Unpacking Your Dark Data To Democratize Your Digital Transformation
Add bookmark
Manish Rai
Vice President of Product Marketing
Automation Anywhere
The first wave of automation in enterprise has been focused on structured processes; repetitive tasks, they follow the same path, using structured data. That has brought tremendous value to enterprises. But only about 20% of the data in a given enterprise is structured. 80% of the data is unstructured. We’re talking over here documents, such as invoices, market applications, and
explanation of benefits. Emails coming in stating, “What is the status of my invoice?” or “What is the status of my order?” That’s just the beginning of your unstructured notes and service tickets.
And there’s where the market is headed. The next level of value is being derived by attacking the semi-structured data of this tremendous opportunity around documents such as invoices and POs. Sometimes invoices and POs represent close to 50% of the volume of documents inside enterprises. So we are seeing a tremendous amount of traction around structuring those documents.
The next thing I think we’ll see next year is not only in solving unstructured data, but looking at the other dimension — making unstructured decisions. Unstructured decisions are predictions made after observing humans and the sequencing of tasks. Cognitive technology will be getting to a point where it’ll be able to predict the next best action and recommend the next best action.
Cognitive technology will be able to take data from decisions people are making. Think about decisions people make around, approving market applications for example. Once the cognitive technology has observed hundreds of those decisions, it can build a model and start recommending whether the market application should be approved or not, based on the models.
McKinsey estimates that AI will deliver $13T in economic activity by 2030. So, when you look at what the market size of RPA is today — even though it’s growing significantly — we’re looking at a billion or two, and we’re merely scratching the surface. So, we have just gotten started on this journey in my experience. There are a tremendous amount of efficiencies that will be gained by enterprises in the coming five to ten years as the AI technologies mature.
Let me put that in context. We have a customer — a large technology firm based here in Silicon Valley. They came to us to try and automate their order processing. They said they have tremendous radiance in demand, like many technology companies. More than 70% of their orders have come in the last few weeks. They have 50 full time employees processing orders. Even that number of employees was not sufficient to meet the year-end demand. They took our cognitive product — IQ Bot, to try and understand these orders, and extract information. Every vendor has a different format. So, it’s a very difficult task. In about six months, they have been live with us; they have automated, end-to-end, 40% of the order volume.
If you can think about it, if they have 50 full time employees — 20 FTEs that have been re-deployed to other work in just six months alone. In the first five weeks, they had attained ROI as they were able to automate close to 50% of the deal volume. And now they’ve gone to multiple times return on investment. So, hopefully, that sheds some light on how big the opportunity is, and how quick the ROI can be realized.
That’s one enterprise example. In terms of realizing ROI, other firms might not be as quick if there are issues around change management within the company. It does depend on the organization; each one approaches it differently — but the opportunity is tremendous.
Executives might think that cognitive AI can do anything, and they think of highly complex processes to automate. That’s when we see the people get disappointed, and the projects fail, because you want to take baby steps in your cognitive journey. You’re doing RPA; this is the very next step. We recommend to people is pick high ROI processes - and invoices and POs are great way to start. You can explore Explanations of
Benefits or Standard Settlement Instructions - the kind of documents we’re seeing that are really good places to start for people.
Then, grow from there and attack more and more. The key is to look at 80/20 rule - sometimes it’s 20% of the customers who generate 80% of the orders. Or it might be 20% of the vendors who’ll do 80% of the order volume. Go after the low hanging fruit and show high ROI quickly. Get by and then deploy it enterprise-wise.
If you’re further along on your cognitive technology journey — you’ll want to benchmark yourself against best-in-class companies.
We have actually put together benchmarking tools that help enterprises benchmark themselves against how best in class enterprises are doing. We have strongly guided recommendations. If we’re truly to make this transformation successful, business and IT have to work hand-in-hand. IT needs to put the right level of controls, governance, and ensure the deployment is secure, and business needs to be empowered to build the bots on their own, to attain the velocity they need to drive the transformation.
We’ve seen iterative CoE’s work. First having a center of excellence, maybe for finance — then adding HR, then having a corporate-wide center of excellence that provides governance. Think of it as though enterprises need to be moving toward the app store model. The Google and Apple app stores are good models — where anyone can build an app or, in this case, it would be that anyone could build a bot.
But when your talent submits the bot — there is a process of governance they can look at, and say, “Hey, is the Bot secure? Does it meet our standards?” Then you have a quick turn-around of 24 hours to 48 hours to approve the bot and get it out into production. So, they have to move to that agile, app store-type model to get everyone to participate in the transformation. It’s about democratizing the digital transformation.
