Use AI image tools to make two or three mood boards in clearly different directions, label them as concept art, and send them with a short approval form. Draw the layout yourself from measured dimensions. AI is quick at atmosphere and poor at scale, so venue measurements, capacities and exit routes stay with a person.
The real gain is not prettier boards, it is earlier agreement. A client who has seen three directions in the first meeting rules out two of them in minutes, instead of rejecting a finished design a fortnight later. The risk is the reverse: a polished AI image makes a promise the budget and the room cannot keep. The process below keeps the speed and removes the promise.
What AI is good at in event design, and where it stops
Split the design job into two kinds of work. The first kind is about feeling: colour, texture, mood, style of florals, lighting temperature, how a room might look at dusk. Image generators handle this well, and a planner who used to spend an evening collecting reference photos can now produce a board in under an hour. The second kind is about fact: how many round tables of ten fit in a 12 by 18 metre hall, where the fire exits are, how wide the aisle must be for a wheelchair, whether the stage is visible from the back row. Image generators do not know any of this, and they will draw a room that looks plausible and is physically impossible.
Think of the output as an illustrator's sketch, not a drawing. A sketch can show that a long banquet table under hanging greenery feels right. It cannot show that the greenery will survive the venue's rigging rules, or that thirty people fit along one side. Keep that line clear in your head and in every client conversation.
| Job | Use AI? | Why |
|---|---|---|
| Colour and mood directions | Yes | Fast, cheap, easy to discard |
| Table, florals and lighting looks | Yes, as concepts | Shows the feeling; sourcing is a separate step |
| Signage and stationery ideas | Yes, then redraw | Generated text is often misspelt; finalise in a design tool |
| Floor plan with real dimensions | No | Needs scale, capacities and exits, checked by a person |
| Sightlines, stage and rigging | No | Safety and venue rules, not styling |
| Cost of the design | Assist only | Supplier quotes decide it; see the budget tutorial |
Start from a design brief, not a prompt
Images are only as good as the brief behind them, and clients are usually vague. Before opening any tool, write a half-page brief with the client: the occasion, guest count, venue type, season and time of day, three words for the feeling, three things they definitely do not want, the budget band and any brand colours or cultural considerations. Where the event has a brand (a product launch, an awards night), collect the logo, brand colours and one existing image the client likes.
Ask the AI assistant of your choice to turn that brief into three mood directions in words before you generate any images. This is cheaper to revise than pictures and surfaces disagreements early.
Here is a client brief for an event. Propose three distinct design
directions. For each give: a name, a one-sentence idea, a colour
palette of five colours with plain-language names, three materials or
textures, a lighting note, and one thing it would NOT suit.
Do not name living artists or brands. Do not mention anything the
brief says to avoid. Directions must differ in mood, not just colour.
BRIEF: [paste half-page brief]Notice the request for "one thing it would not suit". It forces the model to flag trade-offs, and those become your talking points with the client. Edit the directions by hand before generating images, because a name and a palette that you chose will survive client comments better than the AI's first pass.
Worked example: a product launch for a software reseller
Say a small B2B software reseller wants a customer launch evening for 80 guests in a hired loft space. The brief says modern, not corporate, evening, brand colour a deep teal, no balloons, budget band mid-range, and no stage. Here is how the design round runs.
- Directions in words (10 minutes). The assistant proposes "Studio Night" (charcoal, teal, brass, low pendant lighting), "Greenhouse" (teal, sage, white, plants and long tables) and "Gallery" (white walls, teal accent lighting, plinths for product displays). It flags that Greenhouse would need real plants delivered and watered, and that Gallery would feel cold with 80 people standing.
- Images (45 minutes). For each direction you generate four to six images: a wide room view, a table or bar close-up, a lighting detail, a signage idea. You discard the ones with distorted furniture or garbled lettering, which is most of the first batch.
- Board (20 minutes). You assemble one board per direction with the palette, three images, the "would not suit" note and a plain line: "Concept images created with AI to show mood; final styling subject to venue and budget."
- Client review (a 30-minute call). The client picks Greenhouse with Gallery's lighting. The decision is recorded in writing.
- Layout (separate, human). You measure the loft, take the capacity from the venue's licence, and draw the plan with real dimensions. AI had no part in this step.
In total, roughly two hours of design work, against the half-day or more a planner would normally spend gathering references. Those times are illustrative, not guarantees; your first few boards will take longer while you learn which prompts work.
Prompting images for a mood board
Good event prompts describe a scene the way a photographer would brief a shoot: the room, the light, the time, the materials and the camera angle. Avoid brand names and living artists, partly for tone and partly because their use can create rights questions you do not want to discuss with a client.
Wide photo of a long banquet table in a converted warehouse at dusk,
deep teal linen runners, brass candlesticks, trailing eucalyptus and
white ranunculus, warm low pendant lights, shallow depth of field,
realistic event photography, no people, no text.The final phrases matter. "No text" cuts garbled lettering. "No people" avoids odd hands and faces, and leaves room for the client's imagination. For a variant, change one element at a time (the flowers, then the lighting) so you learn what each phrase does and can show the client a controlled comparison.
Some tools help keep a consistent look across a set. Midjourney's style reference and moodboard features work on V6 and later models, so a board of four images can share one visual language. Adobe's Firefly has its own style reference and strength control, but those work only with Adobe's own Firefly Image models, not the partner models offered inside the app. Canva has AI image and design features that draw on a monthly AI allowance, which differs by plan, so check the allowance on the pricing page before you promise a client twenty variants. Check each tool's current terms and plan details yourself, because they change often.
For a deeper look at keeping one look across a whole set of images, this tutorial on Midjourney style references goes step by step. If you are weighing this against the way other design trades do it, how interior designers use AI for mood boards shows a near neighbour of the same workflow.
The layout draft: where a person must hold the pen
A layout is a safety document before it is a design document. The numbers come from the venue, not from an image. Gather these before drawing anything: the venue's floor plan or a measured sketch, the licensed capacity, exit locations and widths, the position of fixed items (pillars, radiators, stage, bar, power), the loading route, and any access requirements for guests who use mobility aids.
Then draw the plan in whatever tool you already use: a simple floor-plan or drawing application, a spreadsheet grid where each cell is a set distance, or squared paper. Add tables and chairs at real sizes. Use AI at this stage only as a thinking partner on arithmetic and options, never as the drawing. A prompt such as the one below works well for the sums, and you check the result yourself.
I have a hall 12 m wide by 18 m long. Fixed items: a bar along one
short wall (2 m deep), two pillars each 0.5 m square at 6 m and 12 m
along the centre line, and exits on both long walls. I need seated
dinner for 80 at 1.8 m round tables of ten. Suggest three table
arrangements with the number of tables and the clear aisle width
between table edges. State every assumption. I will check the numbers
against the venue plan.The AI will return arrangements and assumptions. Your job is to test each one against the plan: do aisles meet the venue's minimum, is there clear access to every exit, and does anyone sit behind a pillar with no view of the speaker. Arithmetic from a language model can look confident and be wrong, so redo any figure that will appear on a document. If the venue supplies its own layout rules or a capacity limit, those win over anything the AI suggests, every time.
A client approval round that actually closes
Design approvals drag when feedback is vague ("make it more elegant"). Give structure and the round shortens. Send the boards with a one-page approval sheet that asks for four things: which direction (pick one), what to keep, what to change (maximum three items) and what to remove. State the number of revision rounds included in your fee, and say that changes after sign-off are costed separately. Note in writing that the images are concept art and that the final look depends on supplier availability, venue rules and the agreed budget.
A sample approval sheet might read (illustrative):
Event: [name], [date]
Direction chosen: Greenhouse / Studio Night / Gallery (circle one)
Keep: ______________________ (up to 3 items)
Change: _____________________ (up to 3 items)
Remove: _____________________
I understand the images are concept art made to show mood. Final
styling is subject to venue rules, supplier availability and budget.
Signed: ______ Date: ______Use the AI to draft the cover email and to summarise the client's feedback into a change list, then read both before sending. If a client's reply is long and rambling, ask the assistant to extract the decisions and questions into a table, and confirm that table with the client in one line. That removes the common failure of a revision round that fixes the wrong thing.
More examples of the same method
A small wedding. The brief asks for "romantic but not fussy" with a spring palette. Two of three directions come back too ornate; the third, "Orchard", works. The planner then asks for florals only, to pass to a florist as inspiration, and states clearly that the florist will interpret it in season. For floral wording after approval, this tutorial on wedding flower proposals follows on.
An away day for a small HR consultancy. The client wants a calm, professional room for 25 and a hand-out pack. The mood board is quick; the real value is a signage set. The AI drafts sign text, but one generated sign had a misspelt word inside the image itself. The fix: use AI for the look, then retype all sign text in the design tool and proofread it.
A charity gala. The planner needs a "black tie, warm and hopeful" look within a tight budget. The assistant's directions all assume imported flowers. Asking it to redo the directions "using only seasonal, locally sourced stems and reusable décor" gives achievable boards. Adding the budget to the prompt changes the output more than any stylistic phrase.
A pop-up dinner. The concept is a long table outdoors. The generated images show a perfect lawn and no wind. The planner adds a wet-weather note to the board, with a second image of the same table under a marquee, so the client has seen the plan B before it is needed.
A trade-show stand for a small manufacturer. The exhibitor wants three stand looks. AI is good at showing graphic styles, but the stand footprint, power points and height limits come from the organiser's exhibitor manual. The planner pins those limits to the top of the board so the client does not fall in love with a three-metre tower that is not allowed.
What goes on a board, and what stays off it
A board that tries to do everything confuses the client. Keep each one to a single page with five parts: a name and one-line idea, a palette of five colours with names (so the client can say "swap the sage for something warmer"), three to five images, a materials and lighting note, and the label that says these are concept images. Leave off prices, supplier names and floor plans. Prices turn a creative conversation into a negotiation before the client has chosen a direction, and plans invite a precision the concept does not have.
Test the palette in the real world before you present it. Screens flatter colour, and teal on a laptop can look grey under venue lighting or on printed stationery. Ask the AI for palette names and approximate descriptions, then confirm real colour values from fabric swatches, paint or print proofs. If any guest-facing text will sit on a coloured background (menus, place cards, signs), check that the contrast is comfortable to read at arm's length in dim light. A beautiful dark-on-dark scheme that nobody can read is a design failure, however good the image looks.
One last tip for presenting: show the boards in the order you would choose them yourself, strongest last, and say which you recommend and why. Clients value a view, and the AI cannot give one. Your judgement about what suits this room, this budget and this client is the part of the service the tool does not replace.
Mistakes planners report, and how they show up
The impossible image. A board shows a ceiling installation the venue will not permit, and the client wants exactly that. Prevention: add a "venue rules" line to every board listing what is not allowed (rigging, open flame, wall fixings) and check it before the call.
Scale drift. The images imply a room twice the real size, because a generator draws airy rooms. Prevention: include the real room dimensions in the board and show a photo of the actual venue next to the concept.
Look-alike risk. A generated image closely resembles a recognisable commercial venue or a competitor's event. Prevention: avoid naming places or brands in prompts, and look at results with fresh eyes before they leave the building.
Unlabelled AI images. A client forwards a board to a supplier as if it were a reference photograph of a real installation. Prevention: the caption line, and a habit of watermarking or labelling concept images; the guidance in how to label AI-generated images and video covers the wording.
Check it worked: three questions before sign-off
After approval, run three quick tests. Could a supplier price from the board and the sheet without calling you? Does every measurement and capacity on the layout trace back to a document from the venue? Would the client be unsurprised if the final room were handed to them exactly as built? If any answer is no, fix it before ordering anything. Then take the approved direction into costing and the run sheet; how event planners use AI for budgets and run sheets picks up from there, and the full proposal write-up is in writing event proposals with AI.
Keep a folder per event: brief, prompts, rejected and chosen images, the approval sheet and the signed-off layout. When a dispute arrives months later, "this is what they chose, and this is what they were told" is a short and calm conversation. Over time the folder also becomes your own library of tested prompts, which is where the speed really builds.
Costs and time to expect
For most planners the tool cost is small: a monthly plan on one image tool, or the free allowance on the design app you already use. Check each vendor's pricing page before you commit, because plans and credit systems shift. The larger cost is your time spent rejecting bad images, which falls quickly once your prompts are tuned. Budget about two hours for a first board and under one hour once you have a reusable brief and prompt set. The time that does not shrink is the layout work, which is the point: it should not.
Questions planners ask about AI event design
Can AI produce an accurate floor plan for a venue?
No. Image generators draw convincing rooms, not scaled plans. Use them for atmosphere and ideas, then draw the real layout from measured dimensions in a floor-plan tool, a spreadsheet grid or on paper. A person should check every distance, exit route and table spacing against the venue's own plan.
Who owns the images an AI tool creates for a client event?
It depends on the tool's terms and on the law where you work, and the major providers' terms assign you rights in outputs only 'if any' exist. Read the terms of your plan, avoid prompts that name living artists or brands, and tell clients the images are concept art, not final artwork.
Should I tell clients that mood boards were made with AI?
Yes, in one plain sentence on the board. It sets expectations that the images show a feeling, not a guaranteed result, and it avoids a client expecting a ceiling, flower wall or lighting effect the budget cannot deliver.
How many mood board options should I show a client?
Two or three. One invites a yes-or-no verdict, five invites a committee. Make the options genuinely different in direction, then ask the client to pick one and name two things to change.
Further reads
- Which Event Planning Tasks Should You Hand to AI First? — Where else in event planning AI saves time.
- AI Event Description Generator: Ticket-Page Copy for Small Events — Write the ticket page once the design is settled.
- How Pubs and Bars Can Use AI for Events, Rotas and Stock — Event-night operations once the look is agreed.
- A Wedding Planner's AI Workflow From Enquiry to Final Timeline — A full wedding workflow from enquiry to timeline.
- Can AI Design a Restaurant? Layouts, Concept Images and Limits — The same limits applied to a restaurant fit-out.
- AI Entertainment for Events: What Party Hire Firms Can Add to a Package — Five AI add-ons a party hire firm can sell, a costed caricature station for 150 guests, and the venue and consent points to settle before listing any.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Midjourney and Adobe Firefly help pages (style reference and moodboard features), Canva pricing page for AI allowances (checked October 2026), OpenAI and Anthropic terms on output rights.