If your organization has put its toes in the AI waters, you may have already experienced a reflexive backlash to the idea of machines being involved in what many consider very human processes.
You’re not alone. In a study of judges published in 2023, observers recognized that AI implemented to help negate disparities in judgment between communities actually expanded those disparities (Chen, 2023). There was nothing wrong with the algorithm, but judges instinctively rebelled against AI suggestions, even if those decisions were inconsistent with their own previous judgments.
In response to the judges’ behavior, the researcher behind the study suggested a five-step approach to introducing AI into decision-making processes:
- Use the AI as a support tool like autofill in forms. Rather than suggesting a decision, show the judge previous decisions made in the past.
- Add a light “nudge” or reminder when a judge seems to be deviating from previous decisions. “Are you sure?” before making a final decision.
- Turn that light nudge into general coaching, explaining historically why this deviation has led to errors. Allow the judge to explain their own reasoning so the AI can learn in tandem.
- Bring in other judges’ decisions into the decision-making process, essentially widening the dataset from which judges are receiving AI feedback.
- Once trust is established, a recommendation of the optimal decision can be made.
As we can see from Chen’s process, incremental change in the adoption of AI is designed to create trust with artificial intelligence itself. In the first step of the process, AI really functions as a data aggregator but all decision-making is in human hands. In the last step of the process, a human still makes the final decision but AI is capable of actually recommending action.
Incremental vs. Revolutionary Change in AI
As far back as 2010, corporate change management strategies have suggested incremental change as the backbone for technological advancement. In their study on the adaptation of floppy disk manufacturers over time, researchers found that large firms that adopt frequent incremental technological improvements are more competitive over time (McKendrick et. al, 2010).
In a time when floppy disks are long gone, the lessons of adaptation are still valuable in our era of AI. In fact, the adoption of powerful algorithms to simplify and add value to work is likely as transformative as the change from floppy disk to CD to USB to the cloud.
Incremental change in the AI era can be explained as the process of adopting machine learning in small bites versus adding AI into every process overnight. Think of it less as Clippy suddenly appearing to dictate decisions to every employee, and more like the judges’ process of building trust described above.
In high-risk environments like cyber security and healthcare, providers and users are rightly skeptical of AI that’s not transparent. With custom AI implementation, stakeholders can have a good grasp of what data AI algorithms are using and can dictate how they would like AI to be involved in their daily functions.
For example, in a systemic review of the factors associated with successful adoption of AI in healthcare researchers found that ease of use, amount of effort, social influence and training were all key components of adoption (Khanijahani et al., 2022). These are not necessarily technology questions, but organizational management questions best handled through incremental change.
Compounding value through incremental change
While adopting AI at a corporate level may primarily rely on organizational management questions, custom AI solutions can be the blueprint for generating value over time.
When adopting wholesale AI solutions, there are limited options for targeting functionality and building trust. Ready-built systems are extensive and usually built by programmers, not experts in your industry. This can mean that your employees will be asked to adopt several processes all at once regardless of how effective they really are in their day-to-day.
With custom AI solutions for your organization, experts are able to hyper-focus on problems that are built for AI. Are your employees sifting through large amounts of data to find needles in haystacks? AI that is built to handle your data specifically can produce quick results that don’t threaten the autonomy of your employees but produce an immediate improvement in work for those on your team.
Like the five part process outlined in the beginning of this article, a custom AI solution can also be created with your employees in mind, training them as they adopt the new workflow. This compounds value over time as employees save more and more of their day for human-first tasks without having to retrain entire teams on how to do their jobs.
Incremental change can be built into the custom solution to sharpen your AI capabilities as employees build trust with new systems.
The Bottom Line
As we’ve discussed before, AI that separates your organization from the pack is unique and hard to copy. In addition to these principles, the AI also has to work within your organization and be adopted widely by your team.
As you add more AI into your processes, employees who have already built trust with the AI solutions will more readily adopt refined workflows over time. Beginning with incremental change will compound in value and increase your profitability at a greater clip than your competitors because your employees will be change agents who understand the value of your next solution. They will especially benefit from solutions that they know were custom-built to your organization’s needs instead of defaulting to the latest AI trend.
Are you ready to get started on incorporating AI into your organization’s future? Reach out today.
References
Chen, D. (2023). Incremental AI. Asian Journal of Law and Economics, 14(1), 1-16. https://doi.org/10.1515/ajle-2023-0018
Khanijahani, A., Iezadi, S., Dudley, S., Goettler, M., Kroetsch, P., & Wise, J. (2022). Organizational, professional, and patient characteristics associated with artificial intelligence adoption in healthcare: A systematic review. *Health Policy and Technology, 11*(1), Article 100602. https://doi.org/10.1016/j.hlpt.2022.100602
McKendrick, D. G., & Wade, J. B. (2010). Frequent incremental change, organizational size, and mortality in high-technology competition. Industrial and Corporate Change, 19(3), 613–639. https://doi.org/10.1093/icc/dtp045