Professional Documents
Culture Documents
IB9KPL Digital Transformation Individual Assignment U2267239
IB9KPL Digital Transformation Individual Assignment U2267239
This student has been formally diagnosed with Specific Learning Differences.
Please make appropriate allowance when marking. Guidance is available at:
https://warwick.ac.uk/services/wss/staff/information/disability/
Disability Team
Masters Programmes: Individual Assignment Cover
Sheet
Number of Pages: 7
Academic integrity means committing to honesty in academic work, giving credit where we've used others' ideas and being proud of our own
achievements.
Upon electronic submission of your assessment you will be required to agree to the statements above
Table of Contents
Masters Programmes: Individual Assignment Cover Sheet .......................................................... 1
Taking the Risk and Making It Your Friend: A Principled Approach to AI Risk Management ..................... 9
References ................................................................................................................................ 12
Appendix .................................................................................................................................. 13
The Ethical Dilemma: Balancing AI Innovation and
Responsibility
This blog provides thought leadership on the crucial need for a principled, responsible approach to
AI implementation, drawing upon academic literature and real-world case studies. I will propose a
comprehensive framework to navigate this complex landscape by critically evaluating the
challenges and risks inherent in AI adoption, fostering trust and transparency while unlocking AI's
transformative power.
However, rather than viewing AI as an adversary to human labour, we must reframe our perspective.
By embracing responsible AI practices, we can unlock a future where humans and machines coexist
symbiotically, augmenting and enhancing each other's capabilities. I will be implementing the
framework at my workplace to test the theory. I also hope others can use this framework for their AI
Digital Transformation.
Embracing Responsible AI: A Principled Approach to Digital Transformation
“It is uncertain whether history precisely replicates its events, yet there appears to be a rhythmic similarity. Pattern
management is the most crucial skill of management. This involves the discernment required to navigate complexity and
identify fundamental truths reminiscent of those encountered in disparate scenarios. Leading to how evolution sees
disruption, leading to mass extinctions and the rise of new.” - (Siebel & Rice, 2019)
Taking this scenario through Nadler-Tushman’s Congruence Model (CM), there are the right inputs, a strategy
that needs to be formalised, a scoped task, and the right level of interactions with the culture, people, and
organisation. The issue is enabling the business to implement AI within tight regulations. Further, there is
ambiguity with current laws and rapid changes in how society perceives AI. The push for change in the
scenario is evident, and there is a valid business case to implement such change. This leads to the need for a
strategy, which is missing in the above scenario due to the early phase of AI. The ADROIT Framework (Mithas
et al., 2020) is what is needed to drive a change in mindset in the business to adopt AI responsibly, an
acronym for:
1. Add Revenue: While a Tax improvement does not seek to add revenue, reducing taxes improves
business growth and revenue (Jackson, 2000).
2. Differentiation: The fundamental issue is understanding and knowledge. Changing the narrative of
let’s take a risk to implement AI to discover what is possible within our constraints using a small
project such as Tax. Here is the differentiator to get the willingness to proceed from the right
stakeholders.
3. Reduce costs: The noticeable cost reduction is the change from a salary to a machine cost. The
added cost reduction is the building block for implementing AI in other parts of the business.
4. Optimise Risk: This is the main crux. Can human risk be reduced by automating, as Lean Six Sigma
teaches (Snee, 2010), or does removing the human pose an ethical risk? Though the significant risk is
competitive risk, BCG reported that 78% of companies are struggling with the risk of using AI. Risk
optimisation is the north star of this CM strategy.
5. Innovating: The innovation department is the driving force behind running the hackathon and showing
the value of using AI.
6. Transforming: Implementing tax should not be the vision; this is a Minimum Viable Project/Product.
The fundamentals and learning of a low-priority, low-risk initiative such as the tax enables what it
would mean to implement AI across the business.
The ADROIT Framework is a valuable tool, and when combined with Agile, operations management,
innovation and leadership frameworks, it drives changes more dynamically. Tata used the ADROIT Framework
to transform themselves digitally, notably their InnoVista program, where they award innovation within their
subsidiaries. They proved an upward trend in participation and outcomes by using this framework.
Figure 3 - Trends in Tata InnoVista participation. DTT: dare to try, PI: promising innovations, TLE-PT: the leading edge-proven
technologies, DH: design honour (Mithas & Arora, 2015).
Figure 4 - Five cross-sector principles for responsible AI (Gallo & Nair, 2024).
The innovation department is not the governing body. The hackathon can be classified as an informal
organisation looking at implementing AI. The governing bodies are the Financial Conduct Authority, Prudential
Risk Authority, and the EU AI Act. The UK has formed the Digital Regulation Co-operation Forum (DRCF) and AI
and Digital Hub pilot with a 2024/25 workplan to support innovators and implementors of AI. This brings five
principles to attention, as shown in Figure 4. This culture brings in multiple formal organisations looking at
Risk. Legal, Governance Risk Compliance (GRC), consumer duty, and procurement exist to resolve business
risk. With AI, we include the Innovation team's optimism and the Risk team's reservation. The congruence
model concludes that this would be a candidate for a successful digital transformation. The diagnosed
misalignment is due to knowledge gaps within financial services and external ambiguity of the usage of AI
within the European region.
Removing the Ambiguity: A Principled Framework for Responsible AI
Figure 5 - The Orville, copywrite Hulu/NBC highlights a broader view of automation’s role.
Level 0 – Cost-Focused Automation: The researchers noted that focusing on cost predominantly fails to
derive the benefits a business seeks. The truth is that when looked at through the Congruence, it only touches
the Task and strategy; the people, culture, and organisation are not considered. The average corporate tax
analyst’s salary is around £30,000 (Glassdoor’s, 2024) , and Workday's Google Document AI is priced at $60
for 2000 pages of invoices per month or a $720 annual fee (Google, 2024).
Level 1 – Performance-Driven Automation: Driven by the business metrics, it can include the people, culture,
and organisation. But it falls short of being human-centred. Take Amazon; their warehouse worker or drivers
are at the edge of their efficiency demands. The warehouse staff “gather, pack, and store” goods using robots
to transport the goods within the warehouse. Their delivery uses predictive AI to plan the best routes for
delivering the most packages (Wu et al., 2022). For the scenario above, the benefit is 24/7 and near real-time
processing of these documents.
Level 2 –Human (Worker)-centred Automation: This type of automation exemplifies the human knowledge
needed to monitor and control the AI system. Here, the analyst's job can be moved to an auditing function of
monitoring the AI and reducing the hallucination risk associated with generative AI and the data output quality
of predictive AI. Thus, the human is the observer and the AI the doer.
Level 3 –Socially responsible automation: The movie Hidden Figures shows how an organisation can
implement automation responsibly. The movie depicts how NASA implemented IBM’s mainframes to displace
workers who performed math calculations. However, one of these workers saw it as an opportunity to reskill
and become a mainframe programmer. Similarly, when implementing this tax system, workers can be
educated on the types of AI and how to interact with them. At this level, the culture will be changed to enable
AI.
Taking the Risk and Making It Your Friend: A Principled Approach to AI Risk
Management
Figure 6 - A systematic approach to identifying AI risk examines each risk category in each business context (Buehler et al.,
2021).
AI and its potential risks are pressing topics that require attention beyond ethics. Risks can surface at any
stage of development or live usage as presented in Exhibit 2 of Cheatham et al. (2019) report. Eloquently put,
the three core pillars of AI risk management when using the six-by-six framework are Clarity - the use of a
structured identification approach to pinpoint the most critical risks; Breadth - instituting robust enterprise-
wide controls, and Nuance - reinforcing specific controls depending on the nature of the risk (Cheatham et al.,
2019). Proper policies and standards can help us navigate these challenges effectively. The six-by-six
framework and a structured identification approach can pinpoint the most critical risks and institute robust
enterprise-wide controls. Laws such as GDPR, the FCA handbook, the PRA handbook, PCI DSS, the Data
Protection Act 2018, the Equality Act 2010, and the Human Rights Act 1998 were utilised to form the decision
tree in the spreadsheet attached. The six-by-six framework must be continuously improved, and each risk and
context must be re-evaluated at an appropriate cadence of daily, weekly, monthly, quarterly, or yearly.
Workday uses Google Document AI to perform these Tax Document Optical Character Recognition OCR and
Tax matching systems. I get a positive outcome when using the six-by-six framework on Workdays Document
AI. The key takeaways were that they used Retravel Augmented Generation (RAG) to reduce hallucinations,
ensure calculations are always the same, and keep a knowledge bank for tax knowledge (Stratton, 2023).
Workday is open to discussing how they have implemented the NIST AI Risk Management Framework (Sauer,
2023). Workday uses a “human-in-the-loop” approach for its AI lifecycle (Trindel, 2022) and champions
transparency and fairness with its ethical AI initiatives (Cosgrove, 2019). This, with Data Protection Impact
Assessment (DPIA) questions, builds a broader. In Workdays and Google documentation, there is a concrete
understanding they want to project when customers use their AI and how customer data will be used to train
the models in isolation to that customer tenant. These answer the data, privacy, culture, and transparency
when answering the questions in the impact assessment attached.
By adopting this holistic approach, as Workday and Google have, organisations can navigate the complexities
of responsible AI implementation, fostering stakeholders' trust and confidence while unlocking these
technologies' transformative potential.
Conclusion: Shaping the Future of Responsible AI
I took the congruence model and six-by-six framework and returned to my department to start the discussion;
the best outcome was how we had already used machine learning and how governance and ethics came from
that implementation. The most important output is how the business changed the T&Cs to ask customers if
they want their data to be used in analytics. Fundamentally, many of these risks have been covered and
mitigated; adding AI only becomes a paragraph amendment to existing governance.
The hardest part is changing the culture and narrative of using AI with ambiguous legislation. I did this post
while I got through the challenge of transforming the business to implement AI; I felt the need to help others in
the same situation without having much guidance. I hope to test the hypothesis once I have implemented this
at my workplace. Embracing responsible AI practices is not merely an ethical imperative but a strategic
necessity for long-term success and sustainable growth. By adopting the principled approach above that
prioritises ethical governance, privacy, fairness, explainability, and continuous improvement, your organisation
can position itself at the forefront of the AI revolution while mitigating risks and fostering stakeholder trust.