Enterprise AI: The new frontier

As Artificial Intelligence and Machine Learning technologies have started to come out of their adolescence, a lot of businesses are trying to adopt these in various parts of their value chain. Companies are realizing that they can get a strong competitive advantage by the insights and automation that is possible using Data Analytics and AI in all functions including marketing and sales, operations, supply chain and even in improving their well-established processes.

While the technology has evolved enough to give strong value enhancement to businesses, the success rate of these solutions at an enterprise level is still not so high. The issue this time is not around the efficacy of the algorithms but more to do with the implementation and integration of these technologies alongside the existing tech ecosystem of the organisations. In this context, a few things that the companies offering Data Science and AI solutions will need to understand and address are -

1.???????It is not just about algorithms and models: While working on Machine Learning models are at the heart of the solution, the algorithms are just one piece in the entire jigsaw. The business problem that needs to be solved and the business context are critical to the success of the solution. AI solution providers have to work very closely with the business user in a consulting mindset to deep dive into the problem and to design the solution.

2.???????You don’t need a team of data scientists – you need a data science team: ?Data science is not a single skill. To make a solution ready for enterprise, a combination of skills such as maths, statistics, technology, business consulting and data engineering are needed. Do not focus on one skillset – build teams that deliver the complete solution.

?3.???????Productionizing the solution: To build a solution and to build a solution at scale are two very different problems. Often due to the nature of AI problems, the solution needs a two-step approach. A Proof-of-Concept (PoC) is often required in the first step to test the feasibility of the solution and to assess the effectiveness of the results. This stage is algorithm-heavy and needs the data scientists to come up with the solution. However, when it comes to taking the solution live into a production environment, a lot of other issues need to be taken care of. Data Engineering pipelines need to be created and decisions need to be taken on data security, platform for deployment, nature of data in terms of volume and velocity, system performance, analytics, QA and DevOps. For a successful enterprise solution, all these parts are critical and need to be addressed well.

4.???????Last mile adoption: A successful AI project is not just about writing a piece of software. It is about implementing it in client’s business environment. The adoption of new ideas is often met with resistance – some reasons could be just because users of technology are required to make changes to their way of working but because they have real reasons. The Data Science team should have a strong feedback mechanism to capture the issues being faced and address them iteratively till the adoption of the system happens well.

?In the coming days, the edge that a provider of AI based solutions would have is not in creating the right mathematical constructs but in providing solutions to the business problem. Large businesses and enterprises would be happy to engage with companies that have the enterprise mindset and can offer ‘at-scale’ solutions using AI.?