When AI keeps changing the work, organizations should stop treating each shift as a temporary change project. Rory McDonald and Will Drover call this “steady-state disruption”: a condition in which new capabilities keep arriving and there may be no lasting return to a settled status quo. Their proposed response is to make AI coordination a standing responsibility, protect fast experiments and slower foundational work as separate cadences, and build role-specific learning into everyday work.
That is a management framework, not a proven formula. The examples below are reported by the authors; they illustrate the practices but do not establish that they cause better outcomes.
As an Amazon Associate I earn from qualifying purchases.
Why does AI change call for a different approach?
Traditional change management often assumes a sequence: announce a change, help people adapt, then return to normal operations. McDonald and Drover argue that this sequence may not fit AI. New capabilities keep emerging, and organizations must repeatedly decide what to adopt, what to defer, and how work should change.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →If leaders manage each development as a standalone upheaval, employees can inherit the ongoing work of interpreting tools, adjusting workflows, and filling skill gaps—on top of their existing jobs. The authors’ alternative is to treat adaptation as part of the operating model rather than an occasional campaign.
#1 Best Overall
The premise should be understood as the authors’ framing, not a settled forecast that every organization will experience AI change at the same pace or in the same way. Their practical question is what a change-management toolkit should look like if disruption continues.
1. Give AI coordination a real owner
Someone needs explicit responsibility for watching AI developments, translating their relevance for the organization, sorting ideas into priorities, and supporting decisions about governance. If these duties are simply added to a committee’s existing workload, coordination can become intermittent and unclear.
Rank #2
What the function should do
- Scan: Track developments that may matter to the organization rather than reacting to every announcement.
- Translate: Explain likely implications for teams, processes, and existing systems.
- Triage: Route ideas toward experimentation, further assessment, or no action.
- Support governance: Bring architecture, security, and other relevant considerations into decisions.
The authors describe Microsoft Digital’s AI Center of Excellence as evolving from an advisory group toward a more centralized coordinating role, including an idea-intake pipeline and architecture and security decisions. They attribute this account to the group’s leader, Qingsu Wu, who described the goal as: “How do we turn AI into consistent, measurable outcomes at scale?” This is an example reported by McDonald and Drover, not an independent assessment of the center’s effectiveness.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches2. Separate experimentation from foundational work
AI projects do not all benefit from the same pace. Teams exploring a new capability may need short cycles and frequent tests. Infrastructure, security, integration, and durable product work often require longer planning and implementation. Applying a single speed expectation to both can leave organizations either too slow to learn or too rushed to build reliably.
Use two operating cadences
- Fast lane: Run bounded experiments, test near-term capabilities, and learn quickly. Define what the experiment is meant to answer and how its results will inform a decision.
- Slow lane: Plan infrastructure and foundational changes on a longer horizon, with the time needed for integration and durable implementation.
McDonald and Drover report that Airtable CEO Howie Liu split product work between a faster group focused on frequent AI capabilities and a slower group for infrastructure bets. Liu’s point about the slower work was that “you cannot ship in a week via a ‘hacky prototype.’” The example illustrates the difference in cadence; it does not show that this organizational structure is right for every company.
3. Make learning continuous and relevant to the role
Periodic workshops and annual certifications may not keep pace with changing tools and tasks on their own. The authors recommend learning that happens continuously, in small increments, and in the context of a person’s actual work. The aim is not to make every employee an AI specialist; it is to help people build the skills needed for their responsibilities as those responsibilities evolve.
That approach requires more than offering a general course catalog. Organizations need ways to identify role-specific gaps and connect employees with useful learning while they are doing the work. McDonald and Drover cite Salesforce’s Career Connect as an example of identifying employee skill gaps and surfacing tailored learning opportunities through Slack. They also mention Agentforce Learning Days. These are reported examples, not an independent evaluation of the programs or their impact.
Free tools Windows power users keep installed
One-click scans. No signup required.
How can leaders tell whether their response is built for ongoing change?
McDonald and Drover’s framework points to three diagnostic questions. They are prompts for organizational discussion, not a validated scorecard.
- Ownership: Is AI scanning, translation, and triage an assigned responsibility, or an extra committee task that competes with other duties?
- Cadence: Can teams experiment quickly while foundational work proceeds at a realistic, slower pace?
- Learning: Do employees get ongoing, role-relevant support in their work, or mainly periodic training separate from it?
A weak answer in one area can undermine the others. For example, experiments without coordination may produce scattered efforts, while new tools without role-specific learning can leave employees to work out new expectations alone. These are practical implications of the framework, not measured causal findings.
What the examples do—and do not—establish
The organizations in the article make the recommendations easier to picture: Microsoft Digital for coordination, Airtable for distinct product cadences, and Salesforce for learning embedded in work. The authors also report that Aon has roughly 60,000 employees and attribute to CEO Greg Case the view that AI can widen what employees are able to do. The workforce figure comes from the 2026 MIT Sloan Management Review article; an underlying Aon source was not verified.
These examples should not be mistaken for comparative evidence that one design works best, or proof that the practices reduce overload or improve performance. The article discusses workload concerns and refers to external research, but the available text does not identify the underlying studies, dates, samples, or numerical findings. Its contribution is a management framework and reported illustrations, rather than an efficacy study.
Article date and source
MIT Sloan Management Review lists McDonald and Drover’s article as dated September 16, 2026, and says it is available as a PDF: MIT Sloan Management Review Store. The syndicated Tribune Content Agency text displays September 10, 2026. The University of Virginia Darden Report also lists the article with a September 16, 2026 date: Darden Report.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
