Case Study: Elena's Slide Deck Upgrade — DALL·E for Professional Presentations

The Problem With Stock Photos

Elena builds decks for a living. As a management consultant specializing in organizational transformation, she routinely creates 40-60 slide presentations for C-suite clients on complex topics: change management, digital transformation, workforce restructuring. The content is rigorous. The visuals, by her own admission, had always been the weakest element.

The problem was not access to images — her firm had expensive stock library subscriptions. The problem was that the images were obviously stock photos, and the subjects were always slightly wrong. A slide about "breaking down organizational silos" would get illustrated with a stock photograph of a brick wall, or a generic "team collaboration" photo that could have been taken in any office in the world circa 2018. The images communicated nothing and sometimes made sophisticated points feel generic.

Elena had resisted trying AI image generation because she assumed the workflow overhead would exceed the benefit. She changed her mind during a late-night prep session before a board presentation on digital transformation strategy.

The Experiment

It was 10pm the night before a major client presentation. Elena had 42 slides done and 8 remaining. She was stuck on the section covering the transformation roadmap — specifically, the visual metaphor that would carry the conceptual framework across three slides.

She was building in PowerPoint. Her current solution was a weak stock photo of a road disappearing into horizon — the cliché of all transformation visual clichés.

She had ChatGPT open for a different reason and, almost as a test, typed:

"I need a visual metaphor for a consulting presentation slide about organizational digital transformation. The concept is: organizations moving from a fragmented, legacy state to an integrated, future-ready state. It needs to feel sophisticated and professional — like something you'd see in a high-end strategy publication. Not clichéd (no roads, lightbulbs, gears, or bridges please). What concept would work, and can you generate it?"

ChatGPT responded with three concept options before generating anything:

  1. A transformation from a scattered archipelago of isolated islands to an interconnected coastal region — fragmentation to cohesion, all in one metaphorical landscape
  2. A before-and-after showing the same building viewed through different light — one flat and bureaucratic, one warm and dynamic — without literally comparing
  3. An abstract representation of signal coherence: visual noise resolving into clear pattern, seen from above

Elena chose the archipelago concept. It felt fresh, could carry across multiple slides, and had enough visual richness to be interesting at large size on a conference room screen.

She asked ChatGPT to generate a first version.

The Generation and Iteration Process

The first generation was interesting but too literal — it looked like an actual satellite map. Elena refined in conversation:

"This looks too much like a real geography map. I want it more stylized — something that reads as illustration, not photography. Think infographic-quality illustration, clean lines, professional corporate design aesthetic. Same concept but clearly conceptual rather than realistic."

Second generation: better stylization, but the color palette was too bright. It would clash with her deck's muted blue-gray color scheme.

"The concept and style are right. Can you adjust the colors? The deck uses a muted navy and slate palette. I'd like this to feel consistent with that — desaturated blues and grays, perhaps a single accent color (I'm thinking a warm amber/gold for the 'connected' side of the image to contrast with the colder isolated side)."

Third generation: Elena was satisfied with the direction. She upscaled the image in ChatGPT, downloaded it, and dropped it into PowerPoint.

But now she had a different problem: the metaphor needed to carry across three slides. The first slide showed the fragmented state. The second needed to show the transformation in progress. The third needed to show the connected end state.

She continued in conversation: "I want two more images that form a progression with the first image I approved. They should look like they're part of the same series — same illustrative style, same palette, same visual logic. Second image: the islands are beginning to show bridges and connections forming between them. Third image: the archipelago has become a cohesive coastal region, integrated, the islands now clearly connected with roads, bridges, and a unified visual identity."

She had to iterate each of these two more times — the "in progress" image in particular kept coming out either too complete or too chaotic. But by 11:15pm, she had three images that worked as a visual sequence.

Building the Prompt Library

After that first session, Elena started systematically building a library of prompts for the image types she generated most frequently. She spent two hours on a Saturday afternoon generating reference images and refining prompts for each type.

Her core five:

Conceptual metaphor (transformation/change themes):

Professional conceptual illustration for management consulting presentation,
[specific metaphor concept], sophisticated editorial style, muted [color palette],
clean minimal aesthetic, no text, suitable for full-bleed presentation slide,
high-end strategy publication aesthetic, not clichéd corporate imagery --ar 16:9

Abstract framework visualization:

Abstract professional visualization of [framework concept — e.g., "interconnected systems"
/ "organizational layers" / "convergence of multiple forces"], minimal vector illustration style,
[color palette], clean white space, suitable for presentation slide background,
corporate design aesthetic --ar 16:9

Human/organizational moment:

Editorial photography of [specific workplace scenario], authentic documentary style,
modern professional environment, natural light, diverse professionals,
candid not posed, warm but serious atmosphere, suitable for executive presentation --ar 16:9

Data/technology concept:

Professional illustration of [technology or data concept], technical but accessible,
sophisticated minimal design, [color palette], suitable for C-suite presentation,
editorial business magazine aesthetic --ar 16:9

Cover/section divider image:

Full-bleed presentation cover image for [topic], [visual concept that represents topic],
sophisticated minimal aesthetic, [firm's color palette], premium corporate communication
design, no text in image --ar 16:9

These templates, plugged into DALL·E 3 in ChatGPT with slight modifications for each specific use, produced consistent and usable results within 2-3 iterations.

Handling Image Rights for Client Work

Elena raised the rights question with her firm's general counsel before using AI-generated images in client deliverables. The guidance she received:

For internal presentations and working documents: low risk, proceed.

For client-facing deliverables where images would appear in final reports, presentations, or published materials: use OpenAI's current terms of service to verify commercial rights, document when and how images were generated for records purposes, do not generate images depicting real people or actual client facilities, include a brief note in the document appendix that visual elements include AI-generated illustrations.

The appendix note reads: "Select visual illustrations in this report were generated with AI image generation tools. All content and analysis are the work of [firm name] consultants."

She found that clients, when the note was there, either ignored it or asked curious questions about the workflow. No client has objected.

The Workflow As It Evolved

After six months of regular use, Elena's workflow:

Planning (2-3 minutes per image): Before generating, she writes a one-sentence description of what the image needs to communicate. This clarity work prevents wasted generation rounds.

Initial generation (5-10 minutes): First generation plus 1-2 refinements in ChatGPT. She has found three iterations is usually sufficient; beyond that, she is chasing diminishing returns.

Color and format adjustment (5 minutes): Download, crop to exact PowerPoint dimensions (13.33 x 7.5 inches at 300 DPI for print quality), and do any color adjustment in PowerPoint or Photoshop to better match the deck palette.

Judgment check: Does this image communicate the concept it is supposed to communicate? Would a viewer seeing this slide understand the metaphor? If not, rethink the concept, not just the prompt.

Total per-image investment: 12-20 minutes for a custom conceptual illustration that previously took either hours of stock photo search (settling for something mediocre) or a design commission (expensive and slow).

What the Upgrade Actually Changed

Elena was initially modest about claiming productivity improvements — she is analytically rigorous and skeptical of her own positive impressions. So she tracked it.

Over a sample of twelve client presentations, she compared the time spent on visual selection and creation: - Previous workflow (stock photos): average 45 minutes per presentation for image research and selection - New workflow (AI generation for conceptual images, stock for others): average 65 minutes per presentation

Her overall investment in visual work increased, but: - Client feedback on presentation quality measurably improved — reviewers specifically mentioned visual coherence and custom feel - She stopped using the weakest-link stock photos (the ones she knew were mediocre but was willing to settle for because search time was finite) - Two clients asked whether her firm had design support — they assumed the visuals were professionally designed

The 20-minute additional investment produced presentations that genuinely looked different from standard consulting deliverables. In a business where the quality of presentation design directly communicates the quality of thinking, that was a real advantage.

She also found that the process of generating images — the conversation with ChatGPT about visual metaphors and concepts — sometimes clarified her own thinking about the underlying ideas. Articulating "what should this image communicate about the transformation journey?" turned out to be a useful forcing function for making sure the analytical framework was visually coherent.

That secondary benefit — using image generation as a conceptual thinking tool, not just an output tool — was the one she had not anticipated.