The structures, technology and information architecture, cultures and decision processes of most organisations were designed and built for a very different market and technical context than today. While organisations increasingly recognise the opportunities of AI and are beginning to explore how it will shape their industries, there are significant risks that must be managed. Additionally, most organisations face significant barriers as they seek to build internal capability and organisational support for AI-enabled business models. Broadly, the challenges faced in adopting AI across the enterprise are four-fold:
- Data Architecture and technology capabilities
- Trust (internal, client)
- Legacy Organizational models, Strategy and Financial Processes
Data architecture and technology capabilities
While data remediation and modernisation of data storage, access and security are difficult, unglamorous and expensive, many organisations are learning that building AI at scale without a strong data architecture is all but impossible. Additionally, the data scientists required to understand, build and develop AI is difficult to source and often do not ‘fit’ clearly into traditional organisational structures and cultures.
A key challenge for all organisations to drive adoption of AI in a scalable way is to build a data architecture and data governance model that ensures data is of a high quality/integrity with the key data relationships defined is available for AI systems to use for machine learning purposes and is secure from misuse. Increasingly, it is also critical to enable the processing of both internally created and externally sourced data, as well as structured and non-structured data. From the broader organisational perspective, building an awareness of how AI is created, where the exploration for high-value use cases could start, and the tools and prerequisites required for deploying AI is critical.
In specialist functions, there is a high demand for data scientists across industry sectors and employers must be prepared to offer a stimulating and meaningful work environment to keep these highly qualified staff engaged. Building the structures and business sponsorship within an organisation that empowers data science teams and drives value from their insights is critical to success.
AI adoption has only increased an already breathless rate of change. With all of the negative media attention directed at some of the more concerning implications of AI, it is unsurprising that we, whether as consumers or professionals, become more sceptical about the extent to which AI can be a force for good in the world. To humanise AI, the following three principles should be incorporated into plans for the development and deployment of AI:
- Democratise access to AI so that members of your organisation can learn about the technology and use cases. Key principles include educating and training management and staff throughout the organisation on what AI is, and how to develop it; promoting co-location by having data scientists physically close to decision-makers; and empowering employees via access to data and tooling solutions that simplify and automate AI experimentation, such as DataRobot.
- Build a safe environment for experimentation. On the one hand, it is critical to creating data architecture and technical architecture that is secure, with anonymised data (where necessary), that supports experimentation within a test environment or ‘sandbox’. On the other hand, it is important to build a culture within the organisation that welcomes data-driven experimentation and adopts a ‘fail fast’ attitude. For instance, rewarding people for taking risks is a source of real empowerment for those willing to innovate in a ‘legacy’ organisation. In support, internal governance, risk and control frameworks must be “baked in” to the design and go-to-market process for AI, as a strong risk and control framework enables innovators to move fast with the confidence that risks (operational, brand, financial) are being actively managed.
- Develop an inclusive discussion about what AI use cases to prioritise and how to deploy. To manage employees’ concerns, consideration should be given to engaging affected groups to define plans and priorities, and communicating proactively on changing job roles and how employees are being up-skilled to operate in a new AI-driven business environment. Additionally, engaging customers in the design process for new or re-engineered offerings is a powerful tool to ensure you are building features that clients want and trust. Moreover, building a dialogue with regulators and community interests is critical in the process of bringing AI-enabled products to market, as they are acutely aware of how AI will shape the industries they monitor, both for good and bad.
Legacy organizational models, strategy and financial processes
The accelerated adoption of AI and increased pace of change means that rather than having years to adjust to a new competitive environment or technological innovation, increasingly organisations are expected to adapt and respond within months. Outlined below are three areas where we can expect significant disruption to how our organisations operate:
- Traditional hierarchical decision and management structures: State-of-the-art organisations have found that building agile capabilities allows significant empowerment of frontline teams to adapt to local challenges and enabled them to move faster and to foster more innovation that can be shared across the organisation. While many companies have enjoyed success in deploying Agile into their technology development processes, there is a need to deploy agile thinking across the business to ensure frontline teams are empowered to respond to client needs and embrace new technologies.
- Strategy, finance and planning processes: As AI enables more data-driven decisions and reduces the need for human-led analysis and decision-making, strategy and planning cycles will shorten, and budgetary and financial decision-making will need to be far more fluid. With better quality data available and better quality insights being drawn from that data, much of the ‘art’ of decision-making will be replaced by a more data-driven, scientific approach.
- The increasing atomisation of value chains, but a concentration of value: Thanks to data, analytics and network eco-systems, organisations will likely specialize even more in the areas where they have a competitive advantage. In many industries such as banking, we notice the atomisation of value chains driven by open architecture information networks. These network eco-systems will direct ‘work packets’ to the most effective and efficient operators and, therefore, severely disrupt traditional monolithic ‘end-to-end’ service, delivery models.
Across industries and across the globe, governments and regulators are establishing the regulatory frameworks that will guide the use of data and the development of AI. Issues such as the privacy of personal data, the ethics of how AI can and should be used, and antitrust matters relating to new data-driven industries are being formed and developed before our eyes.
For enterprises and other organisations, it is critical to have a strong dialogue with regulators in order for plans to be ratified and regulations impacting businesses to be discussed and influenced. It is only via partnerships between regulators, consumers and businesses that the right balance can be found for many of the issues mentioned above.
As AI will be a key driver of the fourth industrial revolution, it is incumbent on organisations of all sizes to understand the wider implications of AI for clients, communities, staff and other stakeholders. The adoption of AI cannot be seen as primarily a ‘technology challenge’ as the impact, as described in the paragraphs above, is much more profound. By building an inclusive approach to enterprise AI enablement, we can help everyone to better live with AI.
Gerald Mackenzie – CREDIT SUISSE