The core idea
Organisation development is a planned approach to improving how an organisation functions, informed by behavioural science and attention to relationships and systems. Armstrong and Taylor describe its historical roots and evolving practice. It overlaps with change management, but places particular emphasis on diagnosis, involvement and learning rather than simply implementing a predetermined solution.
Source and attribution [1]Using it in practice
Agree the problem with the people who experience it and those responsible for results. Gather several kinds of evidence: operational patterns, interviews, observation and employee feedback. Develop a working diagnosis and invite challenge. Choose a bounded intervention, explain the expected mechanism, then review what changed and what did not. Treat unexpected findings as a reason to adapt.
An example, not a reported case
Worked example · illustrative
A team requests a resilience workshop after rising absence. A joint diagnosis finds inconsistent priorities and repeated late changes to shifts. Managers and employees trial an earlier planning cut-off and a clearer escalation process. The team evaluates predictability and workload as well as absence, rather than attributing all strain to individual coping.
What to watch for
Participation can become cosmetic if the decision is already fixed. Be explicit about what employees can influence and what constraints exist. A workshop may help a team communicate, but it cannot compensate for unsafe staffing or incentives that reward the problem.
Distinguish a symptom, a diagnosis and an intervention
A symptom is what has been noticed, such as repeated late changes or falling confidence in managers. A diagnosis is a provisional explanation of why it happens. An intervention is what you intend to change. Keep these separate so that a request for a workshop does not become the diagnosis. CIPD’s organisation-development guidance connects organisational review, diagnosis, intervention choice and implementation. In an original application, a sponsor asking for a collaboration event would first be asked which work is failing, where it fails and what evidence suggests relationships are the cause rather than conflicting targets or missing authority.
Agree the scope of the inquiry with the sponsor
Record the concern, the people affected, the decisions that are open and the constraints that genuinely cannot change. Ask who can authorise action if the findings point beyond the team. Explain how employee input will be used and how sensitive information will be handled. Do not promise confidentiality without considering who will receive the findings and whether small groups may be identifiable. If the sponsor will only accept one explanation, describe the work honestly as implementation support rather than an open diagnosis. Participation loses credibility when an inquiry is used to endorse a decision already made.
Build and challenge several explanations
Combine work records, conversations and observation, each used for a clear purpose. A missed deadline might reflect capability, a dependency, an unrealistic target or several causes together. Write down what evidence would support or weaken each explanation. Include employees whose account differs from the majority and look at occasions when the work succeeds under apparently similar conditions. Avoid collecting a large volume of opinions without a plan for interpreting them. Present uncertainty explicitly: three consistent interviews can identify a useful question, but do not establish how common a problem is across the whole organisation.
Use models for inquiry without turning them into a checklist exercise
The Cultural Web can prompt questions about routines, stories and control systems; a logic model can make the intended route from an action to an outcome visible. Choose a framework because it helps investigate a particular uncertainty. Do not ask participants to populate every box if half the categories add little to the decision. Action research, discussed in the original overview, encourages revisiting an explanation after trying a change. Keep a record of what was learned and what was revised. The value lies in a better account of the work, not in producing a completed diagram.
Worked example: a team blamed for being resistant
Invented case: a new case-management process has low uptake and leaders request a programme to improve attitudes to change. Staff describe having to enter the same information twice because one receiving team still requires the old form. A small review confirms the duplication and finds uncertainty about who can close a case. The intervention therefore combines a single agreed record, clearer closure authority and supported practice. A facilitated discussion remains useful for resolving concerns, but it is not asked to repair the system defect. The original label of resistance is replaced with specific, testable barriers.
Choose a bounded intervention and state its mechanism
Describe the proposed action and why it might help. For example, a shared review of difficult handovers could expose missing information, while a revised escalation rule could shorten decisions that currently stall. Specify the manager time, capability and system access required. Decide what can be trialled safely, which cases are outside the trial and who can stop it. Agree a baseline and a review point suited to how often the work occurs. These are editorial planning prompts, not a universal project method. Some changes need wider formal processes before any trial can appropriately begin.
Treat transformation signals as leads to investigate
AIHR’s member Transformation Signal Scorecard asks practitioners to examine internal and external pressures and consider their business implications. An original extension is to pair each important signal with evidence, an alternative explanation and a decision owner. If leaders report that customers expect faster responses, check the actual requests and which parts of the service are affected before assuming the whole operating model must change. Keep urgency separate from confidence in the evidence. A high subjective rating can focus attention, but it does not quantify the benefit of a proposed intervention or prove that a large programme is necessary.
Evaluate implementation as well as outcomes
Ask first whether the intended action occurred and whether people could use it. Then examine changes in the work and the experience of those affected. In the case-management example, count duplicate entries, unresolved ownership queries and completed cases using consistent definitions. Check workload and error rates as balancing measures. A result before and after a trial is suggestive, not conclusive: staffing, demand or another system change may also matter. Use the linked learning-evaluation guide for an example of cautious attribution. Report mixed results and unintended effects rather than selecting only the indicators that improved.
Leave behind capability and accountable ownership
Agree who maintains the revised process, updates guidance and responds when the original problem returns. Involve those owners during design rather than handing them a finished report. Practise a short review using an ordinary case so that the team can repeat it without the facilitator. Retain the rationale for the change, including assumptions and limits, so a new manager understands why it exists. Stop activities that add no value after the intervention is established. Organisation development should increase the organisation’s ability to notice and address problems, not create permanent dependence on a recurring workshop.
Take it into your next conversation
Three useful questions.
- What competing explanation have we genuinely investigated?
- Can the people involved influence the decision we are asking them about?
- What would show that the intervention happened and improved the work?
Related terms
Go to the evidence
Sources & attribution
Further reading: Michael Armstrong with Stephen Taylor (2023), Armstrong’s Handbook of Human Resource Management Practice, 16th edition, chapters 21, 75. Edition & reading notes ↗
Further reading: Hani Nabeel (2025), Culture Analytics: An evidence-based approach to company culture, First edition, Chapters 3 and 13; PDF pp. 38–39, 190–191 and 214. Edition & reading notes ↗
[1] Armstrong & Taylor (2023), 16th edition, chapters 21, 75 — reading notes ↗
The core idea is an original summary of the cited work. Application notes, examples and sketches are our interpretations, not quotations or reproductions of the authors’ figures. Publisher records may require access to read the full original work.
Original ADM HR teaching guide with attributed background reading. Practical guidance expanded on 29 September 2026. The worked cases, diagnostic questions and suggested decision records are original editorial illustrations, not validated assessment instruments or reproductions of member templates. Source links distinguish further reading from the original authorship of models. The overview review date is retained; this expansion does not establish a new legal review date for linked guidance.
Published 2026-09-25 · Reviewed 2026-09-25. Editorial approach