Artificial Intelligence projects that are coming back online after stalling during the pandemic need careful consideration or they risk falling apart, according to a new report.
The Gartner report states that a strong AI engineering strategy is required by companies to ensure there are no failures within AI projects. The proof-of-concept stage must lead on to full roll-out of the project and to make this happen, strategies must not fall by the wayside.
Obviously, before you embark on an AI strategy, you should really undertake a ‘proof of value’ assessment on the viability of the project and whether it will meet business needs. Of course, PoVs are not where you define the problem, just where you, as the name suggests, prove the value.
AI projects usually fall apart due to lack of maintenance, the inability to scale and failure to govern correctly. A strong strategy will overcome these problems, justifying the investment and helping the project become reliable and easier to interpret.
Briefly, the three areas that must be dealt with are:
- Data – the collection, cleaning and arrangement of data is imperative. If this stage is not taken seriously enough, you will have long-term impacts further down the line.
- Machine Learning – the training, testing and fine tuning of the algorithm the team has designed and built.
- Artificial Intelligence – the AI comes in to play in order to use the data. This must be deployed and closely monitored.
The storage and accessibility of your data should also be understood. Are you working from physical servers or are you using the cloud? What is your security like on the cloud side? Also, have you factored in QA throughout the whole project?
Apart from cloud security, you should always be watchful over all of your data, no matter where it is. Some businesses will be outsourcing tasks but regardless, the data should not be sensitive data and should be treated with the utmost respect.
AI engineering draws together the disciplines of DevOps, ModelOps and DataOps. DevOps is about the speed of changes in code, data or models and is important in AI projects where variables are the norm and this discipline is used for data in DataOps and ML in ModelOps.
A lot of companies, emerging from the Covid-19 pandemic, are engaging in ‘responsible AI’, which takes in trust, accountability, compliance, risk, transparency and fairness, amongst other things. This marks a change from viewing AI as something ‘other’, to understanding that it is now an inseparable part of the business, and thus, must be held to the same standards. As Gartner says, every company is now a technology company.
With companies in the situations they are in, regarding the pandemic, it is understandable that they are chomping at the bit to get going, but speeding ahead without the correct planning would be a mistake. Plans that seemed solid prior the Covid-19 should be reassessed, in light of changing priorities within the business and what has been learned about the business in these testing times.
We have not even mentioned the team yet either, as the success of AI projects stands or falls on the quality of the team involved. This will require a dedicated recruiter, to track down the key players in order to create an AI project fit for the new era of your company.
Curious about how Zenshin Talent can help your organisation? Contact us today for a no-strings conversation about your needs and our experience.