AI problem solving is a horizontal skillset, applicable across a landscape of vertical industry challenges in Telco, Retail, Banking & Financial Services, Oil & Gas, Energy & Utilities, Supply Chain, HR and Healthcare.
Data Scientists don’t tend to prioritise or “proactively” understand these industries and their future challenges in their own right, instead rather absorbing what project leads feed into them as the specific vertical industry problem they should be solving. But this can lead to siloed thought processes, supported by problem definitions that are ill-thought out and can go unchallenged to the point that after weeks of data wrangling, feature engineering and code refactoring, a suboptimal solution gets delivered, or worse, a conclusion is reached 3 or 6 months later that requires a complete rethink of the original hypothesis.
A good example of an industry-targeted solution is one where there is an initial investment in time on business context, capturing key domain features and KPIs both on a descriptive (BaU) level and a predictive (forward-looking) level. Solutions borne out of this approach such as this Telco dashboard and 360° HR engine from Agile Analytics tend to more credible or “high-relevancy” solutions for the sector in focus and are more easily scaled to other clients.
In particular a sophisticated mapping of the business challenges can lead to “natural fit” AI solutions such as the use of natural language generation (NLG) transformers to address volumetric documentation productivity gaps in the legal, public sector and consultancy sectors. A deeper mapping requires a framework of course, but often over-complication can lead to slower delivery or lose sight of a priorities. We use an inverted five-tier KPI implementation at ce.tech to help strike the right balance in delivering what is important, in the right order and aligned with best-of-breed predictive solutions and AI technologies - NoLo if possible.
More on that in our next blog.