Standards in automation? There's only one Lee in the IEEE... - Horses for Sources

Few people can claim to have led shared services and IT for Kraft Foods, built shared services from scratch for Ascension Health, become one of the first true shared services practitioners to kick the tires with RPA… before establishing the industry’s first standards body for Intelligent Process Automation with the IEEE. Plus, he’s going to be at our inaugural FORA Council (The Future of Operations in the Robotic Age) as the voice of standards and reason this September in Chicago. Yes, ladies and gentlemen, meet the reincarnation of the process pontiff himself, Lee Coulter, who’s going to give us a little more insight into why the heck we desperately need to adhere to some standards if we’re going to find that automation haven that exists somewhere between fantasy, reality and failed promises…

Phil Fersht, CEO and Chief Analyst, HfS Research: Good morning Lee it’s great to chat with you again. You have been pretty deeply involved in developing and working on standards in process automation with the IEEE for over a year, would you be able to give us an update on what has been accomplished, and what we can expect next?

Lee Coulter, CEO Shared Services, Ascension Health and Chair for the IEEE Working Group on Standards in Intelligent Process Automation: Absolutely Phil, it has been quite a journey and I am very happy say that after working through the various societies of IEEE, the Board of Governors realised that this work impacted multiple societies and decided to use their reserve prerogative to sponsor a standards effort at the Board of Governors level. The first standard establishes some common terminology for us, it goes for approval on 5 May and that’s the procedural verification, making sure we have followed all the procedures of setting the standard, and we expect it to be published in June. At the same time a part of IEEE called NeSCom which stands for the New Standards Committee that reviews all proposals. The next efforts, which will be referred to as P2756 in the IEEE world and their website, will be technology, taxonomy and classification for intelligent process automation products. Incidentally, in the same meeting where our first standard will be approved, they will also be reviewing and voting on the next standard. We have significantly increased attention for the second standard, which is really where we wanted to start but we realised we couldn’t do a taxonomy until we agreed what words meant. Several new members across the spectrum of providers have become advanced corporate members with IEEE and we expect to have a first working group meeting towards the end of June, as we go down the path of establishing a taxonomy.

Phil: And when you look at the general state of automation in the industry today, where would you say companies are, as a whole, and how does this tie in with the need for standards?

Lee: It’s interesting, I recently presented an update at an event and a bunch of people hung out after the update, these were people new to the world of automation. They came up to thank me and I thought that was very interesting. I talked to some of them about their reaction to the material and it was very consistent with the frustration that led me to begin the effort in the first place. It is bewildering, and virtually impossible, to watch a presentation or listen to someone else speak, or read the marketing materials, or read any papers on the topic, because you don’t really know how to interpret what you are being told or what you are reading. We are finding that there is a great deal of interest in bringing some clarity so that we can have intelligent conversations. What continues to surprise me is how many organizations are just discovering this. You and I been talking about automation for four years, yet there are major Fortune 500 and Global 2000 companies who are hearing about it for the first time and just getting started. It’s interesting to think about the hype curve and the adoption curve in terms of where we are and I think we are just at the bottom of the hockey stick and it’s starting to become more mainframe. It’s a good time for the taxonomy and the standards to emerge as a large proportion of corporations across the world are entering this phase.

Interesting, I have been pulled into several rather vociferous conversations about whether automation was a prerequisite to artificial intelligence and cognitive, and various elements up the value chain, where as other people seem to talk about it as a mutually exclusive concept and framework. How do you see this developing as you look at defining the space and is there a progressive step for companies as they get more experienced with automation? Or, do you think they need to have a different approach for both cognitive and automation?

Here’s the big dependency, the whole reason to do cognitive is for inline prescriptive analytics, so what does that mean, it means that on a real-time basis you have sufficient data for a cognitive system to identify, with high confidence, what is likely to happen and tell you what you need to do about it. Our discovery has been that when you look at the data strategies necessary to accumulate the right kind of data to feed these algorithms, automation is playing a key component in creating or illuminating the information necessary. Now is that to say that there are not potential transactional domains where you have sufficient information? I think that there are some pink unicorns out there, where an enterprise won’t naturally have all the data necessary. But, much like with standard process automation there are pink unicorns out there, big banks that had 2,000 people all doing loan apps, or credit apps, or mortgage processing, these were the pink unicorns. You could build three bots, copy and paste and have 600 bots running in the course of the year. For the rest of us, and for the clear majority of where you apply this stuff, this applies where there is a direct analogue between the pink unicorns and basic process automation into cognitive and the difference is all about understanding your data strategies and to build the data fabric you are going to need to feed your prescriptive analytic engine with. Our experience has been that absolutely, automation, is a prerequisite, and we have had to do a major transformation of how we capture and store data. We had to do a full re-platforming moving into Hadoop Cloudera-based data lake, as opposed to your standard data warehouse, those are just not sufficient to fuel these cognitive engines.

Phil: Now as we look at the sheer noise around the industry – what do you think the gap is between marketing hype and reality? Is it 3 years, is it 5 years? When we even hear Gartner drinking the Kool Aid, surely we need to close the gap a bit so people are getting into a more realistic mindset and road map?

Lee: Absolutely I would say that we are probably at the basics of RPA and RDA, we are probably 18 months, 24 months from a convergence. I think in the world of machine learning and cognitive we are probably still on a 3 to 5-year delta in terms of marketing hype. I’ll just give you a personal example of working with a cognitive service provider. It took us 9 months to get the statement of work and the KPIs in place to prove beyond a reasonable doubt that it was, in fact, machine learning doing the work, as opposed to data scientists and smart people doing stuff in the background. That was a wake-up call for us in terms of how far we are away. I was in a meeting, with a very large well-known provider of these kinds of services, and I told the guy to sit down and put his magic wand away because it’s not very magical and in fact it’s a huge order to get this stuff to produce meaningful value or meaningful work in the enterprise, and I am not going to listen to the marketing hype, we really need to get down to the business of figuring out how to get this stuff to produce value. In a lot of cases, it’s expensive to dip your toe in the water with cognitive and even to get to, a proof of concept, proof of technology, proof of value. In one case we have been working on it for 15 months and we are just now beginning to see some progress. Coming back to your question, I still think we are 3 to 5 years away and I think we are going to see a convergence as big data, data services, data strategies and data science become far better understood as a necessary foundation to fuel the cognitive stuff.

Phil: Right… and who is going to work with clients to help them get there, Lee? Is it going to be the current crop of sourcing advisors? Do you think today’s service providers have the right mindset and commercial models to get clients over the finish line, or do you think different players are going to emerge in the next 18 months?

Lee: That’s a fantastic question Phil, and here’s what I would tell you, what we are seeing is the current crop are not equipped to do this and there are a lot of one trick ponies out there. Process automation is great but the world of cognitive is a totally different domain, different skill sets, different technical competencies, and it’s far more IT intensive than process automation. There are some players out there, if they are smart about how they evolve their organizations and you could probably pick them out, these are people that not only provide advisory and consulting but also technology based solutions. I think they will have an upper hand in terms of being a credible guide, and advisor to buyers in this space, but I think for the folks that are really focusing on the basic process automation today, it’s going to take a significant re-tooling to move into being an advisor in this cognitive space.

Phil: Final question, Lee… if we were to anoint you the ‘King of Automation’ for one week and we granted you one wish, what would that wish be?

Lee: It would be to solve my data needs. Everywhere we look we are finding that access to data, transitional state data, illuminating dark data, information converted into data, or vice versa, it’s one of those things – if I could solve that, then all sorts of horizons open up on the cognitive space and we are finding out that’s just a journey where you fix a few things, you try again, you add some more, you try again. There doesn’t seem to be a well-understood approach about how you think about a knowledge domain and what data you need to make it work, so if I had one wish, and I could rub the Genie’s lamp, then I would want to solve my data needs.

Phil: Thank you very much, great answer, Lee 🙂 You’ve been a great friend to HfS and am sure many of the folks reading this are looking forward to seeing you at the FORA inaugural session in Chicago this September (see link for more details).

Posted in : Robotic Process Automation