AI Planning Template
Vision for AI
Write your vision statement for AI:
How does it serve in a mission statement?
Goals, Objectives, Strategies and Tactics (GOST)
You need to have quantified and time-bound target to pursue, and ideas for how to proceed tactically.
Write your Goals, Objectives, Strategies and Tactics:
- Goals
- Objectives
- Strategies
- Tactics
Business Case
With the strategy clear, a business case can now be built and presented with the goal of obtaining resource. Draft an outline business case using the following headings:
Recommendation
Rationale
Decisions to be taken
Business Drivers
Scope
Analysis
Gaps and blockers
Financial Implications of go vs no-go
Assumptions
Costs
Benefits
Risk
Strategic Options
Opportunity Costs
Conclusion, Recommendation, and Next Steps.
Compliance and Ethics Plan
What are your intentions for the end users of your AI services? Are your intentions fair, honest and transparent?
What laws and regulations cover your proposed AI activities?
Are you using personal data?
In the case of a complaint against an AI-based decision, how will explain the relevant AI process, data and algorithms?
How are you prepared for legal eventualities such as law enforcement access to data and algorithms, and disclosure?
How are you ensuring that 3rd party IP is protected?
Change Management Plan
What are your plans for:
Responsibility for delivery of AI services?
Organisational structural change?
Recruitment of relevant talent?
Training and skills development?
Program Governance
How will you control:
Ethics -
Are you using IEEE Standards for Artificial Intelligence Affecting Human Well-Being?
Openness and transparency -
Do you have openness transparency by design in the following domains:
• Engineering (design and maintenance)
• User
• Professional (AI plumbers)
• Legal
Traceability
Measurement & evaluation
Management of autonomous systems, openness,
Privacy
Security
Skills development
Data Quality Plan
What are your plans for ensuring that data is of the highest quality?
Where is the data coming from?
How was it collected?
Why can the sources be trusted?
To what extent is the data biased?
Can biases be corrected?
Is the data current?
How much data is missing or corrupt in the dataset?
What methods are in place for cleaning and wrangling data?
System Design
What architectural and development paradigms are you using now – for example, Enterprise Architecture, Object-oriented Architecture; Services Orientated (SOA); Agile? Are they appropriate for the future? How will they evolve?
What are the key solution building blocks needed to support greater use of AI?
How will your data management systems need to evolve?
Are you considering external cloud-based services? List the pros and cons.
What would future AI solutions need from other systems?
What would trigger execution and interdependencies with other processes?
How would AI-based predictions cause actions to be taken?
Infrastructure Development Plan
How will you develop your infrastructure so it delivers AI workloads - i.e. support high levels of processing on large volumes of data.
Hardware considerations, including processer choices
Storage
HPC
Hadoop architecture
Software
System management
ADI
Vendor Selection and Management
In what ways does your data security and compliance regulations influence the choice of vendor or force an on-premises solution?
List the pros and cons of 3rd party cloud-based solutions, short term and long term
How will potential vendors encryption, security and data ownership.
What are the risks in terms of building dependencies on an external vendor? Is there a risk of ‘lock-in’?
How can you get your data and algorithms back if you wanted to change vendors?
Control Framework
Have controls in your AI system been built-in at the design phase?
Are you deploying physical systems? If so, have you featured safety engineering as part of the design process?
What ‘red-line controls’ have you set up – for example, maximum transaction value or ‘behaviour inhibitors’ that overrides the algorithm when there is a risk of errors such as regulatory violation.
What tests are in place to verify that the right models have been used on the right data?
Have these tests been used to tune the machine learning system?
Organisational Readiness
Have the following plans been appropriately socialised, and has feedback been solicited and considered: strategic plan; organisational plan; data and data ecosystem; scale-out plan; training plan?
How does your training plan include provision for:
Building AI solutions
Managing AI solutions
Using AI to augment people
Educating consumers or end users
AI skills for leadership
What are your plans for talent acquisition?
Phase 1 Implementation
What problem are you trying to solve?
Why?
How will you know that the implementation is successful?
What deployment method will you use:
Manual – take data files and run models on them using Python, or R
Virtual Machines with ‘sandboxed’ AI environment
Batch/distributed
API
3rd party Cloud Services
machine learning as a Platform
Other - specify
Scale-out plan
What are the plans for scaling infrastructure to deal with additional workload?
How will you scale-out training appropriately?
What new talent will you need, and how will you acquire it?
How will you educate your customers at scale so they can intelligently consume your new products or services?