FUTURE TECHNOLOGY
Champions can be teams of analysts who are skilled in quantifying value returns and presenting them as inspiration for others. Or they can be power users who are good at articulating the benefits of AI. Also crucial are team and departmental leaders who can use their influence to initiate upskilling programmes. And IT managers will be at the centre of deployment, data access and governance. With the right people in place and early successes having inspired decision makers, the enterprise can look to the longer term.
Beyond quick wins
As the marginal returns of each successive quick win diminish, the business will scale its AI capabilities with a fundamental shift in company culture that reduces the incremental costs of AI use cases. It is in this phase of the AI journey that governance becomes indispensable as it focuses teams on risk-adjusted value delivery and on the efficiency of scaling, all while remaining compliant with government standards.
UAE law on AI is varied and continues to evolve, so, as enterprises scale their AI efforts, it will be prudent to formalise a strategy that balances these developments with the business’ s own view of responsible AI and the innovation requirements of MLOps. Governance as it relates to proofsof-concept( POC), self-service data and industrialised data products must be clearly understood. Experimentation is to be encouraged but teams need to know when self-service projects should move into the funding, testing and deployment stages.
Modern MLOps teams embed governance into the AI lifecycle, from data collection to operationalisation. They will be aware of roles and responsibilities across design, performance and maintenance of ML models and will monitor updates and refreshes to manage model drift. They will have established performance and risk metrics to protect potential value, and they will be dedicated to the ongoing auditing of models to check for any deterioration in the quality of results.
Over time, the MLOps team can only prevail if due attention is paid to the diverse training needs of not only the team itself but also non-technical colleagues. Onesize-fits-all training is unlikely to guarantee sustainable success in AI, so time and resources must be applied to strategic
INTELLIGENT TECH CHANNELS 41