Getty Images/iStockphoto
Get AI change management right: A checklist for success
Is your business truly ready for AI -- or just adding expensive tools without adapting workflows to maximize value and minimize risk?
You can lead a horse to water, but you can't make it drink.
Likewise, you can give a team an AI tool, but that doesn't guarantee it will deliver real business value. Worse, if implemented improperly, the tool could increase risks and drive up costs.
Even if AI saves time and effort for specific employees or tasks, it only reaches its full potential if workflows are redesigned to integrate the technology effectively. Strong change management practices ensure that AI becomes an integral, productive part of the business, rather than an expensive investment with limited value.
Checklist: 5 best practices for AI change management
Determining how to adapt workflows -- or measure the value of change -- can be challenging. The following best practices checklist can help organizations streamline AI change management.
- Run a pilot project before full-fledged adoption. Integrating AI directly into production workflows without prior testing can create unnecessary risk. A better approach is to run a limited pilot project involving a small subset of users or a noncritical process. This lets teams test the tool and assess how organizational processes will need to change.
- Observe how employees actually use the AI tool. How employees actually use an AI tool often differs from how vendors or leaders expect them to. Before redesigning workflows, collect real usage data to understand how tools are used in practice and the effect they have on business processes.
- Involve process owners in workflow redesign. The people who perform or manage a process on a day-to-day basis are best positioned to understand how AI is likely to affect it. To that end, consult the people closest to the work when modifying workflow designs to incorporate AI; the input of these process owners is critical for business leaders.
- Iterate on process changes. Treat AI change management as ongoing. To ensure that processes continue to improve over time, periodically revise workflows based on performance, employee feedback and evolving tool capabilities.
- Quantify AI value creation. To gauge the impact of AI, measure baseline costs such as personnel time, labor and spending before and after AI integration. Track these metrics over time to confirm that AI's benefits outweigh its costs.
The role of change management in AI adoption
To generate value from AI, businesses need to change not just the tools they use but also the processes those tools support. After all, most organizations' workflows were designed before AI became widespread, and they reflect assumptions from that era: how tasks are performed, how long they take, who completes them and how risk is managed.
When AI enters the picture, many of these assumptions no longer hold. AI might dramatically speed up the time it takes one person to complete a task, for example, which could change expectations for timelines and staffing. AI could even entirely automate some tasks, reducing the need for human involvement altogether.
AI can also introduce new risks. For example, sharing AI models with third parties could expose sensitive data. If workflow designs don't adapt to these changed realities, businesses might miss out on AI's full benefits.
Take a common scenario: A new generative AI tool reduces the time an employee needs to complete a task. But if the broader process stays the same -- and the employee's freed-up time isn't repurposed for higher-value work -- overall efficiency doesn't improve. The result is a new tool, likely with new costs and risks, but no net gain in business value.
AI change management example: Software development
Imagine that an organization adopts an AI-assisted code generation tool to speed up software development. This AI augmentation, the thinking goes, will reduce the time human developers spend writing boilerplate code and test cases.
But if the AI coding tool is introduced without modifying internal development processes, it's unlikely to add much value. Developers might use the tool inconsistently, leading to inconsistent workflows. They might also overlook new risks associated with AI tools, such as package hallucination and other security vulnerabilities.
To add value while keeping risk in check, the development team would likely need to make process changes across the IT organizations. Examples of such changes include the following:
- Reducing the time allocated to tasks that AI can easily augment, such as writing boilerplate code and low-complexity documentation.
- Adding a step requiring engineers to review AI-generated code before pushing it further down the pipeline, as AI tools can make mistakes that are easier to fix when caught early.
- Automating test case generation with AI, but requiring that humans review the test cases before deployment.
- Establishing clear guidelines for when AI use is prohibited -- for example, in high-risk cybersecurity contexts.
- Redirecting software engineers' time from routine coding to higher-value tasks like software architecture design or UX analysis.
Chris Tozzi is a freelance writer, research adviser and professor of IT and society who has previously worked as a journalist and Linux systems administrator.
