Cognitive Automation (CA) employs Artificial Intelligence (AI) and Machine Learning (ML) methodologies that often deal with understanding ambiguous inputs such as natural language texts or deriving insights from historical data to optimally predict an output by minimizing a cost function or maximizing a likelihood function. Increasingly, CA is being applied in many industry sectors. However, there are some unique challenges that exist. Firstly, data curation for generating training data is a non-trivial data scientist task. Secondly, CA systems are “best effort” and the exact cost savings they bring in cannot be clearly quantified. This has an implication on the investment by companies wanting to adopt CA. Finally, companies building CA systems find it hard to propose an attractive pricing and return on investment model for their customers due to these challenges. I shall further elaborate the same through a case study from real customer data to illustrate how an attractive pricing model can be used for building a CA system.
Gainsharing with investment recovery
A large telecom company X was looking to bring cognition into the automation of their support system – specifically auto-resolution of field service requests (SRs). The yearly volume of SRs for X is about 8000 in network products and 30000 in supply chain. The average cost to resolve an SR manually is about 5 USD in supply chain and 70 USD in network product SRs. An initial study of the SRs by the cognitive solution provider Z projected the following cost savings:
Table 1: Cost savings calculation
Gainsharing with investment recovery is a method to share this cost savings between X and Z such that:
Let’s define the variables and the algorithm steps for this pricing model.
Case B is the most attractive for X since it has no investment requirement and the gains are produced from year one itself. The yearly cost of maintenance can be factored into the model as a fixed recurring cost for X. The gainsharing ratio will be negotiated and baked into the contract after discounting all costs. In the above example, if 40000 USD, the gainsharing ratio will be = 0.6 when no separate invoice to X for is desired. There is no general rule-of-thumb for this gainsharing ratio and may vary depending on different factors.
Conclusion
The above case study can be used as a pricing guideline for customers embarking on a similar cognitive automation journey. Future work entails incorporating the savings realized from continuous assisted learning into this pricing model and also accounting for, the efforts by cognitive bots in partially resolving SRs.
Sridharan Muthuswamy, Distinguished Senior Member of Technical Staff – Wipro HOLMESTM
Sridharan is a technologist with over 17 years of diverse experience in software research and development (R&D), communication technologies, data science and machine learning. He has created several intellectual property (IP) assets at the intersection of business and technology.
At Wipro, he is primarily responsible for accelerating the cognitive automation transformation journey for customers through use case ideation, building minimum viable products (MVPs) and proof-of-concepts (POCs) using Wipro HOLMES™, Wipro’s artificial intelligence platform. In addition, he supports targeted go-to-market (GTM) initiatives for strategic customers.
© 2021 Wipro Limited |
|
© 2021 Wipro Limited |
Pharmaceutical & Life Sciences