How to Plan the MGMT902 Agentic AI Change Management Report
A strong agentic AI change management report does three things: it defines agentic AI precisely, applies a change framework to a specific group of people and a specific use case, and shows how humans stay accountable for what the agents do. Markers reward focus on one or two use cases far more than a broad tour of AI trends.
This guide is written for students working on the MGMT902 Assignment 3 report on agentic AI and change management. It explains how to plan the three parts, split your words and avoid the mistakes that cost marks. You can compare your plan with our reference sample for this report, and if you are working through the whole course, the guides for Assignment 1 and Assignment 2 show how the earlier reports connect to this one.
What the assessment is really asking
The report is a 3,000-word individual piece worth 40% of the course, split across three parts. Although the topic sounds technical, the marks sit mainly in change, governance and people. A report that spends most of its length describing how AI models work will under-deliver.
The company you choose should be a New Zealand business listed on the NZX Main Board and should connect back to the strategy you used in your first assignment. Choose one with plenty of public disclosure so your evidence is easy to find.
Word and mark split
| Part | Marks | Suggested words | Main focus |
|---|---|---|---|
| Part 1: Implementing agentic AI-enabled change | 30 | 850–900 | Definitions, change framework, strategic value |
| Part 2: Barriers and governance | 30 | 850–900 | NZ barriers, assurance, recommendations |
| Part 3: Change management plan | 40 | 1,150–1,200 | Stakeholders, oversight, communication, measures |
| Introduction and conclusion | Within parts | 150–200 combined | Scope, use cases, key message |
Part 3 carries the most marks, so protect its word count. Many students run out of space and rush the measurement section.
A step-by-step plan
- Pick the company and read its latest annual report and investor material for strategy, risks and technology statements.
- Select one or two concrete use cases, such as triaging claims, resolving network faults or handling customer queries.
- Write a clear definition that separates agentic AI from rule-based automation and from prompt-driven generative AI.
- Choose one change framework, such as ADKAR or Kotter’s eight steps, and apply it to a named staff group.
- List the barriers that genuinely apply to your company and rank them by impact.
- Choose a governance reference, such as the NIST AI Risk Management Framework or ISO/IEC 42001, and show how it would be used.
- Draft the stakeholder map, communication plan, training plan and oversight model.
- Define measures for benefits, risks and adoption, then edit to the word limit.
Part 1: Make the change visible
Begin with the distinction that defines the topic. Automation follows fixed rules, generative AI produces content when asked, and an agent pursues a goal by planning and taking actions across systems. State that difference in two or three sentences and then move on.
Choosing and applying a framework
Do not describe the framework in general terms and stop. Show what each stage means for real people. If you use ADKAR, say what awareness looks like for a claims handler who will soon supervise an agent, and what training creates the ability to do it. If you choose Kotter, explain which steps matter most for your company and why.
Finish the part by linking the change to innovation, agility and competitive advantage in your company’s market, not in the abstract.
Part 2: Choose barriers that fit your company
The brief lists several barriers: trust and explainability, data quality, legacy systems, accountability, privacy and regulation, and culture. You do not need equal coverage of each. Pick the most relevant ones and show evidence from your company, such as known legacy platforms or the volume of personal data it handles.
Comparing barrier types
| Barrier | Typical evidence to use | Typical response |
|---|---|---|
| Trust and explainability | Staff reluctance, need for readable decision logs | Shadow mode pilots, transparent reasoning |
| Data quality | Fragmented customer or asset records | Data clean-up before go-live |
| Accountability | No named owner for automated decisions | Executive owner and decision rights model |
| Privacy and regulation | Use of personal information under New Zealand law | Privacy impact assessment and human review |
Each barrier should end with a recommendation that follows from your evidence. A list of problems without solutions earns little.
Part 3: Build a plan that could be run
Treat this part as a short project plan. Name the use cases, the stakeholders and the actions for each group. A simple influence and interest grid is enough, provided each stakeholder receives a specific engagement approach.
The oversight model
This is where many reports are thin. Say clearly which decisions the agent may make alone, which it may recommend for human approval, and which stay entirely with people. Name who is accountable and how actions are logged and reviewed.
Measuring success
Include at least one measure for benefits, one for risk and one for adoption. For example, time to decision, rate of human overrides and proportion of staff trained. If you use sample numbers, label them as illustrative targets rather than facts.
How to use your evidence well
Evidence is the difference between a general discussion and a Masters-level report. Draw your company facts from its annual report, investor presentations and NZX announcements, and use credible industry or government material for the wider claims. Each paragraph should make a point, support it with one piece of evidence and then explain what it means for the change plan.
Keep company evidence and academic or standards material distinct in your own mind. A company disclosure tells the marker what the business is doing; a framework or standard tells the marker how you propose to manage it. Reports that join the two, for example by showing how a stated company risk maps to a governance control, tend to score in the upper bands.
Linking back to your earlier work
The brief asks you to tie the report to the business strategy from your first assignment. Open Part 1 by restating, in a sentence or two, the strategic priority that agentic AI supports, and return to it in the conclusion. This creates a clear thread and shows the marker you have aligned technology with business goals rather than adopting AI for its own sake.
Editing for the word limit
Three thousand words goes quickly. After your first draft, remove repeated definitions, shorten framework descriptions and convert long lists into compact tables. Check that every paragraph answers a requirement in the brief. If it does not, cut it or move it, and keep a margin of about five per cent so that tables and headings do not push you over the allowed range.
Mark-maximising checklist
- The agentic AI definition is precise and distinct from automation and generative AI.
- One company and one or two use cases are used consistently throughout.
- The change framework is applied to real roles, not just described.
- Barriers are specific to New Zealand and to your company.
- A governance framework is named and linked to actions.
- Stakeholders, communication, training, oversight and reinforcement all appear.
- Measures cover benefits, risks and adoption.
- Referencing follows APA 7th edition, and the word count excludes only the reference list.
- The title page, font and spacing match the brief.
Common mistakes
- Writing a general essay about AI without a named company.
- Describing ADKAR or Kotter without applying it to staff.
- Listing every possible barrier with one sentence each.
- Leaving out human oversight, or saying only that humans will remain in the loop.
- Using unsourced claims about productivity gains.
- Overspending words on Part 1 and rushing Part 3.
Where to get extra support
If you want expert guidance on structure, our report writing help and management assignment help pages describe how we support students. You can also read about our approach before deciding.
FAQs
Do I need a real company for this report?
Yes. The brief expects an NZX Main Board company, and your evidence should come from its public disclosures, so a hypothetical firm will not do for the submission.
Can I use both ADKAR and Kotter?
You can, but one framework applied well usually scores better than two described briefly. If you use both, give each a clear role.
How many use cases should I cover?
One or two. This keeps the oversight model and stakeholder plan specific enough to be credible within 3,000 words.
Which governance framework should I choose?
Pick the one you can explain and apply confidently, such as the NIST AI Risk Management Framework or ISO/IEC 42001. Show how its ideas change your plan rather than only naming it.
Should I include figures for expected savings?
Only if you can source them. Otherwise, use measures and clearly labelled illustrative targets, and avoid presenting guesses as facts.
What happens if I submit late?
The brief applies a daily mark deduction for late work, so check the current course documents on Moodle and plan your draft submission early.
When you are ready for a structured plan or full support, send your brief through our order page and our team will respond with the next steps.
