April 23, 2018
In this post I will do a brief overview of the most useful strategic lessons learned from various of the talks given by Andrew Ng.
He mentions on various of his talks that he believes that AI is the new electricity and that in the next
decade every industry will be transformed by this technology. At Coneptum we believe that in this scenario the first players
to adopt AI oriented business strategies will have a significant advantage over the non AI oriented ones so with this post
we start a series of articles tackling the topic of digital transformation using AI
1. Build an AI team that spreads its AI capabilities across the different business units.
Getting AI to work smoothly on the context of generating economical value for a business is not an easy task. For that
reason, those companies that have different business units working on separate AI projects are likely to get those projects
fail. A better strategy is to build a team of AI experts that work together with all the business units empowering
them with all the data collected from other areas and all the models and algorithms already built in other areas. Another
recommended strategy is to build a Central AI Data-Warehouse that gathers together the data generated in all business
areas in order to be reused by AI engineers when they need it.(On a future post we will explain our technical solution to that matter)
2. Use AI to build positive feedback loops.
Sometimes could be tempting to think, well, we will first will correctly digitalize our business processes and collect all the data needed and
then 3 years late we will start an AI project on this collected data, as Andrew Ng points out on this [great Greylock Partners podcast].
The wrong point here is that in this long initial period of time, your IT engineers will not get any feedback about the
quality of the data generated. A better approach if you collected some data is to let the AI team work on this data, get insights
of it and then provide feedback to the IT team in order to be able to prioritize those tasks that are more likely to help
in the development of valuable production AI systems.
3. Decide thoughtfully which part of your AI to build internally and which parts to buy to a third party.
Another advantage of having a centralized AI team is having an AI leader who thinks about the infrastructure, algorithms and knowledge needed to achieve the AI strategy. All companies have a finite amount of resources and the field of AI is advancing very fast thanks in part to the vast amount of resources invested by big companies like Google, Facebook or Microsoft. For that reason the recommended strategy is to focus on building those things that are singular and unique for your business and are likely to generate value. On the other hank it is highly advisable to buy those products or services that a third party has already built in order to reduce the overall project costs.