20 Practical Ideas for Arts and Culture Professionals to Stay Cognitively Sovereign
AI now generates images, writes artist statements, assists with grant applications, and powers audience analytics for galleries and venues. The risk is that tools built on existing cultural production push creative work toward what has already been valued, quietly narrowing what gets made and shown. You need to stay the person who decides what matters.
These are suggestions. Take what fits, leave the rest.
Write your artist statement before using any AI writing toolbeginner
Your statement needs to come from your own thinking about your work first. Use AI to refine grammar and clarity only, never to generate the core idea.
Make your curatorial selection before checking audience databeginner
Select which works belong in a show based on your critical judgement first. Then check what the data says. The gap between your view and the data is worth examining, not erasing.
Reject the first AI image generation output in every visual projectbeginner
The first output reflects the statistical centre of your prompt across the training data. It is not your vision. Use it as a reference point to move away from.
Write your grant application narrative by hand before using AI to refine itbeginner
Funders can recognise AI-shaped prose. More importantly, writing it yourself forces you to articulate what you actually believe about your work and its cultural value.
Identify what your AI audience analytics tool cannot measureintermediate
Dwell time and ticket sales measure engagement. They do not measure whether someone left changed. Decide how you will track the thing the dashboard ignores.
Give a colleague the role of questioning every AI-assisted programme decisionintermediate
Assign someone to ask what perspective the algorithm is missing and what voices the data is underweighting. Make this a formal part of your programming process.
Compare AI-generated programme descriptions to ones written by your teambeginner
Put both in front of a colleague who has not seen either. Ask which one sounds like your organisation. Use this to calibrate how much the AI is flattening your voice.
Research an artist for thirty minutes before asking AI to summarise their workintermediate
Form your own critical view first. Then read the AI summary. Notice what it missed and what it smoothed over. These gaps are where your curatorial perspective lives.
Audit your last three acquisitions or commissions for AI influenceadvanced
For each decision, ask whether audience data or an AI recommendation shaped the choice before your team had time to form an independent view.
Write what you find genuinely difficult about a work before reading any AI interpretationbeginner
Difficulty is where meaning lives in art. If AI gives you an interpretation before you sit with the difficulty, you skip the experience the artist intended.
Organisational Strategy and Community Connection
Talk to ten audience members before acting on any AI segmentation reportbeginner
AI segments audiences by behaviour. People tell you why they come, what they were hoping for, and what keeps them away. That information does not appear in any dataset.
Write your institution's cultural purpose statement without AI assistancebeginner
Your mission is not a content generation task. Write it in a room with your team, in your own words, before any tool is involved.
Review AI-assisted fundraising copy for language that flattens your voiceintermediate
AI fundraising tools optimise for conversion patterns across many organisations. The result is fundraising copy that sounds like every other organisation. Read yours aloud and ask whether it sounds like you.
Map the community relationships your CRM cannot captureintermediate
Your CRM records transactions. It does not record the relationship your education team has built with a local school over three years. Map that separately.
Run one programme decision per season based entirely on artistic judgement, not dataadvanced
Take the AI dashboard out of one programming decision. Make it on the basis of what your team believes matters culturally. Measure the outcome and learn from it.
Require any AI-drafted communication to be read aloud before sendingbeginner
Reading aloud surfaces AI-shaped phrasing faster than any edit. If it does not sound like a person from your organisation, rewrite it.
Ask your community partners what your AI tools are getting wrong about themintermediate
Your community partners have a direct view of where your data-driven assumptions miss the reality of their community. Ask them explicitly.
Identify the artists your AI recommendation engine is systematically missingadvanced
AI recommendation tools favour artists with strong digital footprints and existing audience data. They underweight emerging artists and those who do not self-promote online.
Write a programme note without referencing any AI summary of the workbeginner
Sit with the work, the artist, or the primary sources. Write from your direct encounter rather than from a synthesised summary.
Ask your board what they believe about your institution that no dataset would showintermediate
Boards carry institutional memory and community relationships that predate any analytics tool. That knowledge needs to stay in the room when AI-assisted reporting shapes strategy.
Five things worth remembering
The difficulty in a work of art is not a problem to solve. It is the point. Protect your willingness to sit with difficulty before reaching for an interpretation.
AI tools built on existing cultural production will optimise toward what has already been valued. The work that matters most is often the work those tools would not have recommended.
Your institution's voice is built from specific curatorial choices over many years. AI-generated communications will smooth those choices toward a generic nonprofit tone. Notice and resist.
Write down why you made a programming decision, not just what you chose. Institutional knowledge disappears faster than people expect when the people who hold it leave.
The relationship between an arts organisation and its community is not a data asset. It is built through direct presence and long-term trust. Protect the time that takes.