GuideTransformation
How to run a change impact assessment before rollout
A change impact assessment works out, before rollout, how much a planned change alters the daily work of each stakeholder group and in which ways: process, technology, role, skills, behaviors, measures and location. Done well, it shows where training, communication and support effort should go, which groups are already overloaded by other initiatives, and in what order to roll out. This guide covers six steps and the AI-specific impacts most assessments miss.
On this page
- What a change impact assessment decides, and when to run it
- Six steps from process maps to a sequenced rollout
- Rating guide for the seven impact dimensions
- AI-specific impacts that standard impact templates miss
- A hypothetical claims-triage rollout across three regional teams
- Validating ratings and tracking readiness up to go-live
- Where change impact assessments go wrong
- Questions and answers
- Sources
What a change impact assessment decides, and when to run it
A change impact assessment answers a narrow question: for each group of people touched by a change, what will be different on the first Monday after go-live, and how hard will that be? The answer drives three plans that programs often build on instinct: who needs training and how much, who needs to hear what and when, and where extra support must sit in the first weeks.
Run it once the future-state design is stable enough to describe, usually late in design and before build finishes, and refresh it whenever the design moves. Too early and the ratings are guesses; too late and the plans cannot be resourced. In ColdAI's transformation approach this work sits at the hand-off from Design & Plan to Execute & Adapt1, and it feeds the rollout sequence rather than a report.
It differs from a stakeholder analysis, which maps influence and attitude: the impact assessment is about the work itself.
Six steps from process maps to a sequenced rollout
Define the change precisely
Write the change as concrete differences between current-state and future-state process maps: steps removed or added, decisions moved, systems replaced, approvals changed. Where a service blueprint exists, use its backstage lanes, which show internal work that journey maps hide.
Identify stakeholder groups at the right granularity
Group people by how the change reaches them, not by the org chart. A team that uses the new system all day and one that approves its outputs weekly are different groups. A useful test: if two groups would need different training or a different message, separate them. Include indirect groups such as finance, audit, IT support and customers.
Rate impact by dimension
For each group, rate the degree of change on every dimension in the rating guide below as none, low, medium or high, and write one sentence explaining each medium or high rating. The sentence matters more than the label, because it is what trainers and communicators will act on.
Assess readiness and capacity
Impact alone does not tell you risk. A high-impact change landing on a group with spare capacity is safer than a medium one landing on a team already absorbing two other initiatives. List everything hitting each group in the same window, including upgrades, reorganizations and regulatory deadlines, and mark peak periods such as quarter-end.
Build the heat map and sequence the rollout
Plot groups against dimensions, shaded by rating, with a saturation flag per group. Begin with a group that has meaningful impact but good readiness, so the first wave teaches you something without overloading anyone; move saturated groups later or split their changes into smaller releases.
Turn every rating into a plan
Each medium or high rating should map to at least one action: a training module for a skills impact, a revised procedure for a process impact, a manager-led conversation for a role impact, an updated scorecard for a KPI impact. A rating with no action is either wrong or a gap.
Rating guide for the seven impact dimensions
Use one scale for every group so ratings compare across the program, and write the descriptions down before rating begins.
| Dimension | Low impact | Medium impact | High impact |
|---|---|---|---|
| Process | Same steps, small changes to sequence or forms. | Some steps added or removed; new handoffs. | Workflow redesigned; most steps differ from today. |
| Technology | Familiar system with new fields or screens. | A new tool for part of the work. | Primary system replaced, or an AI agent inserted into the workflow. |
| Role and organization | Responsibilities unchanged. | Some duties added or moved to another team. | Role redefined; reporting line or team changes. |
| Skills | A short awareness briefing is enough. | Structured training before go-live. | A new capability that needs practice and coaching to build. |
| Behaviors | Habits largely unchanged. | New routines that need reinforcing. | Long-held workarounds or informal approvals must stop. |
| KPIs and incentives | Measures unchanged. | New measures added alongside existing ones. | Targets, scorecards or incentive pay change. |
| Location and schedule | No change to where or when people work. | Occasional travel or shift changes. | Relocation, new shift patterns or a move between sites. |
Ratings describe how much changes for the group, not whether the change is welcome.
AI-specific impacts that standard impact templates miss
When part of a workflow moves to an AI model or agent, the visible impact is the tasks people stop doing. The less visible impacts are often larger. People take on review and approval duties for outputs they did not produce, which needs a different skill: spotting answers that are plausible but wrong. Accountability shifts too. When an agent drafts a decision and a person approves it, the approver now owns an outcome built on reasoning they did not see.
Add three questions to every rating for an AI-assisted change. Which tasks move to the system, and does the remaining work still make a coherent job? Which new duties appear, such as reviewing, correcting, escalating and monitoring exceptions? And who is accountable when the system is wrong? For a task-by-task view of which work is exposed, run an AI task exposure analysis first; the impact assessment then takes its findings group by group.
Measures need particular care. If handlers are judged on volume and an AI tool absorbs the easy cases, each remaining case is harder and volume falls, which looks like underperformance unless the scorecard changes with the workflow.
A hypothetical claims-triage rollout across three regional teams
Validating ratings and tracking readiness up to go-live
Ratings made by the program team alone tend to understate impact on the groups it knows least. Validate them before resourcing the plans, then track readiness.
Where change impact assessments go wrong
Rating at department level
Early signalAlmost every cell in the matrix reads medium.
MitigationSplit groups until ratings differ; averages hide the teams that will struggle most.
Treating the matrix as the deliverable
Early signalThe assessment is presented, approved and never referred to again.
MitigationTie each rating to an action owner and review the heat map at every steering meeting until go-live.
Assuming technology is the main impact
Early signalTraining covers screens and clicks but not new decisions or accountabilities.
MitigationRate role, behavior and KPI dimensions with the same care as technology.
Questions and answers
How detailed should a change impact assessment be?
Detailed enough that every medium or high rating leads to a specific action, and no more. For a contained change to one process, a single matrix with a handful of groups is enough. For a program touching many functions, assess each wave separately and roll the results into a program heat map. If ratings change no plan, you have gone too deep; if trainers still ask what changes for a group, not deep enough.
Who should own the change impact assessment?
A named change lead should own it, with process owners supplying current- and future-state detail and each group's representatives validating the ratings. The transformation office or PMO should own the cross-initiative saturation view, because no single program sees everything else landing on a group. The sponsor should see the heat map before approving the rollout sequence.
How does a change impact assessment relate to ADKAR and similar change models?
They answer different questions. Prosci's ADKAR model describes what an individual needs in order to change: awareness, desire, knowledge, ability and reinforcement2. The ACMP Standard for Change Management describes a broader process for managing organizational change3. An impact assessment tells you which groups face how much change, and on which dimensions; individual-change models then help you design how to move each group through it. Most programs use both.
What happens to the impact assessment after go-live?
Its ratings become the baseline for adoption tracking. Groups rated high on behaviors or KPIs are where workarounds and reversion are most likely, so they deserve the closest attention in the first weeks. Cutover and stabilization inside a delivery program are covered by hypercare planning; the impact assessment tells that team where to look first. The benefits each change was meant to produce are then tracked through benefits realization.
Sources
- Transformation capability: approach and offerings — ColdAI
- The Prosci ADKAR Model — Prosci · checked 10 October 2026
- Standard for Change Management — Association of Change Management Professionals (ACMP) · checked 10 October 2026