AI automation application documentation

AI automation application process

1. Scoping phase

Preparation

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):

  • Current business objectives
  • Organisational structure chart, list of stakeholders and resource allocation
  • Current internal resource skills coverage and capacity
  • List of current bottlenecks, resource challenges, or inefficient business processes
  • Relevant process guides, policies, regulations and data dictionaries
  • Relevant assurance or consultancy reports

  • Technology architecture diagrams

  • Existing AI and data initiatives or projects

 

Decisions needed:

  • Define the delivery team
  • Consider change management, stakeholder engagement approach and areas of potential resistance

Kickoff and identify PoC

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:

  • Key Stakeholders – to provide business strategy, impacts of potential changes and decisions on the preferred approach
  • Fifty One Degrees team – to guide the discussion, provide technical feasibility guidance and coordinate discussion

 

Decisions needed:

  • Define the first PoC

Full project scoping

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):

  • Relevant internal policies and regulatory guidelines
  • Data governance frameworks

 

Decisions needed:

  • Define the key areas within the scope as listed in this section

AI automation application process

2. Execution phase

Identify approach to automate process

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:

  • Provide access to datasets and real-life examples for test case development

Build AI automation application

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:  

  • Provide feedback on regular status updates from Fifty One Degrees

Evaluate module against test cases

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:

  • Review the test results and provide feedback on module performance
  • Consider the business impact and compare to the original project objectives

AI automation application process

3. Integration phase

Build exception reporting & notification module

We create an exception reporting and notification system to provide output from the automation to be used by the business.

Client involvement:

  • Collaborate in defining exception criteria and notification preferences

UAT testing & optimisation

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:

  • Participate in UAT and provide feedback for final adjustments

Deployment & launch

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:

  • Given the application is likely to be deployed into your environment, your team will undertake the final deployment

Ready to adopt AI? Check out our guides

Resources home

Want to grow your business or create efficiencies with AI?

Get in touch