If your organisation could reduce one recurring administrative burden, what would you choose?
Preparing a first draft of an internal communication? Turning lengthy public guidance into a briefing for review? Developing training materials from approved organisational resources?
These are useful starting points for an AI conversation because they connect the technology to work your people already need to do. They also give leaders something concrete to assess: whether a particular application could improve a task without introducing disproportionate cost, effort or risk.
For not-for-profit organisations, the value of AI will depend on those practical decisions. A tool may produce an impressive demonstration, but its contribution needs to hold up in the everyday work of delivering services, supporting employees and managing organisational resources.
The starting question is: where could AI make a worthwhile, measurable difference to our work?
1. Begin with a task your people understand
Ask staff which recurring activities consume time, involve repetitive work or regularly create delays. Choose a specific task that can be described from beginning to end.
“Use AI to improve administration” is too broad to assess. “Help our communications coordinator prepare a first draft of a monthly internal update from approved material” gives everyone a clearer understanding of the proposed use.
Identify who completes the task, what information they need, how often it occurs and who checks the result. Establish what currently makes it difficult.
This matters because the underlying problem may have little to do with drafting or analysing information. A report might take too long because its source data is inconsistent. An approval might stall because responsibilities are unclear. Employees might repeatedly ask the same questions because existing guidance is difficult to find.
Clarifying the problem helps the organisation decide whether AI is suitable or whether a simpler process change would be more effective.
2. Choose an application with manageable consequences
An initial use should be useful enough to assess while allowing the organisation to understand and manage what could go wrong.
Possible starting points include:
- Preparing a draft internal update from approved, non-sensitive information.
- Turning public guidance into a draft briefing, checked against the original source.
- Developing training questions from approved procedures.
- Suggesting alternative wording for a public-facing communication.
- Creating an outline for a presentation using material the organisation has cleared for that purpose.
These examples still require review. A summary can omit an important qualification, and a rewritten communication can change meaning or introduce an unsupported claim.
Uses involving sensitive client information or decisions affecting people require closer assessment. Generating a general communication is different from assessing someone’s eligibility for support or making recommendations about their circumstances.
For leaders, the practical question is whether mistakes would be easy to detect and correct before they cause harm. That should influence both the choice of task and the level of oversight.
3. Understand the information involved before choosing the tool
A proposed use may sound straightforward until the organisation examines what information would need to be entered, uploaded or connected.
A request to summarise a meeting, for example, could involve employee information, confidential discussions or details about clients. A document used to prepare a briefing may contain information that was collected for a different purpose.
Before proceeding, establish:
- What information the task requires.
- Whether personal, sensitive or confidential material is involved.
- Whether the task can be completed using less information.
- Which service and account type would be used.
- Who can access the information and resulting output.
- Whether connections to other systems are necessary.
Involve IT and the people responsible for privacy and governance in this assessment. An employee’s existing subscription or a tool’s availability within another application should not be treated as approval for a new organisational use.
OAIC guidance recommends assessing AI products for their intended purpose and taking particular care with personal information. As a best-practice measure, it advises against entering personal information, especially sensitive information, into publicly available generative AI tools.
Leaders do not need to evaluate every technical setting themselves. They do need assurance that the proposed use has been assessed and that staff understand the conditions attached to it.
4. Give human review a clear owner
“Someone will check it” leaves too much uncertainty.
Decide who will review AI-generated material, what they need to verify and whether they have the knowledge and time to do so. For a briefing based on public guidance, that might include checking the source, confirming important qualifications and removing unsupported conclusions.
For a public communication, review may need to cover factual accuracy, tone, accessibility and whether the organisation can stand behind the statements being made.
The reviewer should be able to reject or substantially revise an output. Staff also need to know when to stop using the tool and complete the task another way.
Responsibility for the final work remains with the organisation. A fast first draft is only useful if it can be turned into a dependable result through a practical review process.
5. Test the full task, including checking and correction
The speed at which AI produces an answer is only one part of its value. Assess the complete activity, including preparation, prompting, review, correction and final approval.
For example, a tool might produce a draft briefing in minutes but require extensive checking because it confuses responsibilities or omits important details. Another application might deliver a smaller time saving while consistently helping staff produce clearer first drafts.
Before a trial, agree on what improvement would count as worthwhile. Depending on the task, useful measures could include:
- Total time from starting the task to approving the final result.
- The amount of correction or rework required.
- Accuracy and completeness against agreed criteria.
- Whether staff find the process practical.
- Subscription, training and ongoing support costs.
Use representative examples during the trial. Testing only the easiest material may give an unrealistic picture of how the tool will perform in ordinary work.
Record where it helps and where it struggles. That evidence will support a more informed decision than enthusiasm following a single successful demonstration.
6. Be clear about how additional capacity would be used
Time saved does not automatically become a budget saving. It may allow staff to prepare information sooner, reduce overtime, respond to requests more promptly or give greater attention to complex work.
Those can be valuable outcomes for a not-for-profit. They should be described in terms the organisation can observe.
If an AI application reduces the effort needed to prepare a recurring internal update, what becomes possible as a result? Can the employee spend more time checking information with service teams? Can updates be issued more consistently? Does the change reduce pressure during a busy reporting period?
Be realistic about the scale of the benefit. A few minutes saved across several activities may improve the working day without creating enough capacity to take on an entirely new responsibility.
Discuss these expectations with staff. Their experience will help establish whether the change is useful and whether any additional work has simply moved elsewhere.
7. Decide whether to continue, adjust or stop
A trial should finish with a decision.
Continue when the application delivers a useful result under conditions the organisation can maintain. Adjust it when the task has potential but needs clearer instructions, different information or a stronger review process. Stop when the benefits do not justify the effort, cost or risk.
If the use continues, assign an owner and document the approved task, tool, information boundaries and review requirements. Make sure those arrangements remain accessible to the people doing the work.
Treat expansion as another decision. Success with one type of document does not automatically demonstrate suitability for sensitive records, client communications or a different team’s responsibilities.
A practical starting point for your leadership team
Choose one recurring task and bring together the person who performs it, its operational owner and the relevant IT or governance staff.
Describe the problem, agree on the desired improvement and assess whether AI is an appropriate option. If it is, run a defined trial using suitable information, a named reviewer and clear measures of success.
This gives your organisation a way to build experience while keeping decisions connected to service delivery and responsible use of resources.
AI earns its place when it helps your people complete worthwhile work to an appropriate standard, with a clear understanding of the effort, cost and responsibilities involved.
Comments