By Nutan Bhattiprolu

From the Leaders’ Desks

Artificial Intelligence (AI) will transform companies, industries and businesses. That is the future, and not a very distant one. Having said that, it’s not current reality either. At least, not yet in a complete way.

AI is a broad area of knowledge which is delivered by numerous promising technologies, tech stacks, and companies. For a business user it can be tedious, time consuming and possibly even never ending to explore the technologies available and plan for the right ones which fit the need.

The following few paragraphs are aimed towards business leaders to help them craft the right AI strategy for their organization using benefit driven approach, sans the hype. This essay is organized into two broad sections, first section aims to bucket the various AI approaches & technologies based on a “value delivered to business” standpoint and the second section offers mandatory enablers for AI to succeed and thrive.

The following classification attempts to evaluate the ROI which AI applications bring to the table based on qualitative factors like maturity of underlying technologies, time taken for benefit realization, amount of effort needed to be invested, possible end applications etc. Understanding AI technologies through the value delivery lens and planning to be on top of the wave when it comes to transform your industry is the key to long term survival and success.

  • Automation/RPA – This is the lowest hanging fruit, but has the highest short term ROI.  This new avatar of business process automation is most usually referred to as RPA or robotic process automation and is its smarter cousin. The robots in RPA usually function from a central server & interact with applications the way humans do – like clicking buttons, opening apps, filling up forms etc. They are also capable of tasks which are out of the reach of average BPA like parsing documents to identify invoice numbers etc. Equipped with these capabilities RPA can potentially transform many clerical back office activities and deliver higher efficiencies.
  • Analytical Insights – Analytical Insights need data science & a good dose of business inputs to be successful, compared to the tasks in RPA which is primarily smart automation of operational tasks. Activities like forecasting inventory, identifying trend patterns, insights which are useful to the business form a valuable and main part of the Insights initiatives. These are delivered by machine learning algorithms, statistical analyses including predictive models, learning and classification algorithms extending up to deep learning models. The value can be immediate with insights, or it can be delivered over a period of time as will be the case from learning algorithms which get better over time. A key aspect to consider with this is the seamless integration of model outputs and insights into the usual business workflows. This is where basic Auto Natural Language Generation (NLG) or interactive visualizations might prove extremely beneficial.
  • Human Interfacing AI – Intended to replace human beings one day, these are usually at the consumer interaction end and are capable of conversations – in roles like customer support, tech support etc. Facebook & Bing tried these technologies unsuccessfully & Google Assistant uses it to some degree of success. The technologies might not be fully mature to deal with independent or client facing conversations, these will become better over time with evolving natural language processing capabilities. Even with some downsides they should definitely be developed for internal use and can be worked upon for improvement to make them customer facing over time. This will ensure all the knowledge in the organization is consolidated into a smart repository which will accumulate benefits over time , as these algorithms learn and leverage the increasing base.

There is no one size fits all approach. Factors, like maturity of the industry, type of company, need to differentiate from competitors, dictate the decision on the blend and the amount of tech investment needed in each of the above categories. But as a general thumb rule – planning for incremental steps on high end ML with mature technologies like RPA, Insights acting as support struts will bolster the org strat plans today and lay strong foundations for a sustained value delivery mechanism for the years to come.

Irrespective of the technology mix chosen, the facilitation of the following factors is an absolute must for a thriving AI practice. These factors can act as blockers or enablers based on how they are treated and applied. Business leaders must ensure that their organization leverages these to their advantage.

  • Enabler 1 – Culture & Process Redesign

For AI approaches to be successful, business units should co-operate for collective action. During the advent of digital media, the most effective teams on Twitter were those who were empowered to interact across the business units – retail store fronts, customer service departments, client DB teams. Similarly AI offers benefits across the org like 360 degree customer view which can be achieved and consumed only via an integrated effort. Teams should also be open to taking insights from, say a data science team, without treating it as “outsider interference”. The processes should support this giving data and taking insights too.

  • Enabler 2 – Data and related Infrastructure

Usually business processes generate data as a byproduct which gets accumulated in silos. Analytics, on the other hand, thrives with deep /wide data, even if it just means bringing up previously unseen cross tabs across data sets. Data collection done with a purpose is sharper in delivering value than data generated as a byproduct.
Also the infrastructure in the organization has to allow access & merge of various data sets seamlessly – A siloed MS Access database of the customer service team & the SAP systems of the finance team have to be equally amenable for sharing and merging with a ML model.

  • Enabler 3 – Starting from scratch? – Always get a thought leader first

Not a scientist or an analyst or a technical team lead, get a thought leader who can transform a business situation into an analytical use case. A leader who can carve out a path with a pipeline of projects would be crucial for a thriving analytics team and value adding to the business. This is not to say execution is of the least priority but if the plan is to build a team then the first hire should be someone who can bridge both sides – business and data science, because execution is something which can still be outsourced and overseen but a business problem to use case creation has to be internal to really shine.

  • Enabler 4 – Top down – Business first

This can’t be stressed enough but AI & analytics initiatives have to be top down and have to be incorporated into the strategy tagged to a business benefit. Bottom up approach / free hand exploration of data are welcome but should not be the mainstream way.  A non-directional insights initiative will deliver value which might or might not be consumable by the business. An initiative driven top down also engenders a collective vision for all the teams involved.

  • Enabler 5 – Integrate & Share, but Carefully

AI brings together various elements – business data, public sources, internal teams and external partners. There will be real value in sharing data for analysis & monetization but creation of frameworks and agreements which take care of privacy of the data , anonymity of parties involved and further propagation or use of the data or the insights would be critical.

  • Enabler 6 – Always Pilot

An internal pilot for any large project, especially if intended to face the end customer is always a safe bet. If a customer service bot is being built to support or replace the customer support team, build & deploy as an internal help desk bot which can support internal FAQ for prolonged testing under near real life scenarios. Pilots will warm up the teams for usage, allowing them evolve mechanisms to cope with this new supporting bot colleague and will aide easy adoption and further advocacy.

Here is how an ideal strategy blend would look like with 50-60% of the portfolio delivering immediate value starting from as quick as 3 months and the rest of the portfolio structured to deliver long term value appreciation.

The right strategy mix is one which enables the organization to incrementally embrace AI and builds a robust value delivery pipeline for today and tomorrow.

Please let us know what you think