We will start by reviewing key documentation listed here, and we will often shadow the current team to understand practical considerations for the automation application.
We will also define the Delivery Team, based on their personal appetite for involvement and management priorities. Candidates for this team are often ambitious, motivated and technologically capable individuals from the following functions; engineering, data science or analytics, operations and sales or marketing.
Client documentation requested (if available):
Technology architecture diagrams
Existing AI and data initiatives or projects
Decisions needed:
The first meeting will be a kickoff session to introduce the Fifty One Degrees team to key stakeholders within your organisation.
In this meeting, we will define the potential areas where AI automation can bring value, focusing on the highest-impact areas and processes that can benefit from automation. PoC selection will also consider specific criteria such as feasibility, potential ROI uncertainty, alignment with business goals, and implementation complexity.
Attendees:
Decisions needed:
Once a Proof of Concept (PoC) is selected, we define the project’s scope in detail, including project objectives and success criteria, timeline, milestones, budget, data resources, assumptions and constraints, regulatory considerations, roles and responsibilities, project risk management (including controlling scope creep) and functional specification, such as application features.
We will rely on key business information from relevant internal technical stakeholders. Consideration will also be given to how the PoC could be appropriately scaled in time.
Client documentation requested (if available):
Decisions needed:
At this stage, we work together to select the best AI approach (in terms of cost, speed, data leakage and accuracy) to automate the identified process.
We will extract test cases with the expected outputs to guide the automation. Care will be given to extracting a wide range including edge cases to ensure proper consideration is given to a realistic set of circumstances.
Client involvement:
We then build the automation application, often as a scalable micro-service and API, with Python being a likely choice for implementation. This structure will facilitate the automation of the selected process.
Client involvement:
The developed module is evaluated against the test cases to ensure it performs as expected. Detailed performance reports will be shared.
Consider plans for ongoing solution monitoring (implementation of which will not be included in a PoC).
Client involvement:
We create an exception reporting and notification system to provide output from the automation to be used by the business.
Client involvement:
User Acceptance Testing (UAT) is conducted, involving your team to ensure the system works as intended in your real-world environment. This will involve capturing and prioritising feedback, analysing technical solutions and implementation – it is anticipated this period will last no more than 4 weeks and will run in parallel with other aspects of development and deployment.
Security and compliance review will also be required. Risks that cannot be accepted will need to be mitigated.
Client involvement:
The automation solution is deployed into your environment, and a formal launch is initiated. The launch will be designed to be a phased launch so as to compare performance and ensure a safe roll-back position. Post-launch support can be provided to ensure the system operates effectively.
Suitable training to operate the solution can also be created and provided, as required.
Client involvement:
This documentation details a three-phase process for developing AI automation applications: Scoping, Execution, and Integration. The Scoping phase focuses on defining project goals and teams, while the Execution phase involves building and testing the AI solution. The Integration phase covers exception handling, user acceptance testing, and deployment.
The AI automation application process is divided into three main phases: Scoping, Execution, and Integration.
The Scoping phase involves reviewing documentation, defining the delivery team, identifying potential areas for AI automation, selecting a Proof of Concept (PoC), and detailing the project scope including objectives, success criteria, timeline, and budget.
The Execution phase focuses on identifying the best AI approach, building the automation application (often as a scalable micro-service and API), and evaluating the developed module against test cases to ensure it performs as expected.
The Integration phase includes building an exception reporting and notification module, conducting User Acceptance Testing (UAT) for final adjustments, and deploying the automation solution into the client's environment with a phased launch and post-launch support.