Artificial Intelligence is now everywhere. From a Science-Fiction concept to an widespread technology, it often comes with clichés and misconceptions (Yup, we’ve all seen how it ended in 2001 Space Odyssey, The Matrix or Terminator…)
More and more organisations of all sizes and industries are trying to understand how they can leverage AI and get aboard the hype train. But from a buzzword to an actual solution delivering strong value, their are a few things to keep in mind before taking the leap.
Artificial Intelligence projects can fail if they are not led for a really, really good reason. This might seem obvious, but having witnessed the significant impact they have on an organisation, the business case must be ambitious though pragmatic, focusing on tangible business outcomes. Also, and as sexy as AI can be, is it the most adapted technology to solve that specific business case?
There are many compelling use cases in all industries and departments, from Sales to Finance, Marketing, Health & Safety, Logistics… and more which create competitive advantage and massive savings. This Gartner paper also encompasses good practice when building your business case.
Until future systems can embrace human-like top-down decision-making, current Artificial Intelligence technologies relies on data. A lot of data. It must combine quantity with quality, needs to be clean, comprehensive and avoid siloed information, biais or partial data (limited history, demographics…)
If it is sometimes challenging to figure out what data you will need for your algorithms to learn and deliver efficiently, working backwards from the problem you are looking to solve can help determine where to start.
In today’s digitized economy, the ability to use data represents a real and essential competitive advantage. To get to a future state of mature analytical competency, there’s real work to be done in integrating the data you have already. This is a strategic goal for the entire company and, when addressed properly, will lead you to develop experience and a data infrastructure that unlocks every next step. HBR
In other words, walk before you run and get a good understanding of your data first: Conduct an audit of what data you have and where it sits, making sure you have easy access to all relevant information.
As mentioned earlier, AI, Machine Learning (ML), Robotic Process Automation (RPA) are subject to many misconceptions. Amongst the most common ones are:
- AI is a physical entity – Yes, some people still think of AI as a Terminator-like robot. Well, HAL 9000 would almost seem like a more accurate description…
- AI can think like a human-being – Replicating the complexity and flexibility of a human brain is way beyond existing AI technologies. “The biggest misconception around AI is that people think we’re close to it.” Prof. Gary Marcus, NYU
- AI can learn on their own – You have to teach them how to learn and improve, tweaking their algorithms constantly and feeding them with data
- AI is alsways objective – AI can have massive biais too
- AI will take your job – It doesn’t replace jobs, it makes them more strategic, according to many serious studies
- AI can use and figure out your messy data – As seen above, you need to figure it out first.
Education is key, not only to raise awareness around what AI will bring to the business at an operational level, but also to get executives onboard, promote data literacy and manage their expectations. Support from the top-management is critical to drive change, get access to all data, overcome political hurdles and lead with vision.
Driving AI projects is everything but a one-off process. The best performing organisations have in-house experts and developers constantly improving models and algorithms, making sure they are updated on a frequent basis and adapting to the latest requirements and information.
We have seen how strategic your data is, especially when it comes to AI. Data governance gives you a real competitive advantage, as it will directly impact the quality of your predictions and algorithms. An efficient data governance strategy must cover:
- Security – Managing access to data
- Integrity – Making sure the data is accurate
- Loss prevention & backup – Mitigating data loss/corruption risks and protecting privacy, especially when it comes to sensitive information
- Lineage – Keeping full transparency for everyone to understand where the information comes from and how it’s been processed, avoiding the “black box” effect
- Completeness – Avoiding biais and partial views
- Ethics – This is sometimes overlooked but we cannot emphasize enough how critical this is, in business as in our everyday life.