The Consulting Phase
Setting the stage and understanding the business process
We interact with the stakeholders, understand their roles and responsibilities and map the as-is processes. In this starting phase the key stakeholders from the customer and Spring & River sit down to establish roles to be played by each, which is crystallized in the form of a RACI matrix.
Articulate the problem statement and set out deliverables
This is a collaborative process in which we spend quality time with each of the process owners, as well as other stakeholders, empathise and understand their problems and ideas. Through these discussions we identify the issues and build the problem statement. The process steps are identified and the linkage amongst them is established here. We then articulate the deliverables the customer is looking for and set out the parameters by which the success of the project will be measured.
Define and apprise customer of risks involved
Just like any other project, AI projects will have risks involved. The risks will be identified, mitigation strategies outlined, control and governance steps will be defined. Project feasibility needs to be clearly established. Spring & River will explain and the customer must accept the time, personnel and financial resources needed to develop, maintain and operate the proposed solution and match that with the expected benefits, to assess the practicality of the proposed solution.
Experiment & Execute
The development phase
Check the information flow, data sources, formats and data availability
We now look through the process flow, establish data elements associated with each step , check the formats in which they are captured, understand additional internal and external elements which may be required even if they are not currently captured. All data items with their respective data types be it structured data or textual data or video or images which can go to be a part of the solution are located.
Building the data flow diagram
In this stage we conceptualize how the desired solution is expected to work, set out the data elements required at each stage, map them to the existing data availability, find the gaps, and set down the steps required to capture key missing information and to clean, de-duplicate and complete existing data sets.
Model design - building test data sets
To build the models we need to create multiple data sets for the techniques to be applied. Possible data sets for model building are thought through, their storage requirement firmed up and the methods to prepare them readied.
Model design – candidate AI techniques
At this stage we identify the candidate techniques or a combination of them to arrive at the required solution. The data needs for these techniques are identified and checked with the findings in the previous two steps to ensure that necessary data is available. The best fit techniques are established through a number of iterations.
Present & Deliver
The Implementation Phase
Model design – presentation techniques
Once the model has been finalized, the results presentation methods and technology to be used are thought through. Existing technology in use at the customer premises would be given preference to reduce additional licensing costs.
The infrastructure which will be used for solution delivery is finalized at this stage. Customer preference for public cloud, private cloud or in-house implementation are noted and accordingly Infrastructure decisions are taken
The customer needs to set out how the information generated by the solution will be utilized by the various users, their access rights, the level of detail to be provided, masking requirements for external users (like suppliers), whether downloads are allowed and who are allowed to edit data. The superuser roles and permissions also need to be established. This will define the data access methodology. Apart from this, the information needs to be secured from hacking (encryption, firewalls etc) and breakdowns (redundancies, early warning systems etc.).