costs of commercial hardware robotics are prohibitively high, so if this experimentation yields good result, TagUI can be an alternative free software to be used with open robotics components affordably available from sites such as AliExpress.
Do you have any questions which you would like to pose to Ken? Do share them in the comments below.