ONE FOR YOUR DATA STRATEGY: WHEN BUYING AI
Acknowledging your company needs a data strategy is just the beginning. For a medium corporation, the main challenge consists on the execution. This will not only require allocation of resources to this aim, but also a change of mentality and ways of working.
It is very likely (and also makes sense) you will outsource most of this work, therefore dealing with consultants comes into play. They might be big firms or rather young small companies, and assessing who is the right one for your case gets complicated.
Here we share 3 considerations that will help you decide and get ready for this kind of projects:
ARRANGE YOUR DATA: NOT A HIDDEN COST
Beware! No fancy Machine Learning (ML) or AI without the house in order. Regardless it is offline or real time calculations, in order to get value from data, storage it in a fit for purpose manner, taking into account: business logic and processing methodology. This two aspects will define the data infrastructure as well as the data pipeline.
Usually, consultants do not mention this stage and customers have unrealistic expectations. You have to know it is not fancy, but the outcome of this stage is more valuable than the actual AI, building the foundations for a sustainable product in the long run.
(PD.: If data collection is required, some steps will be means to an end. Calm, your business needs it.)
PARTNERING, NOT OUTSOURCING
Unlike other projects with external consultants, data driven transformation cannot be completely outsource, must not be. An in-house intrapreneur is needed.
Not only this individual needs to look at the data with business lenses and be familiar with the company’s processes, but also from a technical point of view: Is the data labeled or not?, is it real time?, is your aim classifying or predicting?…
The main functions of this person are:
- Clearly understand the goal of the project, define evaluation metrics and convey them to your external team.
- Get your consultants as fast and accurate as possible familiar with the business logic and processes involved.
- Point of contact for data acquisition: more data and more diverse.
- Bringing the key users and stakeholders of the company along in the processes. This is extremely important for a successful implementation.
DATA STRATEGY EXECUTION: SPEED AND ITERATION
Be ready to deal with uncertainty at speed. The approach of moving faster and iteratively will result in uncomfortable situations with bad outcomes at certain points, but overall, it accelerates the process significantly by helping your data partners to better understand your needs and select and tailor the models and information infrastructure to them.
BUT YOU ALSO MUST EXPECT…
Even though we have talked about patience and walking alongside your data consultants, as a customer, you should expect these results in the first weeks:
- Fast data infrastructure assessment
- Secure Cloud system to easily drop and store files
- Cloud-based automated process (no realtime) for analysis and model validation
To sum up, executing your data strategy is not an easy buy, but if done properly it will deliver benefits even before been finished.