The short answer
AI for designers in November 2026 lives inside the tools: the Figma agent, generally available since 6 October 2026, explores directions, makes bulk edits and documents design systems on Full seats of paid Figma plans; Figma Make turns designs into clickable prototypes; Adobe Firefly generates images, video and audio with models Adobe trains on licensed and public domain content; and Canva resizes one design into every format. Hand over production, variations and documentation. Keep concept, taste, accessibility judgement and client decisions human.
| On the canvas | Figma agent: Full seats on Professional, Organization and Enterprise |
|---|---|
| Prototypes | Figma Make, with Gemini 4 Argon in the model picker on all plans |
| Commercially safer images | Adobe Firefly's own models, trained on licensed and public domain content |
| Every format from one design | Canva Magic Resize and Brand Intelligence |
| Exploration | Midjourney (no free plan; images public unless Stealth Mode on Pro or Mega) |
| Training defaults | Figma: on for Starter and Professional, off for Organization and Enterprise |
How designers use AI in 2026
The biggest change for designers this year is that AI moved from separate image generators onto the canvas itself. Figma, Inc. made the Figma agent generally available on 6 October 2026: it explores design directions, makes bulk edits, documents design systems and applies feedback, and it can build plugins and shader effects on request. Adobe's Firefly AI Assistant runs multi-step work across Photoshop, Lightroom, Premiere and Firefly in public beta. Canva AI returns layered, editable designs from a conversation.
That shifts where designers spend time. Producing ten variations, renaming four hundred layers or resizing a campaign into twelve formats is now minutes of AI work and a careful review. Choosing the concept, judging what is good, making sure it is accessible and defending it to a client is still yours. Our best AI for designers page ranks the tools.
The 6 design tasks worth handing to AI
1. Exploring directions early (Figma agent, Midjourney or Firefly Boards)
Ask for several distinct directions from one brief before you commit to any. Firefly Boards is an infinite canvas for mood boards and storyboards with teammates. Midjourney is known for a strong, distinctive style, which is useful for mood but needs care for client work (see the risks below). The "A campaign image set" stack on /stacks runs Midjourney, Adobe Firefly, Figma and Canva in order.
2. Bulk edits and clean-up (Figma agent and Figma's smaller AI tools)
Renaming layers, applying a new token across a file, adding interactions and swapping copy are ideal agent jobs. Figma says AI search and renaming layers cost no credits; other actions draw on your monthly AI credits.
3. Design system documentation (Figma agent)
Ask the agent to document components: purpose, variants, states and usage rules. Then edit for accuracy. Agent skills let you save repeated prompts as slash commands and publish them to your team, so the format stays consistent.
4. Clickable prototypes from a design (Figma Make)
Figma Make turns prompts and existing Figma designs into interactive prototypes and web apps. Since 30 October 2026 the agent and Make can also take context from Google Drive on all plans. A working prototype answers questions a static frame cannot.
5. Images and extensions for production (Adobe Firefly)
Use Firefly to extend a background, generate a product shot variation or create a sound effect for a motion piece. Firefly Custom Models (public beta since 19 March 2026) train a private model on your own images to capture a style or character. Outputs carry Content Credentials showing AI was used.
6. Every format from one master (Canva)
Once the key visual is approved, Canva's Magic Resize turns it into every size, Brand Intelligence keeps fonts and colours on brand, and Translate localises it. This is the part of a campaign that used to take a junior designer a week.
What designers should keep human
- The concept. AI generates options from what already exists. The idea that makes a brand different is your contribution.
- Taste and final selection. Pick, combine and refine. Never ship the first generation.
- Accessibility. Check contrast, type size, focus states and alt text yourself against your accessibility standard. AI misses things.
- User research. Watch real people use the prototype.
- Client relationships and rights decisions. Whether AI imagery is acceptable for a client is a conversation, not a default.
A starter AI stack for designers
| Job | Tool | Plan | Why |
|---|---|---|---|
| Interface design, agent, prototypes | Figma | Professional (Full seat) or Organization | Figma agent and Make; Organization has content training off |
| Images, video, audio | Adobe Firefly | Free daily generations, paid for volume | Adobe's own models are trained on licensed and public domain content |
| Formats and brand production | Canva | Pro or Business | Magic Resize, Brand Intelligence; Business content not used to train |
| Mood and exploration | Midjourney | Pro or Mega for Stealth Mode | Distinctive style; public images on lower plans |
| Briefs, rationale, case studies | Claude | Free or Pro | Careful writing for the words around the work |
Compare image tools in Midjourney vs Adobe Firefly, and see the Figma guide for credit details.
A 30-day AI rollout plan for a design team
- Days 1 to 3: settings and rights. On Figma Professional, open the team's Admin, Settings tab, AI section and decide on the Content training toggle (it is on by default for Starter and Professional). Agree with clients or legal which image tools are allowed for commercial work.
- Days 4 to 10: the boring half. Use the Figma agent for layer clean-up, bulk edits and documentation only. Note hours saved and credits used.
- Days 11 to 17: exploration. On one live brief, generate directions with AI before sketching, then compare them with your own sketches in the critique.
- Days 18 to 24: prototype. Build one flow in Figma Make and test it with five users.
- Days 25 to 30: production and rules. Run the final campaign resize in Canva, then write a one-page team policy covering tools, client disclosure, credits and accessibility checks.
The risks for designers, and how to manage them
- Rights and licences. Adobe's commercially safe promise covers its own Firefly models only; for partner models such as GPT Image or Nano Banana, Adobe says it is up to you whether they suit a commercial project. Midjourney's Terms of Service grant it a perpetual, worldwide, irrevocable licence to what you input and create, and companies with more than USD 1,000,000 gross annual revenue must use Pro or Mega to own their assets.
- Leaking unreleased work. Midjourney images are public on midjourney.com unless you use Stealth Mode on Pro or Mega, and turning Stealth on does not automatically make earlier public images private. Figma trains on content by default on Starter and Professional unless an admin turns it off.
- Credits. Figma's agentic features share one monthly credit budget, and a multi-screen flow can cost 300 or more credits. Weave tools cost 13 to 110 credits per run.
- Sameness. AI styles converge. Start from your own references and sketches, and use AI to widen the options, not to set the direction.
- Disclosure. Some clients and platforms require you to say when AI was used. Content Credentials on Firefly output help here.
For image prompts, see Midjourney prompts, and for picking an image tool, best AI for images.
Prompts for the 6 design tasks
One prompt per task. Replace the square brackets.
Prompts to copy
1. Five distinct directions from one brief
From this brief: [paste], propose five visually distinct directions. For each give a name, the mood in three words, colour and type approach, one reference style to avoid, and who it would appeal to most. Make them genuinely different from each other, not variations of one idea.
You get: A spread of directions to react to in the critique
2. Clean up a file (Figma agent)
In this file, rename all layers using the pattern [component/element/state], replace hard-coded colours with our tokens from [library], and list anything you could not match to a token instead of guessing.
You get: A tidy file and a short list of exceptions
3. Document a component (Figma agent)
Document the [component] component: purpose, every variant and state, spacing and token usage, do and do not examples, and accessibility notes (contrast, focus, keyboard behaviour). Use our documentation format from [page].
You get: Draft documentation to check and publish
4. A clickable prototype (Figma Make)
Turn the frames in [page] into a clickable prototype of the [flow] flow. Keep our components and spacing. Add realistic placeholder content, working navigation between screens and an error state for [step].
You get: A prototype ready for a user test
5. Extend an image for a new format (Adobe Firefly)
Extend this image to a 16:9 banner, continuing the background naturally on both sides. Keep the product and lighting unchanged and leave clear space on the left third for a headline.
You get: A production-ready extension with room for copy
6. Every channel from the master (Canva)
Resize this approved design for Instagram feed, Instagram story, LinkedIn post, a 300 by 250 display ad and an email header. Keep the headline readable in each, keep the logo clear space, and use our brand kit.
You get: A full set of formats to check, not rebuild