The Role of AI in Large Scale Automation Programmes, C-Suite Guide (Part II)

Rage Office
Rage AI – Enabling Financial Institutions to Unlock The Power And Opportunity of AI
February 8, 2016
AI-for-enterprise-ecosystem-map
EXCLUSIVE: The first ever AI for business landscape map!
February 24, 2016

The Role of AI in Large Scale Automation Programmes, C-Suite Guide (Part II)

Process Automation on the Mechanism of Metal Gears.

Introduction

Part I in this series which was provided by Virtual Operations discussed the best practices and principles for managing automation transformations across an enterprise. This instalment describes the operational framework involved in the application of these practices.

Businesses continuously struggle with managing volumes of data – receiving, processing, recording, analyzing, and reporting. This is where a well-executed automation programme can return unprecedented savings and benefits for an organization. Many media forums have misled the masses to believe these tools are new arrivals to the market. However, these tools have been alive and in production across many businesses today. Many have also been grossly exaggerated by science fiction enthusiasts and many uninformed sceptics who have their own agendas to promote. In reality, well-orchestrated automation programmes have consistently proven their value in industry for many years, and will undoubtedly continue to.

With the wealth of automation technologies available, it can be challenging to determine which of these are best suited to effectively address BPI objectives. With the bounty of misinformation inundating the marketplace, it is understandable there are so many misconceptions and suspicions about their definition, intent, and benefit. Clarity on these points is needed, and then we can explore how these technologies can be leveraged and synergized to provide a robust, imperforate, and transformative solution to address the ubiquitous and costly inefficiencies obscuring your business profitability.

Artificial Intelligence

When asked how to resolve war, Confucius, the ancient Chinese philosopher, replied “I’d insist on the exact definition of words.” An abbreviation, which covers a very broad spectrum, “AI” is a difficult term to both qualify and define.

Arguably the most progressive field of computer science today, AI is often used synonymously with a variety of terms – cognitive computing, automated intelligence, pattern recognition, machine learning, predictive reasoning, etc. Regardless of the semantics, it’s a collective term that describes a subset of technologies, or tools, each with its own capabilities, limitations, and maturity. Familiarity with this suite of tools provides the cornerstone of an automation strategy.

Managing a disparate array of incoming data streams from paper, fax, scan, email, voice, social media, etc., many AI tools are highly adept at data capture. These tools employ some method of pattern recognition to the sampled input stream. Examples of these include discerning an object of interest from an image, written transcription from a voice call or verbal command, search and data aggregation across common themes, fingerprint and DNA sequence identification, character recognition from cursive script to text, and many others. The ability to identify and isolate these patterns from raw, unstructured, extensive amounts of data has countless applications in business, science, engineering, research, medicine, and a host of other industries.

Some divisions of AI such as machine learning, cognitive computing, and natural language processing are evolving areas of computer science. Patterns are captured, classified, and compared against very large libraries of data, which is a very time/resource-intensive process. Using complex algorithms, these platforms aim to model the cognitive, learning, and reasoning abilities of a human. This is a monumental task, as computers don’t natively understand human (natural) language, expression, behavior, or emotion. Increasing investment of capital and talent to these areas within the last decade have produced some note-worthy achievements, but many challenges remain before these reach maturity in business markets.

Let’s put some meaningful context to these concepts by exploring a common enterprise example:

Human-Free Accounts Payable Example:

  1. The business receives a digital pdf invoice into an allocated, process-monitored email inbox.
  2. The incumbent email triggers the automated process, which creates, queues, and tracks a new invoice work item.
  3. The invoice field level data (supplier account number, PO number, net amount, etc.) is automatically extracted from the pdf metadata.
  4. The field level data is captured for further action and passed downstream to the waiting RPA process.

Robotic Process Automation

RPA doesn’t attempt to model the cognitive or abstract process that resembles human intelligence, and any argument to the contrary is emphatically false. Instead, RPA models the actionable tasks of a human business user. Essentially, a virtual agent is engineered to perform the low-level executable duties of a human full-time equivalent (FTE). These duties are traditionally discharged by clerical, repetitive, rules-based processes across organizational business units.

Data entry, calculations, record-keeping, correspondence, reporting – in fact, most administrative or operational tasks that are enforced by established governance are viable candidates for RPA. Each of these categories have the potential to produce a magnitude of error-prone and time-consuming inefficiencies. RPA tools mitigate the inherent risks, which create very costly disruptions in quality, customer service, compliance, security, and profitability.

RPA uses a virtual software “robot” to identically perform the same activities of a business user. This virtual agent models (through software programming and rule-based criteria), the same series of actions needed to complete a given unit of work, or process. For example, this could include identifying and managing incoming/outgoing data streams; navigating, accessing, and interacting with applications; performing lookups and transposing data; making calculations and decisions; managing and engineering correspondence and documentation; and routing to a human party when necessary for exceptions outside of the robot’s programming.

Among its most notable advantages, the virtual components of RPA allow it to scale very easily on demand to manage seasonal workloads with 24/7 operational continuity, its architectural footprint is relatively non-invasive to existing IT infrastructure because it reuses the existing application framework utilized by a human business user, and it yields to significant reductions in force without compromising revenue margins.

Human Free Accounts Payable Example (continued):

  1. The listening RPA process takes the captured data elements (step 3) as process inputs.
  2. In accordance with the process definition (rule-based criteria), the process automatically:
    • Performs lookups with other line-of-business systems
    • Performs each actionable task in the (manual) process workflow to execute or triage the invoice with same capabilities of human FTE
    • Suspends operation and delegates to human FTE where unhandled exceptions occur
  3. The process records and reports the results back to the process owner.

Data Analytics

Data mining and data analytics can also be used with automation platforms to perform statistical, historical, and predictive analysis. Big data can be consolidated into useful information, providing valuable perspective to previously obscured patterns hidden in copious amounts of data.

Used primarily as an enterprise reporting tool, data analytics drives the key performance indicators, dashboards, and multi-dimensional reports that are used by strategic business leaders in an organization. These decisions frequently involve chronological, comparative, and trend analysis which requires large amounts of data to be aggregated. Transactional, line-item details – stored across multiple tables, in multiple disparate databases/data warehouses, across multiple disparate geographies, doesn’t provide the needed visibility to key decision makers –we need a ‘bigger picture’.

Human Free Accounts Payable Example (continued):

  1. A data analytics report retrieves the archived and current data of all paid invoices by supplier, service type, cost, region, and year, over 3 year period.
  2. The report exposes a higher cost per same service delivery for suppliers A & B than for supplier C.
  3. Leadership decides to promote supplier C to the preferred supplier, renegotiates reduced rates for supplier C based on promoted partnership and increased service delivery volume, and demotes suppliers A & B for contingency only.

Balancing Technologies

Process improvement across many organizations has attempted to use “piece-meal solutions” to try and remedy business process deficiencies. This may appear to be the most cost-effective strategy, but this limited scope often negates the contributing and influential factors upstream which arbitrate the negative results downstream. Rarely is there single solution or tool to address the entire business landscape.

However, there are synergies between these platforms that, when used correctly, yield unprecedented results and extraordinary competitive advantage. It’s tangible (and inevitable) to have fully-automated virtual workforces to manage complex and long-running business process workflows, with little or no human intervention required.

To summarize the previous example, Virtual Operations has worked with enterprises managing multi-million invoices per month. Successfully implementing an automation programme for this Finance & Accounting business unit provided a leaner and more manageable operating business model, and contributed significantly to enhancing revenue margins as influenced by the following factors:

  • Cost Savings
    • Automated end-to-end invoice process resulting in significant FTE reductions
    • Invoice processing time to payout/posting significantly reduced
    • Preferred supplier SLA discount * less expensive cost of service
  • Compliance
    • Automated process mitigates propensity for human mistakes caused by miscalculations, typographical errors, oversights, and inexperience
    • Improved quality throughput and regulatory obedience
  • Scalability
    • Virtual workforce matches fluid demand
    • No costly disruption to IT infrastructure
    • Virtual agents are immediately proficient and require no training investment

While this use case was centralized to Finance & Accounting, these merits are obviously not confined to any specific vertical of industry, nor is it a ‘one-size fits all’ proposition. This framework is a guiding principle meant to illustrate the collaborative nature of the many automation tools in the marketplace and their general application in many automation programmes:

  1. AI tools gather, consolidate, and capture key data elements (numeric, text, imagery, etc.)
  2. These inputs feed into RPA-managed workloads, automating the FTE tasks
  3. Data analytics feed KPI metrics to strategic leaders, who make informed decisions

Without a measure of doubt, Automation has transformational impact across the business landscape. To ensure success, this impact must be measured and managed correctly. As stated in part I of this series, “Sustainable transformation and the creation of real shareholder value are truly attainable.”

But no single ingredient makes the recipe. It requires the right strategy AND the right methodology, the right talent AND the right tools. At Virtual Operations, this is the experience we’ve gained, and it continues its purchase with each new large scale client automation programme we deliver.

“Think big, move fast and get organized!”

 

Leave a Reply

Translate »

We use cookies. By browsing our site you agree to our use of cookies.Accept