open langage menutoggle menu
All News

Cognitive APIs: the future of RPA and Smart Automation?

In order for RPA – Robotic Process Automation – to be more efficient, it needs to be combined with Artificial Intelligence. And if there is a need for greater speed, cognitive APIs can be added, leading to considerable time-savings in the development stage.

RPA, Robotic Process Automation, is aimed at automating certain business processes in order to free up time for the teams. However, it rarely operates alone. Most projects of this kind are actually Smart Automation, a discipline midway between RPA and Artificial Intelligence (AI). “Already 80% of our projects are no longer strictly RPA projects, but in fact Smart Automation. And in the near future, that figure will rise to 100%,” explains Philippe Lerique, Partner in charge of AI, Data and Digital Transformation Projects at Talan Consulting, co-facilitator of the Talan webinar (in French) “cognitive API and RPA”.

To integrate this intelligence into apps, the simplest way is to add so-called “cognitive APIs” to existing RPA applications.


It’s raining APIs

As a reminder, API stands for Application Programming Interface. A cognitive API offers a valuable service using AI, of which other products or services can make use - without needing to know its implementation details. An API can for instance retrieve weather data, a contact person’s location, certain web content, etc.

Let’s take the practical case of opening a bank account as an example:

  1. It all starts with the application review, which gives rise to a score.
  2. If the application is accepted, an automated evaluation of the supporting documents follows.
  3. Then, if all goes well, the applicant is assigned an account number.
  4. The bank then sends the new customer a personalised confirmation letter, a welcome pack and a credit card.

In steps 1, 2 and 4, the existing cognitive APIs are sufficient. “This is what institutions such as Boursorama Banque, Orange Bank and CA-CIB do very effectively.  There are thousands of applications using APIs of this kind in their RPA and demand is soaring,” says Philippe Lerique.

API covering functions as diverse as opening and understanding an email, responding to a telephone message, analysing feelings, detecting fraud, reading identity documents, consulting, completing and optimising agendas, preparing an action file for a site with orders for the equipment involved, etc. The sky is the limit – provided that technology can support it.

It is relatively easy to create ambitious application models integrating these APIs, in particular by making use of the AI in their learning functions – the cognitive API is trained to recognise certain fields of a form, for instance.


Which API, and at what cost?

The question then arises: how can I choose the right service, given that there are more and more APIs available?

There are many answers. To name but a few, there are proven service libraries in Microsoft’s Power Apps or UI-Path (the market leader). Some players have also released quality APIs that can be used right on-line: Wirk, Quick Sign, Vonage, etc.

It remains to be seen what price companies are willing to pay to use such an API. Many factors come into play. The supplier’s reputation for reliability is certainly important. But customers will look for more. Some “ready-to-use” APIs are called upon for a given function, while others are more summary and will require significant internal development. The “top of the basket” consists of APIs that are not only ready-to-use but also equipped with support services in the event of trouble – such as manual support services offered by Wirk when automatic data recognition does not succeed.

Points to consider include parameterisation, built-in security, the ability of the API service to meet high demand, and learning management – including knowing where data are stored to continue accessing it even after changing providers.


Why preference should be given to short contracts

Once the API set-up costs have been taken into account, the operating costs vary depending on the volume of data involved, the frequency of use and the time range envisaged, the number of calls to the API, the duration of calls, etc. Philippe Lerique points out that there are evaluation grids for each of these factors.

One final point: should customers commit for the long term, when choosing to adopt an API? “Things are moving so fast in this area, and APIs are improving so swiftly, that I would recommend short-term contracts!”  advises Philippe Lerique. “It is good to know that the usage price of certain cognitive APIs has dropped off sharply, including in recent months. This means customers can avoid being tied hand and foot to their solutions...”

Through all these considerations, let us remember to stay focused on the essentials: the use of cognitive API leads to considerable time savings in development.