Artificial Intelligence – Let’s get specific

Article

Artificial Intelligence – Let’s get specific

As futurist opinions on the impact of artificial intelligence reach fever pitch, the real question for business is what do we do right now? Alan Marshall, lead partner for analytics at Deloitte, looks at how firms should approach the landscape.

General purpose artificial intelligence - think Iron Man’s Jarvis - that can successfully imitate human behaviour is an exciting prospect, but it is still far from a reality.

However specific artificial intelligence is real.

Specific artificial intelligence is when computers mimic certain mechanisms of the brain or activity that we previ­ously thought was the domain of humans only; for example, processing language and images. This capability is readily available and is broadly speaking called cognitive comput­ing.

The rise of cognitive computing along with robotic process automation (RPA), big data and analytics, provides access to a new ’robotic’ workforce that can redefine how financial services functions can achieve significant perfor­mance improvement.

 So let’s get specific

To best meet the disruption financial services face on many fronts, ‘intelligent machines’ should be at the heart of how institutions respond.

So where can we put this ’robotic’ workforce into action in financial services?

1. Customer service – intelligent machines provide an opportunity to personalise customer communication, en­able self-service, and automate service fulfilment. For ex­ample:

  • Deploying ’Cognitive Agents’ to work ‘on the shoul­der’ of humans, or directly with customers to answer queries and execute service requests. More advanced than the current crop of ‘chat-bots’, cognitive agents like IPSoft’s Amelia are able to maintain short term mem­ory in conversation, access knowledge repositories to answer questions, and automate process execution;
  • Personalising communication to customers using natural language generation engines that interpret movement in financial markets, assess the impact on a customer’s investment portfolio, and automatically generate a personalised message to keep the custom­er informed and/or provide suggestions on trading opportunities; and
  • Personalising omnichannel marketing and service conversations using recommendations generated by complex analytic models de­ployed in real time decision engines. These decision engines centralise intelligence, ena­bling consistent personalisation at every ser­vice event as well as proactive intervention on customer experience to address problems before they escalate.

2. Automating internal processes – use of intelligent machines to automate repetitive, knowledge and natural language rich process­es, for example:

  • Using basic algorithms to automate ma­chine-to-machine process steps in shared ser­vices functions delivers significant reduction in manual effort and increased accuracy in both process execution and data entry;
  • Adding cognitive capability e.g. natural lan­guage processing, image recognition, and ma­chine learning to algorithms to automate bank­ing and insurance operations. This includes application and claims processing, customer no­tifications, renewals, and product administration, reducing manual effort, increasing both the ac­curacy and faster response to customers; and
  • Automating compliance and business reporting using natural language gener­ation engines that ingest data from traditional business intelligence systems and produce text for summary in­ternal reports. Compliance checks can be codified into these engine reports to widen the scope of use to in­clude reporting for regulators.

3. Deepen insight – cognitive analytics detect patterns and relationships in unstructured data as well as under­standing context, to explain why the patterns and relation­ships are important , for example:

  • Using cognitive advisors like IBM Watson to pro­vide recommendations to financial planners / advisors on suitable products for clients. They can also answer specific questions clients may have, drawing on accu­mulated detailed knowledge of a wide range of financial products;
  • Using natural language generation software to col­late customer data including known needs and prefer­ences to generate a meeting preparation report. This re­port can provide advice in natural language to the sales teams on current opportunities and how best to sell to this customer’s needs; and
  • Using natural language interfaces as the gateway to consuming complex analytics on both structured and unstructured data. These interfaces will remove the in­termediary (e.g. analyst) and allow decision maker to rapidly obtain insight and increase agility in decision making.

How do we do start assimilating?

Now let’s look at how we start assimilating machine capability into organisational functions?

Traditional technology implementation follows a production line path: establish requirements; design to meet requirements; build to meet design; test and rollout. Cognitive and process automation technology however has ‘training’ at the heart of its approach. The key steps include:

  1. Establishing a process set or domain in which the technology operates;
  2. Completing an initial upload of data or define initial algorithm;
  3. Training the technology through repeated execution of tasks or response to questioning; and
  4. Monitoring accuracy, which should improve as the machines learn through the feedback provided, primarily by humans.

In parallel with the processes above, the function into which the ‘robotic’ workforce will be deployed needs to be prepared. Initially this will be focused on process redefinition and / or updates to role and responsibilities. However, as the disruption to the overall process ontology reaches a tipping point, it becomes necessary to redesign the function to scale the impact of the ‘robotic’ workforce.

What about the human workforce?

No doubt the future of human work will change sig­nificantly as machines continue to assert their competitive advantage in executing routine activities and identifying patterns in data of all shapes and sizes.

However, we expect to see mass ‘augmentation’ rather than mass ‘substitution’. The future of work will be defined by partnerships between the human and the ‘robotic’ work­force.

These partnerships can create more than just efficiency – these partnerships have the potential to deepen our work and produce outcomes we previously thought were not pos­sible. They will be able to create new jobs in fields we had not previously imagined, and allow humans to focus on work that is more fulfilling.

This article was first published in Asia-Pacific Banking & Finance.

Did you find this useful?