Every creative professional in 2026 has the same conversation at dinner parties. Someone asks: “So, are you using AI?” The answer is always more complicated than yes or no.
The reality on the ground — in studios, agencies, freelance workflows, and independent practices — is that AI tools have moved from novelty to infrastructure. Not everywhere. Not for everything. But in specific, high-leverage parts of the creative process, AI is now as unremarkable as Photoshop or AutoCAD.
This is a field guide to what is actually being used, by whom, and for what — stripped of launch-event hype and doom-scenario anxiety.
The framework: augmentation, not replacement
Before the tools, the principle. The creatives doing the most interesting work with AI share a common approach: AI handles volume; humans handle judgment.
Generating two hundred logo variations is AI work. Choosing the one that communicates the brand’s emotional truth is human work. Removing a background is AI work. Deciding whether the image should exist at all is human work.
The tools below are organized by discipline, with honest assessments of what they do well, what they do poorly, and where the human must remain in the loop.
For photographers and image makers
Adobe Firefly (integrated across Creative Cloud)
What it does: Generative fill, text-to-image, style transfer — embedded inside Photoshop and Lightroom rather than standalone. Best for: Retouching workflows, extending backgrounds, removing objects, rapid compositing. Limitation: Output can look “Adobe-clean” — technically proficient but lacking the idiosyncrasy of specialized models. Who uses it: Commercial photographers, retouchers, advertising studios already in the Adobe ecosystem.
Midjourney v7
What it does: The highest-quality text-to-image generation available for stylized, aesthetic-forward imagery. Best for: Concept development, mood boards, editorial illustration, fashion look development. Limitation: Not photorealistic for all subjects; no native PSD export; Discord-based interface still frustrates some professionals. Who uses it: Art directors, concept artists, editorial designers, indie publishers.
Magnific AI
What it does: AI upscaling and enhancement that adds detail rather than just enlarging pixels. Best for: Making low-resolution reference images usable; enhancing AI-generated concepts to print resolution. Limitation: Can hallucinate detail that was not in the original — problematic for documentary work. Who uses it: Print designers, large-format exhibition work, game asset pipelines.
Aftershoot and Imagen AI
What they do: AI-powered culling and editing of photo sessions — selecting the best shots and applying consistent Lightroom presets. Best for: Wedding and event photographers drowning in volume (a 500-image wedding becomes manageable in an hour). Limitation: Culling taste reflects training data, not your specific aesthetic — requires calibration. Who uses it: High-volume portrait and event photographers.
For graphic designers and brand studios
Figma AI (native features)
What it does: Auto-layout suggestions, text generation for placeholder content, asset search, design system management. Best for: Accelerating the mechanical parts of UI/UX design — wireframes, component libraries, responsive variants. Limitation: Cannot make brand-strategic decisions; generated copy is obviously placeholder unless heavily edited. Who uses it: Product designers, UI teams, agency wireframing workflows.
Canva Magic Studio
What it does: Template-based design with AI generation for social content, presentations, and marketing materials. Best for: Small businesses and solo creators who need professional-looking output without a design degree. Limitation: Everything looks like Canva — a aesthetic ceiling that brands outgrow quickly. Who uses it: Social media managers, small business owners, non-designers in marketing roles.
Runway ML
What it does: AI video generation, motion graphics, green screen removal, image-to-video animation. Best for: Short-form video content, animating still images, rapid prototyping of motion concepts. Limitation: Video generation quality varies wildly; temporal consistency remains unsolved for longer clips. Who uses it: Motion designers, social content teams, music video directors in pre-production.
Krea AI
What it does: Real-time AI image generation with style control — sketch input becomes rendered output live. Best for: Live concept sketching, industrial design visualization, rapid iteration in client meetings. Limitation: Still maturing; best for exploration rather than final deliverables. Who uses it: Product designers, automotive concept teams, architectural visualization studios.
For writers and editors
Claude and GPT-4o (via API or interface)
What they do: Long-form drafting, research synthesis, structural editing, tone adaptation. Best for: First drafts, summarizing research, generating variations of headlines and copy, overcoming blank-page paralysis. Limitation: Factual errors, generic phrasing, and the subtle wrongness of machine-generated prose under time pressure. Who uses it: Content teams, copywriters (for drafts, not finals), journalists (for research, not publication).
Grammarly and ProWritingAid (AI-enhanced)
What they do: Style, grammar, clarity, and tone analysis — now with generative rewriting suggestions. Best for: Polishing existing human-written copy; catching errors in high-volume publishing workflows. Limitation: Can homogenize voice; aggressive suggestions flatten distinctive writing styles. Who uses it: Editors, content marketers, non-native English professionals.
Descript
What it does: AI-powered audio and video editing through text — edit a transcript and the media edits itself. Best for: Podcast production, interview editing, video content with heavy dialogue. Limitation: Overdub (AI voice cloning) is ethically fraught and technically detectable. Who uses it: Podcasters, YouTube creators, documentary editors.
For filmmakers and video
DaVinci Resolve (AI features)
What it does: Auto color matching, facial recognition for grading, AI noise reduction, smart reframing. Best for: Color grading workflows, post-production on tight deadlines. Limitation: AI color suggestions require human correction for creative intent. Who uses it: Colorists, indie filmmakers, post-production houses.
Topaz Video AI
What it does: Upscaling, de-noising, frame interpolation for slow motion, stabilization. Best for: Restoring archival footage, upscaling content for modern displays. Limitation: Interpolation creates artifacts in fast motion; not magic. Who uses it: Documentary filmmakers, archive restoration, content repurposing teams.
Sora and competing video models
What they do: Text-to-video generation — describe a scene, receive a moving image. Best for: Pre-visualization, storyboarding, concept reels for pitch meetings. Limitation: Not yet reliable for final production; rights and licensing unclear for commercial use. Who uses it: Directors in pre-production, advertising agencies for client pitches.
The workflow that actually works
Creative professionals who integrate AI successfully tend to follow a similar pattern:
- Define the human judgment first — what decision must a person make?
- Use AI for the volume leading to that decision — variations, drafts, options, processing
- Apply human taste at the selection point — choose, refine, reject
- Use AI again for mechanical finishing — resizing, formatting, export, delivery
- Never publish AI output without human review — the cost of being wrong exceeds the time saved
What AI cannot do for creatives (yet)
- Understand brand soul — the intangible emotional truth that separates good design from correct design
- Make ethical decisions — about representation, manipulation, attribution, and consent
- Replace client relationships — the conversation that reveals what the client actually needs versus what they asked for
- Generate genuine taste — the accumulated judgment that comes from years of looking, failing, and refining
- Take responsibility — when the work is wrong, the human is accountable, not the model
The honest bottom line
AI tools for creatives in 2026 are not revolutionizing creativity. They are revolutionizing the parts of creative work that were never creative to begin with — the resizing, the culling, the background removal, the tenth draft of boilerplate copy, the wireframe that takes forty-five minutes of dragging rectangles.
The creative decision — what to make, why it matters, who it serves — remains entirely human.
The tools are good now. They will get better. The practitioners who thrive will be those who learn to wield them without letting them wield the work.
That distinction is everything.
Lumen is edited by Leo Hartmann. Related: AI in Architecture · Ethics of Synthetic Companions