Use AI with intention.
Practical starting points of AI for arts and culture organizations.
↪︎ Implement controlled uses of AI where you have the most structure, not where you have the most need.
A useful automation needs a crystal-clear assignment
For an arts organization, the question is rarely whether AI can produce something. It is whether using it makes the organization’s work better, easier to sustain, or more accessible.
A faster draft may help. A hundred drafts may create more review than the team can absorb. The difference depends on the assignment, the source material, and the process around the tool.
Start with a specific recurring task that already has a clearly articulable, even templated, workflow, intake, and goal.
Separate preparation from interpretation
There are many places where a tool can help discretely prepare work: organize approved information, suggest variations, reformat a description, compare a draft against direction, or produce a first outline from supplied material.
There are also decisions that deserve a person who understands the institution: what a program means to stakeholders, whether a claim is honestly supported, how a person should be represented in public, whether a communication is appropriate for the moment.
The boundary will differ by task. Make it explicit. A useful system tells the team what the tool can do, what needs checking, and who is still responsible for the AI’s output.
Build a source before building an automation
An organization’s best information is often scattered among people, documents, and old announcements. If that material is inconsistent, AI can reproduce the inconsistency with confidence.
Begin with a small, clean, well-maintained source document or archive: the mission as it is used, current program facts, approved descriptions, audience knowledge, names and dates, useful examples, and language the organization wants to avoid.
A source library does not need to be elaborate. It needs someone to own it and keep it current. (This structure is valuable even if a person ends up doing all the work and writing.)
Test against a real standard
“It sounds good” is too weak a test. Compare the assisted output with an example everyone on the team would be proud of.
Does it sound/look like us? Does it represent the facts accurately? Are all claims traceable? Does the tone fit the audience? Can someone act on it? How much correction did it require? Would the organization feel comfortable taking responsibility for the work and disclosing AI’s role?
If an assisted draft takes longer to revise than a direct draft takes to write, start over. The objective is improving existing work, not letting the tool dictate the workflow.
Keep the arrangement small enough to maintain
An organization can lose the benefit of automation when maintaining it becomes another job. A workflow that requires several integrations, frequent troubleshooting, and one technically confident person may be poorly suited to a small team. (This is the sticking point with early agentic AI.)
Start with the simplest arrangement that meets a defined need. It may be a shared brief, a reusable instruction, a few approved examples, and a familiar editing tool.
Keep source material and finished files accessible outside any one product. A change of SaaS platform or AI vendor should not require the organization to rediscover or recreate an archive of assets.
Apply extra care where the creative relationship is the working relationship
Arts and culture organizations work with artists, communities, histories, and material that may be sensitive or entrusted to them. A tool’s ability to process information does not determine whether the organization should put that information into it.
Use only approved material for initial trials. Be transparent with everyone involved from the start. (Strongly consider best practices for AI disclosures in contracts.) Review the relevant product settings and contractual terms before using confidential records.
This can stay practical, not legalistic. Identify the information involved, the people affected, and the conversation needed to create clear, shared expectations. The point is to preserve trust first, and make the work easier—or better—second.
Example: a first project worth trying
Choose one recurring communications need with a consistently reliable source material feed and one clear reviewer: for example, a programming team already feeds a succinct set of descriptions to marketing, which marketing copies and pastes into various channels.
Could AI be used to automate and scale distribution of approved program copy?
Build a brief. Produce a small set of examples. Record what needed correction. Agree what a good result looks like and who approves it. Only expand to other workflows when that first task works.
A communications design sprint can do this preparation in a bounded engagement. It establishes intent, working material, and the way a team can use assistance without restarting the argument each month.
How is AI working of not working for your team?
↪︎ Set up a call to discuss AI for arts and culture organizations
The honest opportunity
AI can make some kinds of production more available to organizations with limited budgets. The opportunity is to use that capacity to support distinctive programs and real relationships.
The challenge is keeping AI in the passenger seat and not letting it take the wheel.
That requires intention before volume. It requires someone to design how the work will happen. And it requires the organization to retain responsibility for the choices that give the work its value.
Find a useful place to begin.