Responsible Artificial Intelligence, RAI is a governance framework that consists of developing artificial intelligence systems that integrate human empathy and care, and therefore ensure that they work in the service of good while achieving transformative business impact.
As the technology continues to evolve while making notable impacts in many industries, organisations are proactively measuring their RAI capabilities to deliver on their commitments ethically and within reason, as well as drive positive and mature business outcomes, and ensure they are mature.
The practice of implementing RAI means to ensure that the outcomes of AI systems are unbiased, fair, and explainable; that AI systems are robust and safe; that they preserve user privacy by following best practices in data governance; that they minimise environmental impact; and finally on the fact that they do not replace human capacity but rather augment it.
Creating an RAI programme
In simple terms, an RAI programme embodies the structures, processes, and tools that help organisations ensure that their AI systems transform businesses in their entirety while serving as a force for good. To do this, a company needs to start from its purpose and values and translate them into concrete RAI principles.
Companies, public organisations, and associations, have reached this point and published high-level principles of RAI, however, and unfortunately, most of them have stopped there.
To operationalise and make an RAI programme operational, associations and organisations today must take two more additional steps, which are to design specific policies corresponding to each of these principles and operationalise these policies by developing the processes, delivery models, employee training, and the necessary tools.
Benefits for shareholders
RAI directly affects every executive and shareholder because the greater RAI maturity their company has, the more opportunities the company will enjoy, due to the robustness of each RAI dimension. For executives, RAI should mean a small number of new constraints that bring a large number of benefits in terms of brand differentiation, stronger customer relationships, improved employee recruitment and retention especially for younger people and, it goes without saying, lower reputational risk.
And for shareholders, RAI materialises the alignment between their values and interest in a better world and the practices of the company of which they possess.
Digital transformation and RAI
The relationship between successful digital transformation and a positive RAI score is especially relevant. A digital transformation cannot anymore be considered complete and fully successful without taking into account and acting effectively upon the RAI prerogative. And a high RAI score measures how successful a company is in doing so.
Companies that have been proactive in their efforts to ensure ethical, transparent, and accountable usage of AI technologies have experienced success as part of their transformation journeys and improved their RAI scores in the process.
AI has emerged as a vital technology for all companies. Those that appreciate its vitality have worked to develop a comprehensive understanding of their RAI maturity, avoid restricted investments in this field, and ensure they realise the transformative potential of AI to boost RAI scores and harness the power of digital.
Real-life manifestations of RAI
A recent survey that was conducted shed light on real-life manifestations of high and low RAI in organisations. In fact, news in the press everyday report about companies that are sued because their algorithms are proved to be racist, or sexist, or, more generally, biased.
There are also news stories on public administrations sued because discriminatory rules were implemented in their AI systems and whistleblowers who question the AI practices of their companies, which happens quite often.
The financial services and healthcare industries, for example, have reported higher RAI scores and maturity. Both these areas are heavily regulated and have a history of strong compliance and risk management, boosting overall performance. At the same time, industrial goods and automotive were also more mature than anticipated, with business-focused AI applications a driving force because they are simpler to implement.
In terms of low RAI, organisations in the consumer industry have lower scores and maturity, primarily because of complex customer-centric use cases that transpire.
Perceived and real RAI
Today, a large number of organisations do perceive that their RAI maturity is higher than it is. This occurs since companies underestimate the effort necessary for implementing an effective RAI, a conclusion that was substantiated by the survey.
Many of them consider that the issue has been addressed when they publish a series of high-level principles and, as we discussed earlier, this is not enough.
However, even companies that went beyond the principles have failed in some other of the key points. Results from our survey show that their list of policies was not comprehensive, the processes related to these policies were not fully implemented, tools have not been built and, importantly, the company employees and leaders have not been made aware of, and trained on, RAI.
This also applies to the region, where similar findings were recorded amongst other parts of the world. Ethical, transparent, and accountable use of AI technologies, and companies behind in their AI maturity development should expand their efforts to boost performance in these areas and realise the business benefits they envisaged when initially engaging in RAI.
There is some correlation in the sense that larger and more successful companies have the human and material means allowing them to easily develop a successful RAI programme. But, of course, the human and material means are a necessary condition, not a sufficient one. Leadership willingness and sensitivity to RAI issues are key. And high IT spend, or business dominance, do not guarantee these leadership principles.
Geography and IT spending
Research does show that RAI is not an IT issue, even if, of course, IT will have the responsibility of implementing some of the tools developed. RAI is a leadership and managerial issue and managerial practices are often albeit not always more correlated to specific geographies a country, or a region than a sector, spread across multiple geographies with different practices.
It is important to note that each industry – whether it be automotive, consumer, energy, finance, healthcare, industrial goods, or technology, media, and communications – varies in terms of RAI maturity. Moreover, industries that lead in some regions lag in others, which helps explain the lack of differences between them from a statistical standpoint.
- An RAI programme embodies structures that ensure AI systems transform businesses while serving as a force for good.
- To make an RAI programme operational, organisations must take two more additional steps.
- Design specific policies and operationalise these policies by developing the processes.
- The greater RAI maturity a company has, the more opportunities the company will enjoy.
- Companies that have been proactive to ensure ethical usage of AI have experienced success as part of their transformation journeys.
- A large number of organisations perceive their RAI maturity is higher than it is.
- Companies underestimate the effort necessary for implementing an effective RAI.
- The financial services and healthcare industries have reported higher RAI scores and maturity.
- Industrial goods and automotive were more mature than anticipated.
- Organisations in consumer industry have lower scores and maturity, primarily because of complex customer-centric use cases.
- Policies was not comprehensive, processes related to policies were not implemented, tools have not been built.
- Often employees and leaders have not been made aware and trained on RAI.
- Larger and successful companies have material means allowing them to develop a successful RAI programme.
- Leadership willingness and sensitivity to RAI issues are key.
- High IT spend, business dominance, do not guarantee leadership principles.
- Research does show that RAI is not an IT issue, even if IT will have the responsibility of implementing tools developed.
- RAI is a leadership and managerial issue and managerial practices are not always correlated to specific geographies.
- Each industry varies in terms of RAI maturity.
Responsible AI ensures outcomes are unbiased, fair, explainable, robust, safe, preserve privacy, and follow best practices in data governance.