The move to artificial intelligence

Machine Learning


The Academy of Health Management defines Five levels of AI maturity in healthcare:

Level 1 – AI Awareness

Level 2 – AI Experimentation

Level 3 – AI in Production

Level 4 – AI is used extensively and systematically

Level 5 – AI is transformative and has become part of the organization’s DNA.

When Academy members asked health care and technology workers where they thought their organizations were on the AI ​​maturity curve, most chose Level 2 – AI Experimentation.

Similar results were found in other industries.

It is also clear that whether AI is introduced into the enterprise through a dedicated data science department or through user functions, the primary responsibility for AI adoption rests with the CIO, as only IT has comprehensive enterprise-wide knowledge of data, applications, infrastructure, and how AI will impact them.

Since most organizations are still in the pilot phase of AI and aren’t quite sure what the ultimate outcome of AI will be, it’s up to CIOs to think about the challenges and opportunities of adopting AI when it becomes a reality in their organizations.

The best way to transition your enterprise to AI

Here are six items that should be included in your plan:

1. Assess your existing data and IT architecture.

AI works best when the data it operates on resides in a data repository, and the data in the repository is of high quality and interoperable with other data types. Achieving and maintaining high-quality, interoperable data takes time. This process starts with the use of ETL (extract, transform, load) tools that take data from various sources, on-premise and off-premise, cleanse and standardize the data, and make it interoperable with various data types in a single data repository for AI to consume.

This process impacts the IT infrastructure because it involves many systems that may not be well integrated with each other. Systems owned and operated by third parties outside the enterprise must be inspected for data compatibility and compliance with security and governance standards.

On the network side, bandwidth will need to be increased, network traffic and deployment patterns may need to be rethought, and on storage, more storage will definitely be needed, and data backup and recovery schemes for AI will need to be established.

Processing must be tuned for parallel stream processing, which differs from the linear processing used in everyday IT transactions.

That's a lot of information to unpack. It will impact how IT operates and require new data management, processing, and strategy skills that IT may not have.

2. Assess your skill level.

Most IT staff will need to upskill in AI.

Staff will need to learn new programming languages, technical IT support in data centers will need to master parallel processing environments, and network staff will need to provision additional bandwidth and faster throughput for AI, likely developing a dedicated network.

Application and business analyst groups need to learn the mechanics of building AI applications, starting with defining algorithms, building machine learning learning models, and proceeding with iterative QA testing until the AI ​​results are within 95% accuracy of the expert conclusions. On the user side, experts need to be recruited to help develop the algorithms.

3. Set compliance and governance guidelines.

Companies, regulators and governments are only just beginning to respond AI Compliance and GovernanceSo it's up to companies to define their own guidelines.

As AI matures, incidents and use cases will arise that will dictate regulations, and those regulations will be developed. In the meantime, the goal for IT is to avoid being caught up in those incidents and use cases.

4. Evaluate user acceptance.

Employee resistance is a big cause of project failure, and employees will resist if they believe AI will take their jobs. The solution is to create a human roadmap so individuals know in advance where their jobs and responsibilities will evolve. If there is a possibility that jobs will be lost, it is best to let employees know about it in advance and help them find other jobs.

5. Assess the risks.

In the state government, Top priority Cybersecurity and risk management are top of mind for CIOs, but they're not alone.

AI poses a huge security risk because IT security solutions are designed for standard transactional IT, not big data.

A growing one AI Security Threats “The data collected for training deep learning has been intentionally compromised with malicious information, resulting in contaminated data.” The data is compromised and the results the AI ​​derives from it are intentionally false and misleading.

The second risk is that AI results will degrade over time. This occurs when business or other conditions change and the algorithms that query the data, or the data itself, cannot keep up with the rate of change. IT and end users need to develop a maintenance strategy for AI systems to continually monitor accuracy and alert them to declining accuracy levels so that IT and end users can make the necessary adjustments to restore accuracy.

6. Stop the experiment.

While most AI systems are still in the pilot stage, the time has come to pursue the right methodologies, technology adoption, and staff upskilling before AI moves into production.

AI is sure to make its way onto the production floor (and ultimately into the DNA of your company), so it's not too early to start rethinking and redeploying your IT operations, methods, and skill sets in preparation.





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *