Unfortunately, there is no magical number of lines or colunmns. From our experience, we realized that we always need to clean data beforehand. This leads to a reduction of the data ranging from 20% to 70%, depending on the business case, the way data were entered (manually vs automatically), the sensors quality, the frequency of datacapture, the amount of outliers and many other parameters. The bottom line: the more the merrier.

We have three solutions to transfer data to ensure maximum security. Either through a secured SFTP server, with a virtual machine that we can create at your premise or in special case and when the data is too sensitive, we come with our harddrive directly at your premises. Data is not supposed to be uploaded in other server and be migrated to other premises without your permission. Finally, we are used to signing agreements to ensure confidentiality and we always commit to delete data after end-of-projects.

1 - Use-case generation that could arise from a business need, identified by a tech-savvy business person or business-savy tech person
2 - Define sucess KPIs: what is the size of the potential returns
3 - What are the data we need? Do we already have it or do we need to generate these data?
4 - We develop models
5 - Integration and scaling across the organization

It ranges from well-documented APIs you can call in the future running our models to full integration in your IT systems.

Only according to us: it means the data cannot fit into your computer, or that Excel is starting to lag.

AI needs a language, processes, ressources, human reasoning (yes yes still), and data to run.
Languages are often open-source languages (e. g. Python or R), processes are machine, deep and reinforcement learning algorithms, optimization algorithms and ressources consists of storage and cloud computing. What AI does, is comparing and understand structure that are already created with an incoming datastream.

A computer that understands everything. If it does, it means it probably pre-recorded.
The AI that says "I love you" and means it does not exist.

AI needs a language, processes, ressources and data to run.
As said before, AI consists in particular of processes. A processes that can be executed million of times. AI executes a plan instantanly, while introducing flexibility in the plan executed, which makes it powerful. A computer can do this all 24/7 without being tired.