metamorworks - stock.adobe.com
4 ways to prepare IT departments for AI-driven change
CIOs must guide IT through AI-driven role shifts by reskilling teams and updating governance. Start planning now to ensure your operations remain resilient.
AI and automation will reshape IT staff responsibilities in application development and network operations, and they are likely to redefine operational responsibilities in systems support, data operations and even data management. What changes to IT workflows and responsibilities are coming, and how will this affect IT staffing and deployment?
Although AI and automation are being aggressively deployed in IT, they are unlikely to replace many IT workers, as some fear.
There are two main reasons for this:
- IT personnel can be deployed to other areas of IT that need attention.
- There is likely to be a correction point within the next year or two, which will reveal AI's limitations and require real human beings to address them.
How AI adoption will reshape IT roles and workforce needs
Let's examine these two statements in more detail.
A shift in IT responsibilities
The areas of IT most likely to be affected by AI and automation are application development and computer operations. App dev will see more AI-powered low-code/no-code applications deployed by end users without much help from the IT development team. In computer operations, certain activities -- such as nightly batch processing -- can be largely automated, if they aren't already. This could mean that personnel in applications and operations lose their jobs, but a more likely scenario is that they get reskilled and redeployed.
The argument for reskilling and redeploying personnel is compelling. IT staff continue to operate on the lean side, and it's becoming increasingly difficult for CIOs to request additional personnel in budget meetings. Consequently, if a computer operations team member can be reskilled to work in networks or IT infrastructure -- and many can -- it could be a win-win situation for everyone.
The same situation applies to application programmers. If they are no longer needed to develop the easier applications that can be created with low-code/no-code generators, they can be reskilled and redeployed in more demanding application development areas or in areas tangential to applications, such as data management, AI system development, QA or even IT infrastructure. The value of reskilling and retaining these personnel lies in their existing knowledge of the company and its systems, which a new hire wouldn't have.
A looming AI correction point
As seen with distributed computing, the virtualization of systems and cloud computing, there will come a point when the novelty and initial excitement of AI wear off. This is when people begin to examine how well the technology is actually performing for the company.
Already, there are signs of this. One such example is when companies acknowledge the existence of AI hallucinations and errors. Another is when users and customers experience disappointment with AI and automation-driven systems that "just don't work" in their interactions with humans or in complex business processes and exception handling.
These are new issues, and both IT applications and QA staff will need to engage with AI systems in new ways to resolve them. CIOs see this, too: In the larger picture, they understand that changes to IT job responsibilities and staff reskilling will be the order of the day. But how can they relieve the staff anxiety that's likely to occur?
How CIOs can prepare their IT staff for AI-driven change
1. Keep staff in the loop
When jobs are redefined and new skills are required, employees get nervous. The best way to ease anxiety is to communicate openly and often. For employees changing positions or responsibilities, give them a roadmap to show them where they fit within the technology changes that are coming.
If there is a case where someone must be released from their role, provide them with every possible opportunity to find another job within the company -- or, if that fails, assist them in finding placement elsewhere. Your staff will notice this gesture.
2. Align the pace of IT staff reskilling with the pace of AI and automation adoption
AI and automation won't transform companies overnight, so there will be ample time for IT staff to reskill. Reskilling begins with an assessment of the skills and tools required for AI and automation.
With the skills and tools identified, determine if any of these skills reside in-house by evaluating IT personnel. If not, determine who will receive training in these areas. By aggressively reskilling IT staff members, staff readiness is built for the AI and automation overhaul before it arrives.
3. Define new governance and security practices
AI and AI-driven automation will necessitate revisions to existing IT governance and security frameworks. The government and industry guidelines for AI ethics are sketchy at best, so companies will need to define their own and train employees in these guidelines.
Security threats to AI models, applications and data also differ from those in traditional networks and systems. For example, an AI security breach could occur through AI prompt manipulation or data poisoning, which rarely occur in traditional systems. To protect and defend against these threats, the IT security staff must develop new security detection and mitigation practices that are adapted to AI.
4. Revisit QA
Today, IT and end users test and tweak application prototypes, refine them, and then use automated test scripts and stress load testing to verify the final quality of an application before launching it into production. This methodology won't work for AI because the quality goals of AI applications differ.
The gold standard for AI system readiness is to achieve a 95% accuracy rate when AI outcomes are compared to those from a group of experts in the same subject area. If the AI outcomes begin to decline in accuracy to below 95%, a refreshed AI model or application might be needed.
There are several reasons why the accuracy of an AI system declines. It could be due to deteriorating data quality, misuse of AI prompts by end users or even rapid changes in external business conditions.
Because the cause of AI degradation is variegated and could be technical or business-related, QA and end users must work collaboratively whenever AI performance degradation is detected. For the IT QA area, this means that QA monitoring and engagement with AI systems must be continuous and that it will be much more expansive and collaborative with end users. The QA staff will need reskilling in soft skills and new AI tools to successfully assume this more expansive and collaborative role.
Mary E. Shacklett is president of Transworld Data, a technology analytics, market research and consulting firm.
