AI Workflow for AV
Introduction
At eleven at night after a show, a crew chief has a show report to write, a quote to revise and an email to a client in Spanish. An AI assistant can draft all three in ten minutes. It can also put a wrong power figure in the quote, a phrase in the email that means something else in Bogotá, and the client's unreleased product name into a system that keeps it.
This book is about getting the first without the second. It covers where AI already sits in AV, both on the laptop and inside the gear; the jobs it does well; how to ask it properly; and the checks that make its work safe to send: every number against its source, nothing confidential pasted in, nobody's voice or face used without their consent, and no AI at all when your hands should be on the show.
It names tools as examples, not recommendations; they change every few months. The habits do not.
How to Use This Book
Read it with an AI tool open. Each chapter has a task you can try on real work, and each answer shows what to check before you trust the result.
AI as a Tool in AV
Understanding AI's place in AV work.
AI in the workflow
AI (artificial intelligence) tools — for generating and editing content, drafting and summarising text, automating tasks, and more — are increasingly part of how work gets done, including in AV. AI Workflow for AV is about using these tools as practical aids across AV work: speeding up tasks, helping with content and media, assisting operations and communication, and enabling smarter automation. The key framing is that AI is a tool — a powerful aid that, used well, makes AV work faster and better, but that requires judgement to use wisely. AI is a tool to aid AV work, used with judgement.
This matters because AI is changing how work is done across industries, and AV professionals who can use AI tools effectively (and wisely) will work faster and better. AI won't replace AV craft (the hands-on, live, physical work), but it's becoming a valuable aid across many AV tasks. AI is an increasingly important tool for AV professionals to use well.
Most people in the trade already use it. AVIXA's research reported nearly 65 per cent of pro AV people surveyed using AI at work in 2025, and 76 per cent in a later AVIXA Index survey, mostly generative AI and assistants, while noting that use for technical work such as design and troubleshooting remained limited. Across all US workers, Pew Research found in 2025 that about one in ten used AI chatbots at work every day or a few times a week, mostly for research, editing and drafting.
In AV it turns up in two places. On the laptop, as a chat assistant such as ChatGPT, Claude, Gemini or Copilot, helping with the writing, reading and planning around a show. And inside the gear, in cameras that follow a presenter, microphones that find the talker, video bars that frame a room and software that captions speech. The first you choose to use; the second you need to understand, because it is already in the rooms you work in.
What AI can and can't do
Using AI well starts with understanding what it can and can't do. AI is good at generating and drafting (content, text, ideas, images), assisting and speeding tasks (summarising, organising, suggesting), and handling certain automation — fast, tireless, and capable, but it is not perfect or infallible: it can be wrong, make things up, lack real understanding, miss context, and produce generic or flawed results. So AI assists, but the human judges, verifies, and decides. Understanding AI's strengths and limits is essential to using it well. AI is a powerful but fallible aid — it assists; you judge.
This realistic view (AI as a capable but imperfect assistant) is the foundation for using it wisely: leverage its strengths (speed, generation, assistance) while compensating for its limits (verifying, judging, providing context and direction). Over-trusting AI causes errors; using it as a judged tool gets the benefits. The professional uses AI's strengths while managing its limits.
A chat assistant predicts plausible text. That makes it good at drafting, summarising, rewording, translating, explaining and organising, and unreliable as a source of facts. OpenAI's own research describes hallucinations as a model confidently generating an answer that is not true, and says they remain a fundamental challenge for all large language models, its own included.
In AV that means invented specifications, wrong pinouts, a menu path that does not exist on your firmware, a power draw off by a factor of two. It sounds exactly as confident when it is wrong. Treat every number, model name and procedure it gives you as a lead to check, not an answer.
Practice it
An assistant tells a technician that a projector draws 450 watts and that its lens shift is set from the main menu.
Where AI Helps in AV
Content, operations, and automation.
Content and media
A major area where AI helps in AV is content and media. AI can generate and assist with visual content (images, graphics, and increasingly video and generative visuals — useful for content creation, media servers, LED, and creative work), help with editing and media tasks (drafting, processing, organising media), and speed creative production. For creative-tech AV especially, AI content tools are a growing part of the workflow (generating or assisting with the visuals shows use). AI is a powerful aid for content and media work. AI helps generate and assist with AV content and media.
Used well (as a starting point or assistant, with human judgement and refinement), AI content tools speed and augment creative work; used poorly (blindly, without taste or rights-awareness), they produce generic or problematic results. The professional uses AI to aid content work while applying their creative judgement. AI augments content and media work when used with judgement.
AI already inside the gear
Automatic help in AV is older than chat assistants. Dan Dugan, widely recognised as the inventor of the automatic microphone mixer, developed gain-sharing automixing about fifty years ago as a theatre sound designer: however many people talk, the overall system gain is held at the level of one open microphone, which keeps a panel discussion clear and out of feedback.
The new generation uses machine learning. Panasonic's AW-SF100 auto-tracking software follows a presenter with a PTZ camera using body matching, face detection and deep learning, with no transmitter on the presenter. Shure's MXA920 ceiling array focuses pickup on talkers in its coverage areas, 9 by 9 metres by default and up to eight areas, and its IntelliMix processing includes AI echo cancellation, denoising and de-reverb. Logitech's RightSight 2 on the Rally Bar frames a room as a group, an active speaker or a grid of individual faces. NVIDIA Broadcast removes keyboard clatter, fan hum and room echo from a voice.
Live captions turn speech into text in tools such as Microsoft Teams, and Microsoft's own guidance is the AV technician's job description: captions become less accurate as the talker moves away from the microphone, and they need clear speech, little background noise, no talking over each other and good certified equipment. The AI is only as good as the signal you give it.
Know how to turn each one off, and when to. Auto-tracking can follow the wrong person, auto-framing can cut to an empty seat, and noise removal can eat quiet speech. On a show, a person decides what the audience sees and hears.
Paperwork: quotes, emails and show reports
This is where an assistant saves the most time. Give it your notes and ask for a draft: the reply to a client's change request, the show report from your scribbled timeline, a quote's scope and exclusions from the email chain, a pull sheet from an equipment list, a polite chase for an overdue invoice. Ask it to tighten, to make it shorter, or to make it firmer.
Then read every line as if a new assistant had written it, because one did. Prices, quantities, dates, model numbers and names are yours to check. The assistant cannot know that the client changed the start time yesterday.
Reading manuals and spec sheets
A 300-page manual is where an assistant can point you quickly: paste or upload a section and ask where it covers EDID, or what the difference is between two modes. Use the answer to find the page, then read the page. The manual is still the authority, and the assistant may be answering for a different model or firmware.
Troubleshooting with AI
Describe the system and the symptom precisely, the way you would to a senior colleague on the phone: the signal chain, the models, what works, what does not, and what you have already tried. A good assistant suggests what to check next and why. Then the meter, the tester and the method from AV Troubleshooting Methodology decide. Never change a setting on a live system because an assistant said to without knowing what that setting does.
Translating for international crews
Translation is one of the most useful things an assistant does on a show with an international crew or client: a safety briefing in Spanish, a reply in Portuguese, a menu in German. Keep it simple and literal, avoid slang, and for anything about safety, money or a contract, have a fluent person check it. A mistranslated rigging instruction is not a style problem.
Operations, communication, and automation
Beyond content, AI helps with operations and communication — drafting and summarising documents, emails, and plans; helping organise information; assisting with research and planning; and speeding the administrative and communication side of AV work. And AI enables smarter automation — AI-assisted systems and tools (and increasingly AI features in AV gear and software) that automate or enhance tasks. Across the AV workflow, AI can speed and assist many non-craft tasks. AI aids operations, communication, and automation across AV work.
These uses (drafting, summarising, organising, planning, automating) apply across AV roles (coordinators, managers, technicians, creatives) to work faster and handle the information and admin load. Used well, AI is a productivity aid across the workflow. AI helps across the operational and communication side of AV.
Practice it
A crew chief pastes her handwritten show notes into an assistant and asks for a show report. It comes back polished, and says doors opened at 6 pm and the keynote ran 12 minutes over.
- You know which gear in the room uses automatic features, and how to turn them off.
- Every draft is read by a person before it goes out.
- The manual or spec sheet confirms every technical fact.
- Safety and contract translations are checked by a fluent person.
Using AI Tools Well
Prompt, iterate, and get good results.
Using AI effectively
Getting good results from AI is a skill. It involves clear direction/prompting (telling the AI clearly what you want — the task, context, and desired output — since clear, specific input gets better results), iterating (refining your request and the AI's output through back-and-forth, rather than expecting perfection first try), and providing context (giving the AI the information it needs). Used skilfully, AI produces much better, more useful results. Using AI tools effectively is a learnable, valuable skill. The professional directs AI clearly and iterates to get good results.
This is like learning any tool: those who use AI skilfully (clear prompts, good context, iteration) get far more from it than those who use it crudely. Developing this skill (often called prompting or AI fluency) is increasingly valuable. The professional develops the skill of getting good results from AI.
Writing a good prompt
Anthropic's prompting guidance compares the assistant to a brilliant but new employee who lacks context, which is how to brief it. Be clear and direct about what you want. Give the context and the reason. Show an example of what good looks like. Say the format and the limits: length, tone, headings, a table.
Compare "write a quote" with this: "Draft the scope section of a quote for a one-day corporate conference in a 400-seat hotel ballroom. Include audio, two projection screens with IMAG from two cameras, and a three-person crew. List what is excluded: power distribution, rigging points and union labour. Plain English, under 200 words, as a bulleted list. Here is last month's quote for the style." The second gives the assistant what a new coordinator would need.
AI in your real tasks
Using AI well means applying it to your real AV tasks where it helps: drafting that email or document, generating or starting that content, summarising that information, brainstorming that plan, or automating that repetitive task — using AI as an aid in your actual workflow, with your judgement guiding it. The professional identifies where AI can speed or improve their work and uses it there, while doing the craft and judgement themselves. Applying AI to real tasks (well) is how it delivers value. The professional applies AI to their real tasks where it genuinely helps.
This practical integration (using AI where it helps, skipping it where it doesn't, always with human judgement) is how AI becomes a productivity boost rather than a gimmick or a liability. The professional weaves AI into their workflow thoughtfully. The professional uses AI as a real, judged aid in their work.
Keep the prompts that work. A short library of your own, for the show report, the change-request reply, the quote scope and the crew briefing, turns ten minutes of explaining into one. Keep them with your templates, not only inside a chat history.
Practice it
A technician asks an assistant "write a crew briefing" and gets a generic page.
Using AI Wisely
Judgment, verification, and ethics.
Judgment and verification
Because AI is fallible, using it wisely requires judgement and verification. Judgement — deciding when and how to use AI, evaluating its output critically (is this good, correct, appropriate?), and remaining the decision-maker (AI assists, you decide). Verification — checking AI's output for accuracy and quality (AI can be confidently wrong or make things up, so its output must be verified before relying on it). Never blindly trust or ship AI output — review and verify it. Judgement and verification are essential to using AI responsibly. The human judges and verifies; AI never decides unchecked.
This is the core discipline of good AI use: leverage AI's speed and capability, but apply human judgement and verification so errors don't slip through. The professional uses AI as a powerful assistant while staying firmly in control and checking its work. The professional always judges and verifies AI's output.
Checking every number
The most public lesson came from a courtroom. In Mata v. Avianca in 2023, a federal judge in New York fined two lawyers and their firm 5,000 dollars for submitting court filings that cited six cases that did not exist, invented by ChatGPT, and for standing by them when questioned. The same judge wrote that there is nothing inherently improper about using a reliable AI tool. The failure was not checking.
In AV, check every number against a source you could show someone: the manufacturer's spec sheet, the manual, the venue's power schedule, a calculation you did yourself, a meter reading. If the assistant gives a source, open it. If you cannot find the number anywhere but the chat, it does not go in the quote.
Confidentiality: what never to paste in
In 2023 Samsung engineers pasted faulty source code, an equipment problem and a confidential meeting's notes into ChatGPT to get help; Samsung then restricted generative AI tools on company devices. Whatever you paste into a tool leaves your control.
What happens to it depends on the tool and your settings. OpenAI lets users of its consumer ChatGPT turn off model training in its data controls, and says it does not train on data from its business, enterprise and API products by default. Anthropic says it uses consumer Claude chats for training only if you allow it, or if a chat is flagged for safety review. These terms change: read your tool's, and your employer's policy, before you paste.
Never paste in client names with confidential content, unreleased products, show content under NDA, passwords, network addresses and credentials, other people's personal details, or anything your contract says stays private. Describe the problem without them: "a 40-room corporate client" does the same job as the client's name. The CTS Code of Ethics forbids revealing what you learn in your work without consent, and that includes revealing it to a machine.
Ethics, rights, and limits
Using AI wisely also means respecting ethics, rights, and limits. Accuracy and honesty — don't pass off unverified AI output as fact, and be honest about AI use where it matters. Rights and originality — be careful with AI-generated content and intellectual property (rights, ownership, and originality are evolving and important issues with AI content). Appropriateness and disclosure — use AI appropriately and disclose its use where expected. And limits — recognise what shouldn't be left to AI (judgement calls, the human craft, things requiring real accountability). Using AI ethically and within its proper limits is essential. The professional uses AI ethically, respecting rights and limits.
This responsible approach (accurate, honest, rights-aware, appropriately disclosed, within sensible limits) protects you and others and uses AI as a force for good rather than harm. As AI grows, using it ethically and wisely is increasingly important. Combining AI as a tool, where it helps, using it effectively, and using it wisely is AI workflow for AV. The professional uses AI to work better, with judgement, verification, and ethics. This is using AI well. (AI tools and norms evolve fast — keep learning and applying good judgement.)
Copyright in the US needs a human author. The US Copyright Office said in January 2025 that AI-generated material can be protected only where a human has determined enough of its expressive elements, and that merely writing prompts is not enough; using AI as a helper inside a human-made work does not stop that work being protected. A show graphic generated entirely by AI may belong to nobody.
A person's voice and face belong to them. Tennessee's ELVIS Act of 2024 added voice to the state's law protecting a person's likeness against misuse, including by AI. The FCC ruled in February 2024 that AI-generated voices in robocalls count as artificial voices under US telephone law, and the EU AI Act requires anyone publishing a deepfake to disclose that it was artificially generated or manipulated. Never clone a presenter's voice, fill in a missing line or alter someone's face on screen without their written consent, and say so when content is synthetic.
When not to use it
Some moments need a person and nothing else: a live show, a safety decision, a rigging calculation, a call on whether a system is safe to energise, and a conversation with an upset client. Look things up before the doors open, not during the keynote. During the show, eyes and hands stay on the show.
Practice it
A video engineer wants help with a network fault on a client's show and is about to paste the switch configuration, which includes the client's network names, IP addresses and passwords.
- Every number has a source outside the chat.
- Nothing confidential, personal or secret went in.
- Any synthetic voice, face or image has consent and is labelled.
- A person read it and would sign their name to it.
AI as a Wise Tool
Working better with AI, thoughtfully.
AI in AV understood
Pulling it together, AI workflow for AV is using AI as a practical tool — understanding it as a powerful but fallible aid, applying it where it helps (content, operations, communication, automation), using it effectively (clear direction, iteration), and using it wisely (judgement, verification, ethics). Used well, AI makes AV professionals faster and more capable; used poorly, it causes errors and problems. Understanding AI as a tool to use wisely positions you to benefit from it. AI in AV understood is using AI to work better, thoughtfully. AI is a powerful tool for those who use it with skill and judgement.
AI is a fast-growing, increasingly important part of how work is done, and AV professionals who use it well (effectively and wisely) gain an edge, while AV craft itself remains human. Understanding AI workflow — what it does, where it helps, using it effectively and wisely — is an increasingly valuable capability. AI is a tool worth using well in modern AV work. (As AI evolves rapidly, keep your knowledge and judgement current.)
Developing AI fluency
Developing AI fluency means using AI tools regularly and thoughtfully on real tasks (learning what they do well and how to direct them), keeping up with the fast-evolving tools (new and better AI tools appear constantly), building the skill of getting good results (prompting, iterating, integrating AI into your workflow), and always practising judgement, verification, and ethical use. AI fluency is built by using AI a lot, wisely. The most valuable approach combines effective use with strong judgement — getting AI's benefits while avoiding its pitfalls. Deliberately developing AI fluency, thoughtfully, is the path forward.
From here, growing means deepening your AI fluency, applying AI across more of your work where it helps, and keeping up as AI tools and their role in AV evolve — while keeping human judgement and craft central. AI is a tool that will keep growing in importance; using it well is an increasingly valuable skill. The principles this book provides — AI as a judged tool, used effectively and wisely — are durable even as the tools change. Keep using AI thoughtfully, keep your judgement sharp, and you'll work better with AI in AV. (Tools change fast; the principles of skilful, wise use endure.)
Fluency is a habit, not a subscription. Try the assistant on one real task a week, keep the prompts that work, check what it gives you every time, and learn which features in the gear you install and operate are automatic and how they fail. The tools will change; drafting fast, checking hard and keeping the client's secrets will not.
Practice it
A presenter's microphone dropped out for one sentence in a recorded keynote, and the producer asks the editor to "just generate the missing words in her voice".
Glossary
| Generative AI | Software that produces text, images, audio or video from a request |
| AI assistant | A chat tool such as ChatGPT, Claude, Gemini or Copilot |
| Prompt | The request you give an assistant |
| Hallucination | A confident answer from AI that is not true |
| Automixer | A mixer that opens and balances microphones automatically |
| Gain sharing | Holding total system gain at the level of one open microphone |
| Auto-tracking | A camera following a presenter automatically |
| Auto-framing | A camera framing the people in a room automatically |
| Live captions | Speech turned into on-screen text as it is spoken |
| Data controls | Settings that decide whether a tool learns from what you type |
| Deepfake | Synthetic audio, image or video of a real person |
| Consent | A person's permission, best in writing |
Where to Go Next
AV Troubleshooting Methodology is the method an assistant's suggestions must fit into, Client and Stakeholder Communication covers the conversations it helps draft, and AVIXA Membership & Ethics the confidentiality and honesty duties that apply whatever tool you use.
What the gear actually looks like
Every device named in this book, in one place. Recognition is the whole point of Chapter 4 — use this page to attach a picture to each name before you meet it on a dock at six in the morning.
Find the destination.
Work along the path between them.