Unleashing information intelligence with Info Insight

By Bill Dawes

Kodak Info Insight Platform is a new solution platform from Kodak Alaris that employs Artificial Intelligence (AI) to help organisations deal with the challenge of managing a multitude of inputs from different communication types, such as email, traditional mail, social media, text and online forms. Kodak Alaris' General Manager, Software and Solutions, Roderick Hughes recently travelled to Australia to outline the capabilities of Info Insight to major enterprise, government and BPO customers in the region. IDM editor Bill Dawes asked Rod to explain the business proposition.

IDM: Rod, this looks to be a very different product to Kodak’s traditional Digital Imaging business?

RH: Kodak Alaris has a long history of developing document scanners and the ability to take a really good picture of a piece of paper. About 15 years ago we saw the need to expand into the next step of most business transactions, the capture layer of document processing.

The natural evolution of our business has been to look at how can we move beyond (a) just paper and (b) add more intelligence and more value to the overall transaction.  Where we saw the gap was the explosion of unstructured information as a whole.  

For every chat, every email, every postal letter that somebody sends in, the average enterprise has trouble in knowing how to handle that consistently, and how to eliminate the enormous amount of labour it takes to read and understand and last but not least respond to those types of communications.

Email is completely unstructured and for most firms it’s a problem.  

The traditional call centre often has to create a separate silo of labour to process and respond to emails.  Then you have the problem of people getting a different response than the phone centre personnel would give, and in this case it’s in writing so therefore it’s a little bit more legally binding and a little bit easier to follow up on.  

Social media continues to grow and creates another silo of people in contact centres reading tweets and responding.  

The challenge in a contact centre is how to cost effectively handle all of these different inputs and be able to manage them and provide back to customers a timely and consistent response.

With Info Insight they can have the software read and understand any of those inputs and begin quickly to classify it.  What is it?  Is it simply a change of address request, is it something more than that, a customer complaint or accolade?  And then depending on what type of request it is we can assign it to different business processes.  Often what this will require us to do is integrate into your existing strategic systems, your data warehouse, your ERP, your CRM, and enrich that input.

For example if a prospective customer was to write “Can I buy the iPad in blue?” we can see who you are, we can understand what you’re asking, and by that input we can marry it up with the static information in your firm using that same artificial intelligence ability to read and truly understand language presents your contact centre agent with an approved and consistent response no matter what the input.

When you have a customer that’s impatient and hitting you four different times via four different channels, that’s often 4x the cost for a contact centre.  We can unify all four and provide one consistent and timely response back to the customer across all four channels that they came in which lowers the cost to the parent firm by 4x in that scenario.  

IDM: What is the competitive landscape for Info Insight?

RH: Other players in the market may offer capabilities that partially overlap with what Kodak Alaris can provide, however, none cover the full spectrum of text based documents and communications as we do.  Out platform’s ability to actually interpret and understand natural language provides superior assets.   So we’re not picking key words or a series of key words and wrapping rules around them.  

The Kodak Info Insight Platform makes use of virtually all available algorithms and adds several of its own to arrive at the best possible and most reliable result. In addition to being able to understand natural languages, the platform can also apply continuous self-learning to improve and adapt. So it learns the decisions it’s made up to that point and if it comes across a customer interaction or a document that it hasn’t come across before, you need to give it instructions.  So that’s the only time a human would be involved and this would be your super user.  

Once an expert user has made a decision or guided the AI in terms of how to handle that particular document or interaction the system remembers that so the next time it comes across a similar case it doesn’t have to go through that expert user again.  So you’re continually building this knowledge and capability.

This leaves you able to automate a lot of your very simple cases and that means your staff, which are more expensive in a contact centre environment, only have to deal with really high value or complex cases.  That also means as a contact centre operation you’ll be able to scale up or down your operation.  Because typically when you look at call centre or contact centre operations they’d be resourced based on the peak volumes that they anticipate.  What that means is they’re spending on resources that they probably utilise only about 60% of the year.  So this allows you to optimise that as well.

IDM: The AI technology that you’ve developed to do language processing, where did that come from?

RH: Kodak Alaris is strategically involved with a German firm that is strongly aligned with the University of Koblenz.  Unlike almost every document capture vendor that grew into intelligent document recognition by figuring out how to read data from structured forms, this company’s founder chose to tackle a much greater  challenge and started by figuring out  how to build a way to handle unstructured information in emails with which they approached  a major German airline.  At the time they thought email wasn’t such a big problem yet; “fortunately”, dozens of emails a days have grown to many thousands, necessitating an intelligent solution that happens to work just as well with structured and semi-structured content. The technology was quickly applied to eliminate the time consuming task of reading those emails, figuring out what the problem was and then routing it to the right person who could solve the problem. Going far beyond mere routing, today’s solution actually provides suggested responses to the agent so that the entire process of responding to the customer emails can be reduced to just a few mouse clicks. 

That allowed the airline to continue to grow as new mediums came out like social media, like chat and like everything else, their operation was able to remain consistent and scalable because no matter the input, no matter the channel that the input comes in, the people who are well trained and understand how to respond are able to do that all in the same way, whether it’s a phone call or any type of other input.

IDM: Can’t firms do the same thing using Web forms with menus that route stuff where it needs to go?

RH: I would say it’s a Band-Aid solution and the reason I say that is twofold.  One is most customers get really frustrated with those forms.  If it’s the case where you’ve simply lost your luggage and you know all of the information the airline’s requesting on the form, what was your ticket number, your route, etc. they work okay.  Most customers or a good subset of customers either hate the forms or the vast majority of the information that they put in is at the very bottom, in the comments or the “describe your problem” field.  That area still needs human intervention to read, understand and figure out in that case where to send it. So behind those forms there is simply lots of people for most industries.  

With Info Insight we can take the comment, enhance it with the CRM, look up the flight that you were on and why your baggage was lost and present that seamlessly to the contact centre agent.  What that eliminates is the agent having to query multiple systems, then route it to the correct person to handle the problem and all of the back office stuff that occurs in a contact centre to make that work. Even so, Info Insight can handle from web forms just as well; after all, web forms can contain unstructured customer input as well. 

IDM: These solutions require a very different channel than scanners and capture software. What is Kodak Alaris’ strategy for getting into organisations at that level?

RH: Our strategy is and remains partner driven.  It will likely not be our traditional hardware or software reseller that does this, but someone who is interested in building a professional services capability with deep business understanding of customers.

The technology in the Info Insight platform goes beyond just contact centre operation. The competitive landscape today means organizations cannot afford for online quotes to be processed manually, they need to be able to respond instantly whether it’s on the Web or via social media.

IDM: It’s a very young industry that’s evolving rapidly, with a lot of players attempting to make a stake in this market and really no obvious leaders or anyone with entrenched market share in this because it’s so new.  Who is deploying this technology for automated classification today?

RH:  Some leading companies in the world, many of which are in the Global 100 and 500, are currently using this technology in their systems, either as input management to handle the first couple of steps of classify and extract, or overall in response management or full customer experience management.  Eliminating the mundane manual tasks really allows your business to grow in scale over time without having to add a lot of cost to the structure.  Furthermore it keeps employees happier because they’re doing higher value tasks versus again just sorting through a bunch of paper to see if a signature is there or not.

For one major online-only insurance company it used to take about two hours for an employee to query different systems just to generate a quote for a customer.  Now they can do that in about two minutes.  The AI can go through the different systems, read through the customer's email to understand what type of quote they’re looking for, assemble the quote, and what the agent does now is simply reads it and proofs it and hits send to deploy the quote out to the customer.  

Many contact centres have taken strategies around deflection.  Whereas most firms will post a static frequently asked question on their website and hopefully you can go find it and hopefully understand it based on your language and understanding, Info Insight can automatically present solutions.

The applications that we’re all chasing at this point is how to get the right information to the right people at the right time.  

In the past people would digitise documents purely to store them and archive them.  Nowadays it’s about extracting information and doing something with it. 

So in some ways the new Info Insight business is just an evolution of what we have always done. Paper is just one input and we treat it no differently than an email, to a mobile app, to a social media post, to a forum in a community, so all of these are what I would class as conversations and conversations by the nature of it means you need a person trying to understand what it means.  That’s where our AI comes into play.

Also the natural evolution of those big tiff archives is to use them to build analytics, to build business intelligence.  Currently the customers that have been scanning to archive are perfectly positioned to use the technology that’s in the market right now to go through and extract the relevant data from those systems and build their data warehouses in the right way.  So it’s a pivotal time to actually be able to use all of those “scan to archive” applications, and now take them to the next step to help manage your business more appropriately.

IDM: What is the technology stack behind Info Insight?

RH: Info Insight is a versatile solution powered by an artificial intelligence engine that can read and understand unstructured data, it’s an integration platform to enable you to tie into your back end strategic systems, and then finally it’s a business process design engine.  

It is built in Java and runs as a Web service in a RESTful architecture, and integrates with any of the major database players from DB2 to Microsoft SQL, to Oracle. The vast majority of our customers are on MySQL which is the open source evolution now owned by Oracle.  That allows us to be completely agnostic, whether it be database or operating system, inside of an environment.  The system considers that every business process step is a web service that’s advertised throughout the enterprise so you have the ability to hook in and out of Info Insight with any step of a business transaction.  And that gives you the flexibility as well as the database layer integrations and everything else to really use it with any modern system that has been built in the last 25 years.

IDM: Where does it sit in a spectrum with workflow and BPM technology?  Is it something extra, a replacement, or an alternative?

RH: I wouldn’t refer to it as a workflow engine in the traditional sense of a K2 or Nintex or something like that.  What we do well is document and outline a business process that involves something to do with either an input or an input management or static information and then drive it throughout the different systems inside of your enterprise.  You wouldn’t use it for your expense report internal approval, there’d be cheaper, easier ways to do that at the end of the day. What we do really well is anything that requires a human reading an input, reading information and doing a business process over and over again, we automate that so that the human doesn’t need to be bothered with it.

IDM: The business case for automating business processes largely hinges on the rate of exception processing. Are you able to estimate what this will be for info Insight in different scenarios?

RH. Never is it the case that 100% of transactions can be automated.  There is always some bit of end user validation or exception processing that needs to happen but one of the key aspects of this solution is that the artificial intelligence engine has the ability to learn from the exception process. The system is always getting better at what it does but any new input, any new question, any new type of response, will absolutely fall on humans to figure out how to solve the problem.

Also, when we talk about classification we speak about it at the transaction level.  Not at the field level, not at the character level, but at automating the entire input or, for lack of a better word, the entire document through the system.  We don’t advertise a classification rate, however, depending on the complexity of the input stream, large customers in the US can achieve straight through processing rates up to over 90%. But even lower rates can yield a very compelling ROI; think of companies utilizing hundreds of agents to respond to customers. 

For traditional scanning to a document management system we can achieve around 95% such that you don’t need any manual rekeying.   Enriching the scanned data through your back end systems is the next step and that really becomes dependent upon your CRM data.  It also depends on the threshold levels you set which depends on your confidence levels and the testing we do during the install.

If it’s a question like does the iPad come in blue you may choose to just let that go through the system without human intervention.  If it’s a question about my billing or my contract you may want to ensure that an agent is always looking at the question before you send it to your customer.

IDM: There are many well-known obstacles to deploying automated capture and OCR technologies, whether it's organisational cultural change or user resistance to workflow change. What are some of the obstacles that prevent people considering automated classification solutions?

RH:  I don’t think the obstacles are fundamentally different.  With the economic volatility of recent years firms have started to get a better handle on the cost of manual intervention in every step of the business process and begin to look at outsourcing to a BPO for rekeying and certain steps like that.  

Firms have become very aware of the cost of getting new customers, very aware of how important customer experience is to retaining customers, and therefore closely examine traditional mailroom applications and the way they manage receiving complaints. The need to serve customers more quickly and more effectively has become really, really important to the long term viability of the firm.

The traditional call centre role may disappear in three to five years’ time as more organisations adopt technology and better ways to actually handle their customer interactions.

For simple cases you can handle them with Info Insight and then for really complex cases you put that in the hands of a contact centre agent. We think the whole contact centre operation will be a hybrid combination.