We recently caught up with Infosys to discuss where its Infosys Validation Solutions (IVS) testing practice is currently investing. This is a follow-up to a similar discussion we had with Infosys back in July 2016 that centered on applying AI and making sense of the data that client organizations have (see here).

Our most recent discussion looked at technologies such as AI, chatbots, and blockchain. The focus of IVS has expanded from immediate opportunities within software testing to Infosys’ overall development of new IT services offerings.

AI: more use cases are the priority

AI remains a priority for IVS, with the attention to date having centered on developing use cases in test case optimization and defect prediction. Its PANDIT IP correlates software new releases with past defects, feature changes, test cases, and determines what part of the new release’s code is responsible for defects. IVS points out that its implementation (in identifying the lines of code responsible for the defect) is relatively difficult. IVS is taking a gradual approach, and starting with COTS, the underlying rationale being that new releases of COTS are much more documented than custom applications: identifying the part of the code that is responsible for a bug is therefore easier and is likely to be in the custom code of the COTS.

Chatbots: testing response validity

The use of chatbots/virtual agents challenges the traditional functional testing model, which largely relies on a process, and on executing a test case (e.g. a user tries to login to a website), and to make sure the transaction outcome is valid (e.g. user is indeed logged in). With chatbots, the goal is not so much about process testing, but lies in response testing, for example:

• Interpreting questions correctly
• Dealing with the wide range of expression options end-users have for the same idea
• Selecting the most appropriate response from a high number of potential responses.

Of course, as with any ML, this requires multiple iterations with SMEs for the virtual agent to learn, in addition to using language libraries; this is a work in progress with early PoCs with clients.

Blockchain: integration complexity & business rules testing

The complexities with blockchain are different from those with chatbot testing. With blockchain, as with IoT, the complexity lies in its principles: a decentralized architecture, and many parties/items involved. IVS is assessing how to conduct testing around authentication and security, communication across nodes, also making sure transactions are processed and replicated across nodes.

Looking ahead, there will be a challenge with functional testing, in testing the underlying business logic/ rules, while also complying with different local business regulations, and languages. IVS is developing approaches to validate these contracts and is in early phase of PoC with clients.

Conclusion: the challenge is to automate testing of complex software at scale

The challenge of testing chatbots and blockchain, also IoT, and physical robots, is not so much about effective functional and non-functional testing but about moving the testing of such technologies to an industrial level, using automation software that only exists partially today.

The good news is that the ecosystem of testing start-ups is vibrant, and larger software testing services providers like Infosys are investing now in preparation for the surge in adoption of such technologies.