Why how you deploy AI reveals what your organisation truly values
Many organisations are somewhere on the Responsible AI spectrum right now. There is a policy, or one is being written. There is a working group, or one is being stood up. And somewhere in legal, someone is reading the UK Government AI Principles or the EU AI Act with the same expression they had when GDPR landed: alert, slightly anxious, and hoping it will resolve into a clear list of things to do.
That framing is the problem.
Responsible AI is not a compliance exercise. It is not a risk register entry or a governance addendum. It is a values test. And unlike most tests, organisations do not get to choose whether they sit it. The question is whether leadership has recognised that yet.
The same mistake, made faster
There is a pattern that plays out in large-scale transformation programmes with remarkable consistency. An organisation invests heavily in new systems, restructured processes, and rebranded operating models. It communicates the vision from the top. It manages the rollout carefully. And then it wonders why adoption is slow, resistance is high, and the intended outcomes are not materialising.
The reason is almost always the same. People were treated as a variable to be managed rather than as the foundation on which the change was built. They were recipients of a decision that had already been made, not participants in shaping it.
AI deployment is following the same pattern, at greater speed and with higher stakes.
When the question of how to implement AI is answered entirely at the top, when the technology choices, the use cases, the governance structures, and the risk trade-offs are all determined before the people closest to the work have had any meaningful input, the result is a system that may be technically sophisticated but is organisationally fragile. It reflects the perspective and values of the people who built it, not the people who will live with it.
Sustainable change only happens when people feel it was designed with them, not done to them. That principle does not stop applying because the change in question involves artificial intelligence.
Who is in the room when the system is built
The bias problem in AI is well documented. Models trained on underrepresented data produce skewed outputs. Systems designed to automate decisions in hiring, credit, or healthcare have, in practice, reflected the inequalities already embedded in the data they were trained on. These are not theoretical risks. They are documented outcomes from real deployments.
The instinct is to treat this as a technical problem, something to be resolved in the model. But the technical choices sit downstream of the human ones. Who decided what the system should optimise for? Who chose the training data? Who defined what a good outcome looks like? Those decisions shape everything downstream, and they are made by people, in rooms, with particular perspectives and particular blind spots.
The best ideas about how an organisation should change often come from the people closest to the work. The same is true of the best insights into where an AI system might fail, cause harm, or produce outcomes that look efficient on paper but land badly in practice. Creating the conditions for those perspectives to surface and be heard is not a nicety. It is a design requirement.
Organisations that have tried to address structural inequality through intention alone, without changing who has a genuine voice in material decisions, have learned that it does not work. The same principle applies to building AI systems that function equitably across a genuinely diverse range of users and contexts.
Representation at the point of design is not a philosophical commitment. It is a practical one. Teams with a wider range of perspectives are more likely to ask the right questions before deployment rather than after.
Technology in the service of people
There is a question worth asking about every significant AI investment, does this genuinely improve how people work, or has it been chosen for its own sake?
The answer matters because technology deployed without a clear understanding of the human need it is meant to serve tends to create new frustrations rather than resolve existing ones. It generates resistance, not because people are averse to change, but because they can see that the system was not designed around them. Resistance of that kind is not a management problem. It is a signal. It is telling the organisation where the design fell short.
There is a broader version of the same question. Beyond the immediate workforce, does the AI system serve the communities it affects? Does it account for the interests of people who never had a voice in its design but will experience its decisions nonetheless?
For organisations that have made public commitments on sustainability or net zero, there is an additional dimension that is not yet being counted consistently. A single large language model training run can consume as much energy as several hundred transatlantic flights. The infrastructure required to run AI at enterprise scale is energy intensive, water intensive, and reliant on global supply chains with their own embedded footprint. If an organisation would ask rigorous questions of any other major supplier about environmental impact, it should be asking the same questions here.
The gap between sustainability ambition and AI procurement practice is not evidence of bad faith. It reflects where AI sits in the cognitive hierarchy of most leadership teams right now. But the gap will be noticed, not first by regulators, but by employees and stakeholders who understand the organisation's commitments and are watching to see whether they hold.
Governance that reflects values, not just risk appetite
The EU AI Act is the most significant piece of AI regulation currently in force. It is risk-tiered, extraterritorial in reach, and will catch more organisations than those operating directly in European markets. We have to believe that in time, equivalent frameworks will exist across the UK, US, and beyond.
The organisations best positioned to navigate this are not those with the most sophisticated compliance functions. They are the ones that have done the harder work first, equipping their people to understand AI, building governance that reflects genuine accountability, and establishing the kind of psychological safety that allows honest conversation about where the technology is falling short before the consequences become visible externally.
Regulation sets a floor. It defines the minimum standard below which legal consequence follows. It does not define what it means to be trustworthy. It does not tell an organisation how to be the kind of business its customers, employees, and communities actually want to engage with.
The GDPR parallel is instructive. Some organisations treated it as a box to tick. Others used it as a forcing function to genuinely rethink how personal data is held and used. A decade on, the difference in data culture between those two groups is visible and consequential. The EU AI Act and others will create the same divergence.
The organisations that navigate this best will be the ones that treat regulatory change not as a constraint to be managed but as a prompt to build something better.
That capacity, to keep changing and stay adaptive, is ultimately what determines whether an organisation remains competitive in environments that will not slow down for anyone. And that capacity comes from people, not systems.
The question worth asking
72 Degrees Consulting works with leadership teams on AI strategy, and one question consistently produces more useful conversation than any readiness assessment, if your employees, customers, and the communities you operate in could see every decision your AI systems make, and every trade-off made in building them, would the organisation be comfortable with what they saw?
The question is not designed to generate anxiety. It is designed to generate clarity.
The organisations that can answer it with any confidence have done the work. They have considered whose interests their systems serve. They have given genuine voice to the people closest to that question. They have been honest about what the data carries and what the technology costs. They have built governance that reflects their values, not just their risk appetite.
Responsible AI is not a retrofit. It is a foundation. If an organisation's values are real, they have to show up in how its AI is built, deployed, and governed, not as a policy commitment, but as a set of actual decisions that can be pointed to and defended.
Every organisation already has values, whether or not they have been written down. The question is whether the AI practice reflects them.
That is the test. It is already underway.
This article was written by Chay Blyth for 72 Degrees Consulting
72 Degrees Consulting is a boutique consultancy partnering with Fortune 500 and FTSE 500 organisations on digital maturity, technology strategy, and enterprise transformation.
www.72degrees.co.uk
No comments yet