Snapshot: artificial intelligence or robotics?
Artificial intelligence (AI) can now take over the steering wheel of a car, order last minute additions to the holiday gift list or find an instantaneous answer to that obscure piece of trivia on the edge of your memory. But robotics software may be more useful to fund managers.
A standard definition of AI continues to elude even the experts, but a rule of thumb is that true AI handles unstructured information with algorithms to continually improve results.
While earlier theories of AI focused on understanding semantics and adherence to rules-based decision frameworks, recent advances have come from a model based on human neural networks. So-called “deep learning” and “cognitive” AI models are being used for challenging problems like image recognition and natural language processing.
Asset managers and advisers currently use AI for intelligent trading, but this is a comparatively simple problem. More profound implementations will see AI advisors perform holistic estate planning while AI compliance officers actively manage regulatory risk across every jurisdiction in real time.
The brilliant futurist Arthur C. Clarke famously observed that “Any sufficiently advanced technology is indistinguishable from magic”. The inverse, however, remains equally true – if it’s not magical, it’s not perceived as sufficiently advanced.
Building robot users
In financial services, there are a growing collection of not-quite-AI but wonderfully useful pieces of software collectively labelled robotics, or more precisely, robotic process automation (RPA).
You can think of RPA as a virtual human user, clicking on the same parts of the user interface, going through a workflow with deviations as a human user would, and building up a library of processes over time.
In the fund operations world, a generic accounting process could look something like this:
- open business application
- open counterparty website
- reconcile ledger balances
- run ledger report extract from reporting application
- drop report on FTP site
- send email to notify recipient of report availability
The traditional IT-centric automation of this process requires significant technical expertise as the steps occur in several applications and environments. Queries, stored procedures, file monitors, transformation scripts, exception handling and scheduling services would likely be necessary.
RPA approaches this six step process from a different perspective. Instead of solving the integration problem at the deeper, technical level, RPA acts as a user. The robot will perform each step by clicking (literally or virtually) through the applicable application.
For example, opening the website and checking a value is just that: opening up a website through a browser, scanning the page for a particular value and capturing that value before moving to the next step.
Training the system
Sophisticated robotics systems can be “trained”. They perform a set of workflows thousands of times to store possible permutations in the process and learn what actions to take next. Conditions like a failed balance reconciliation are handled as a user might: by opening a ledger to see that all entries have been posted, then checking the FX rate, then checking the trades import time stamp, etc.
In fund operations, robotics software can power a variety of business processes, whether monitoring a broker website every 30 seconds to find updated data, filtering through a vast number of reconciliations to identify thresholds for human intervention or interrogating systems to gather information for KYC/AML procedures.
Artificial intelligence will continue to receive tremendous industry focus as it matures, and appropriately so, but robotics is the emerging technology that can help your organisation right now.
8th March 2017