Student ID: 2267239

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

Student Number: U2267239

Module Code: IB9KPL

Module Title: Digital Transformation

Submission Deadline: 24/04/2024

Date Submitted: 24/04/2024

Word Count: 2246 Of 2500

Number of Pages: 7

Task: Drawing on theories and concepts covered in the

module as well as in your extended reading,
write an up to 2500-word blog that demonstrates expert
knowledge and thought leadership in an area
or application of digital transformation of your choice for an
organization of your choice.
Definition of Digital Transformation: As part of this Digital
Transformation module, we define digital
transformation as building on the triad of ‘data – technology
– people’ to shape the strategic approach
to digital transformation. We encourage you to develop
unique value propositions and innovative
Question Attempted: business models to find solutions and to capture and to
(question number/title, or description of
create value.
As part of the individual assessment, we aim to encourage
assignment) you to strengthen your digital transformation
and digital leadership expertise and online presence. We
recommend that you critically evaluate one
area of digital transformation. With one area we mean that
you do not need to cover all aspects of the
aforementioned triad of ‘data – technology – people’ in
depth. Instead, you could focus on a multi-sided
platform application or an aspect of a technology, e.g.,
Machine Learning or Blockchain, and then decide
on a possible use case, problem that you are going to
solve, or business opportunity for an organisation
of your choice. You will need to analyse the proposed use
case, including potential hurdles that need to
be overcome to achieve the impact that you are aiming for,
by referring to academic literature. Please
note this is not an organizational behavior-type module,
while we recognise the importance of people
management, this assignment should not focus on people
management.
Some further background information:
The blog needs to be presented as ‘ready to post’,
including, for example, images, hashtags and approx.
time for reading. Please ensure that you only use images,
diagrams, graphs and tables for which you do
not infringe copyright and/or have requested permission.
The blog should display your candidate
number rather than your name and should at the time of
submission not yet be ‘live’ on the internet.
The ideas / content that the blog conveys need to be
convincing and should demonstrate critical thinking
grounded in literature (with WBS Harvard referencing). The
blog should be thought-provoking rather
than explaining concepts in a textbook-style-writing. The
blog should be seen as written by a subject
matter expert and driver for change in a specific field of
digital transformation whose opinions matter.
The language used should be formal, yet appropriate for a
blog, i.e. more informal than an academic
journal article. You might want to write the blog using a first-
person positioning (e.g. ‘A recent WBS MBA
module on Digital Transformation made me consider…’ or ‘I
suggest…’), yet most sentence will not
include any writing in the first person (e.g. ‘Miller (2021)
raised the point of….’ Or X means Y (Jones,
2020).
We expect to see an engaging ‘hook’, i.e. an abstract that
makes the reader want to read on, possibly
achieved by asking thought-provoking questions, a clear
structure in your blog which is aligned to
academic essay writing without using the same headings.
For example, you might start in larger letters Page 2 of 3
the ‘hook’, then show an image. You could then introduce
the topic (without calling it introduction) and
provide a review of existing material (without calling it
literature review). You then might structure your
main body in different sections using content-related titles
in bold before having your concluding
thoughts and acknowledging others. You could then show
your list of references / bibliography under a
heading such as ‘Useful articles in the domain of X’ and
adding hashtags to relevant areas.
Here a link to sample blogs of students who published their
blogs last year (note students came from
different modules, please align yourself with the outline
above rather than the sample blogs):
https://blogs.warwick.ac.uk/wjett/entry/scholarly_blogs_part/
Have you used Artificial
Intelligence (AI) in any part of this yes
assignment?
Academic Integrity Declaration
We're part of an academic community at Warwick. Whether studying, teaching, or researching, we’re all taking part in an expert conversation
which must meet standards of academic integrity. When we all meet these standards, we can take pride in our own academic achievements, as
individuals and as an academic community.

Academic integrity means committing to honesty in academic work, giving credit where we've used others' ideas and being proud of our own
achievements.

In submitting my work, I confirm that:

▪ I have read the guidance on academic integrity provided in the Student Handbook and understand the University regulations in relation to
Academic Integrity. I am aware of the potential consequences of Academic Misconduct.
▪ I declare that the work is all my own, except where I have stated otherwise.
▪ No substantial part(s) of the work submitted here has also been submitted by me in other credit bearing assessments courses of study
(other than in certain cases of a resubmission of a piece of work), and I acknowledge that if this has been done this may lead to an
appropriate sanction.
▪ Where a generative Artificial Intelligence such as ChatGPT has been used I confirm I have abided by both the University guidance and
specific requirements as set out in the Student Handbook and the Assessment brief. I have clearly acknowledged the use of any generative
Artificial Intelligence in my submission, my reasoning for using it and which generative AI (or AIs) I have used. Except where indicated the
work is otherwise entirely my own.
▪ I understand that should this piece of work raise concerns requiring investigation in relation to any of points above, it is possible that other
work I have submitted for assessment will be checked, even if marks (provisional or confirmed) have been published.
▪ Where a proof-reader, paid or unpaid was used, I confirm that the proof-reader was made aware of and has complied with the University’s
proofreading policy.

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

The Ethical Dilemma: Balancing AI Innovation and Responsibility ................................................ 5

Removing the Ambiguity: A Principled Framework for Responsible AI .................................................... 8

Taking the Risk and Making It Your Friend: A Principled Approach to AI Risk Management ..................... 9

Linking back: A Holistic Approach to Operationalising Responsible AI .................................................... 9

Conclusion: Shaping the Future of Responsible AI....................................................................... 11

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.

I was recently presented with a scenario wherein AI

could replace human workers in financial tax
prediction. This scenario encapsulates the ethical
dilemma at the heart of this discourse. While the
business case for cost savings through AI
implementation is compelling, we must grapple with
the profound implications of displacing human labour
with machines.

As Whittlestone et al. (2019) cautioned, the ethical

risks of AI extend far beyond the immediate
economic ramifications. Unchecked AI systems can
perpetuate biases, infringe privacy, and exacerbate
Figure 1 - Image generated using Bing Copilot and DALL-E AI
societal inequalities (Barocas & Selbst, 2016; O'Neil,
2016). (Mehrabi et al., 2021) underscores the
pressing need to address bias and fairness in
machine learning models, lest we inadvertently
entrench existing disparities.

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)

Figure 2 - Congruence Model (Nadler & Tushman, 1979).

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

The scope involves introducing policies and standards to clarify

the business, focusing on multiple risk functions. External
knowledge will debunk AI fears, aiming not just to save costs
but also to build an AI centre of excellence. AI predates Chat-
GPT, with Geoffrey Hinton's backpropagation algorithm (Plaut &
Hinton, 1987) leading to Deep Neural Networks (DNNs) used in
Large Language Models (LLM), of which Chat-GPT is a product.
AI encompasses Generative and Predictive forms, plus
automation, like predicting outcomes and executing actions
based on those predictions. Sampath & Khargonekar’s (2018)
Socially Responsible Automation (SRA) Four-Level Model

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.

Linking back: A Holistic Approach to Operationalising Responsible AI

Implementing responsible AI requires a holistic, cross-functional effort that permeates every aspect of an
organisation's operations. A comprehensive approach to operationalising responsible AI, using the scenario
corporate tax scenario:
1. Leadership Commitment: Using the framework described, I can create a better use case for the
stakeholders and get their commitment. This commitment should be codified in a mission statement
and guiding principles that inform all AI initiatives.
2. Governance and Risk Management: what has been produced in the spreadsheet starts the risk
management analysis on how the AI will interact with the aspects of the CM.
 3. Technological Enablers: The framework here gives an enabler and shows that the next stage will align
with corporate standards and governance. Here, the outputs of a strategy runway and standards and
policies on how AI interacts with the business will be produced, allowing others in the industry to
implement AI safely and ethically.
4. Cross-Functional Collaboration: While most of this blog was done through research, the next steps
will be to run the spreadsheet through other AI use cases. I will collaborate with technical teams, risk
management, legal, ethics, and compliance functions to ensure a holistic approach to responsible AI
implementation.
5. Stakeholder Engagement: I will actively engage with diverse stakeholders, including employees,
customers, regulators, and civil society organisations, to gather feedback, address concerns, and
foster trust and transparency.
6. Continuous Learning and Adaptation: As the UK considers publishing more information on its AI
regulations, new learnings and adaptations will be made to mitigate unknown risks.

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.

#ResponsibleAI #DigitalTransformation #AIEthics #TrustAndTransparency #FairnessAndAccountability

#AIFATE #FairnessAccounabilityTransparencyEthics #FinancialService #Banking #WBS #ThoughtLeaders
References
Barocas, S., & Selbst, A. D. (2016). Big data's disparate impact. Calif. L. Rev., 104, 671.
Buehler, K., Dooley, R., Grennan, L., & Singla, A. (2021). Getting to know—and manage—your biggest AI risks.
Mckinsey and Company.
Cheatham, B., Javanmardian, K., & Samandari, H. (2019). Confronting the risks of artificial intelligence.
McKinsey Quarterly, 2(38), 1-9.
Cosgrove, B. (2019). Workday’s Commitments to Ethical AI. https://blog.workday.com/en-us/2019/workdays-
commitments-to-ethical-ai.html
Gallo, V., & Nair, S. (2024). The UK’s framework for AI regulation.
https://www2.deloitte.com/uk/en/blog/emea-centre-for-regulatory-strategy/2024/the-uks-
framework-for-ai-regulation.html
Glassdoor’s. (2024). Corporate Tax Analyst Salaries in United Kingdom.
https://www.glassdoor.co.uk/Salaries/corporate-tax-analyst-salary-SRCH_KO0,21.htm
Google. (2024). Document AI Pricing. https://cloud.google.com/document-ai/pricing
Jackson, A. (2000). Tax Cuts: The Implications for Growth and Productivity. Can. Tax J., 48, 276.
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A survey on bias and fairness in
machine learning. ACM computing surveys (CSUR), 54(6), 1-35.
Mithas, S., & Arora, R. (2015). Lessons from Tata's Corporate Innovation Strategy. IT Professional, 17(2), 2-6.
https://doi.org/10.1109/MITP.2015.26
Mithas, S., Murugesan, S., & Seetharaman, P. (2020). What is Your Artificial Intelligence Strategy? IT
Professional, 22(2), 4-9. https://doi.org/10.1109/MITP.2019.2957620
Nadler, D. A., & Tushman, M. L. (1979). A congruence model for diagnosing organizational behavior.
Organizational psychology: A book of readings, 442-458.
O'Neil, C. (2016). Weapons of math destruction: how big data increases inequality and threatens democracy.
Allen Lane. https://go.exlibris.link/lL4k5fyW
Plaut, D. C., & Hinton, G. E. (1987). Learning sets of filters using back-propagation. Computer Speech &
Language, 2(1), 35-61. https://doi.org/https://doi.org/10.1016/0885-2308(87)90026-X
Sampath, M., & Khargonekar, P. P. (2018). Socially responsible automation: A framework for shaping the
future. Nat. Acad. Eng. Bridge, 48(4), 45-52.
Sauer, R. (2023). Responsible AI Governance at Workday. https://blog.workday.com/en-us/2023/responsible-
ai-governance-workday.html
Siebel, T. M., & Rice, C. (2019). Digital transformation: survive and thrive in an era of mass extinction.
RosettaBooks. https://go.exlibris.link/LvrrsKbW
Snee, R. D. (2010). Lean Six Sigma - getting better all the time. International Journal of Lean Six Sigma, 1(1), 9-
29. https://doi.org/https://doi.org/10.1108/20401461011033130
Stratton, J. (2023). How Workday Is Leading the Enterprise Generative AI Revolution.
https://blog.workday.com/en-us/2023/how-workday-leading-enterprise-generative-ai-revolution.html
Trindel, K. (2022). Workday’s Continued Diligence to Ethical AI and ML Trust. https://blog.workday.com/en-
us/2022/workdays-continued-diligence-ethical-ai-and-ml-trust.html
Whittlestone, J., Nyrup, R., Alexandrova, A., & Cave, S. (2019). The role and limits of principles in AI ethics:
Towards a focus on tensions. Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and
Society,
Wu, C., Song, Y., March, V., & Duthie, E. (2022). Learning from drivers to tackle the amazon last mile routing
research challenge. arXiv preprint arXiv:2205.04001.
Appendix