RISING GIANTS II: USING AI & MACHINE LEARNING
How collaboration between traditional companies and tech start-ups creates a winning team
On my first RISING GIANTS (I) article I wrote about how AI & Machine Learning can be used to leverage massive hidden data potential of traditional enterprises, specifically SMEs, with the help of tech data Start-Ups. Now I try to explain to you how this operatively works.
In a nutshell:
- Start very small, very fast, at low cost and with top-level support.
- Involve only a few people but choose people who love to work with data.
- Pick the low-hanging fruit: Identify one use-case and solve it within 4 weeks.
- Implement this use-case in your daily work to make working life much easier and more effective.
- Spread the word about your success and develop new fans of your work - and then move on to the next use-case.
1. Start very small, very fast, at low cost and with top-level support
Make sure you have support at the top C-level and convince them to start out with a pilot project as a proof of concept (PoC). The advantage of a PoC is that we can start quickly, with a low, well-defined budget and a clear time-frame to finish (usually within 4 weeks). This way you can prove that the new methodology is superior in terms of results and return to the status quo approach (which usually means no approach). In rare cases, the results of the PoC might turn out to be negative, but a lean assessment of the approach saves time and money at every step and stops you early on from running into the wrong direction. The risk on overall investment in terms of money and time stays very low.
2. Involve only a few people, but choose people who love to work with data
When beginning a data science project with a proof of concept, it is always recommended to involve just a handful of people. All of them should believe in the PoC. In a best-case scenario, they should identify themselves early on as true “fans” and should always have an open mindset towards new technology and new ways of doing things. Basically, true fans of data and all things digital! Do not hide the PoC from others who are not involved, but do not spread word about it unnecessarily before having initial results.
3. Pick the low-hanging fruit: Identify one use-case and solve it within 4 weeks
As part of the proof of concept, the core team should define one use-case that does not need to involve the whole organization and where most of the data, if not all, is already available. This use-case should be solvable in 4 weeks’ time and might represent - for simplicity’s sake - only a sample of the corresponding data, if necessary. Finding a suitable use-case is one of the biggest challenges and not an easy task. However, experienced data science experts usually can accomplish this quickly. Finally, after the PoC is finished, results should be presented to all people involved in operations to increase the number of true fans who understand the beauty and simplicity of making working life much easier by eliminating routine work.
4. Implement this use-case in your daily work and make working life much easier and more effective
The goal for the proof of concept is to demonstrate that the technology used, principally machine learning models, works on the selected (sample) data. Members of the PoC team see that much routine work is eliminated and effectiveness is dramatically increased. Now the PoC needs to be shifted into a dynamic mode, based on dynamic, (nearly) real-time data and not only on static or selected data. Simply put: Go from projectto product. Be careful not to try to implement the next front-end system for your teams, at the same time you want your teams to start using the results of intelligent machine learning. They are often tired of being introduced to another new IT system. Instead, simply make the existing systems smarter by implementing the new results in the existing front-ends your teams are used to work with. You’ll gain their support much faster this way.
5. Spread the word about your success and develop new fans of your work - and then move on to the next use-case
Assuming the PoC is successful, spread the word by pointing to concrete results. For example, “We outperformed the internal sales forecast by 25%, where each percentage point gained is worth $100,000 a month. The money saved is due to efficiency increases along the entire supply chain. We made our principal improvement by reducing stock costs.” Also emphasize the fact that increasing efficiency, usually by replacing routine jobs with machine learning, increased on-the-job happiness dramatically, as employees now have time to work on much more fulfilling tasks. People in operations tend to become fans quickly if their daily work environment improves this way!
If your first project is finished, do not forget to start on the next proof of concept, bringing in more people who need to be turned into happier employees and true data fans.
In my next article (Rising Giants III), I’ll write about specific industries where AI and Machine Learning use-cases create significant returns on investment.