AI is seldom as simple as business leaders would like it to be, or as simple as some AI service companies claim it to be. The technology is still far from perfect, and without significant preparation and positioning, there are substantial risks for false positives and false negatives from AI models – an issue which not only impacts the usefulness of specific models but which can significantly sour a company’s leadership on the overall potential of the technology.
There has been an explosion of interest in AI in recent years, coinciding with an explosion in solution providers in the Artificial Intelligence space. Not surprisingly, we find a wide spectrum in these providers of realism and reasonableness of claims about. AI readiness. At one extreme are those providers who view AI as a panacea for all ills and make extravagant promises about transformative change that their technology can bring about. In the middle are more sober-minded providers, who understand that while AI has great potential, it is not a silver bullet and requires significant prerequisites – organizational, process and data – to be effective.
At the other extreme are skeptics who believe that AI is nothing more than hype and overblown marketing claims. We would argue that there is certainly some truth to this perspective but also note that many businesses have failed to adopt new technologies in the past simply because they don’t understand them or don’t see how they could be relevant to their business. The key question then becomes how to determine whether your organization is ready to adopt this technology. The following considerations should be helpful:
Is Your Data Integrated?
What we have noticed is that a considerable number of companies that approach us to find AI solutions need more preparatory data work before they can safely assume that their business is indeed ready to make efficient use of AI. Since we have always had our customer’s best interest at heart, and we would never want to promise them anything without making sure that they can make the most out of our service, we have compiled the checklist below to help potential clients know whether they’re ready to cut a check for a data science/AI project.
The first step, as with any business project or new technology adoption, is to determine whether there are significant benefits that you will be able to realize by using this service. The next question then becomes what kind of benefit are we talking about here – cost reduction, improved efficiency/speed/performance, increased revenue opportunities etc.? There may also be certain use cases where your organization needs automated solutions – marketing automation and CRM come immediately to mind – but even in these scenarios it makes sense to consider alternatives before assuming that an AI solution is necessary. Our experience has shown us that often a data engineer with some coding skills can build models which work just fine without requiring expensive services from expensive Data Scientists.
For any form of AI to synthesize effective solutions, it needs data, and there are several aspects of that data that should be pondered over because the better the characteristics of the data, the better it can be used by the technology. Someone needs to ascertain whether the data from the operating systems are in integrated form or not. Deciding how much of it is to be used also poses a question sometimes e.g. To what extent data from external sources is to be used? Furthermore, the metrics about the quality of data should also be carefully checked, and if there are any quality issues with the data, there need to be processed to address them. How the data is visualized (e.g. using a business intelligence dashboard) also has to be decided.
Is Your Data Cleansed?
Another prerequisite for data readiness is that it must be cleansed before being used by AI. Most of the time, when data is dumped into a big data lake or warehouse, there are many inconsistencies and inaccuracies due to its original collection process. As humans we are biased in our decision-making so if this bias is not eliminated from the data set then any models built on top of it will also inherit these biases. The first step towards addressing this issue is to identify which fields in the data contain errors and then work to resolve them.
One way to do this is through automated methods such as cleansing algorithms, but manual cleansing (i.e. cleaning by human intervention) is often still necessary to correct values that fall outside of normal ranges. Many firms have arisen in recent years specializing in just this type of cleansing.
Is Your Data Enriched?
Once the data is cleansed, it must be enriched so that the AI system has enough context to work with. This means adding more information to each record in order to make it as rich as possible. There are many sources of enrichment data including free open-source data sets, purchased third-party data, and customer or partner data.
The key is to make sure that the enrichment data is relevant to the task at hand and properly aligned with the business objectives. For example, if you’re looking for new customers, then using demographic or socioeconomic information about current customers would not be very helpful.
Is Your AI Use Case Well Understood?
The final step is to make sure that the AI use case is well understood. This doesn’t just mean understanding the technical aspects of the problem, but also understanding the business context in which it exists.
Many times we’ve seen clients who have a good idea of what they want their AI system to do, but don’t really understand all the implications and how it will fit into their overall business strategy. It’s important to involve stakeholders from different parts of the organization in order to get a complete picture of what’s required.
Once you have answered these questions, you should have a good idea about whether your organization is ready for AI. Of course, we’re available to help if you have questions about your current readiness, or about how to accelerate your progress on this journey.
