A client sends over a 40-slide deck. It was built with an AI tool in eleven minutes, and it looks, at a glance, fine. Then you open it properly, and every chart title is centered slightly differently, the brand blue shifts three shades across four slides, and slide 22 tells a completely different story than slide 4 does. Nobody asked for that — the tool just didn't know it mattered. That gap between "looks fine" and "actually works in front of a client" is where this whole question lives.

So will presentation design still be a real job in five years, or are we watching the last few years of it happen right now? The honest answer isn't "yes" or "no." It's that the job is already splitting into two very different jobs, and only one of them survives.

What the machines are actually good at (and it's more than people admit)

I'll say the unpopular part first: a large chunk of what presentation designers have billed hours for over the past decade was never really design work. It was formatting labor. A GfK-commissioned study on corporate PowerPoint use found that roughly 37% of the total time employees spend working in PowerPoint goes to formatting — nudging fonts into alignment, fixing color drift between slides, rebuilding tables that broke when someone pasted new data in. That's about 2.6 hours a week, per employee, spent on work that has almost nothing to do with communication and everything to do with software friction.

That's exactly the layer AI tools have gotten genuinely good at. Consistent margins, matched color palettes, auto-generated bullet hierarchies, first-pass layout suggestions — a machine doesn't get tired doing that, and it doesn't skip step 14 because it's 6pm on a Friday. If your value as a designer was "I make the fonts match," that value is mostly gone already, and no amount of hoping changes that.

The 60% wall — and why it's not where you'd expect

Here's where it gets more interesting than the usual "AI can't replace human creativity" line, which is true but says almost nothing useful. Figma's State of the Designer 2026 report, based on a survey of 906 designers across five regions, found that generative AI tools reliably get a project to around 60% complete, fast — options generated in seconds, a working structure in minutes, a blank-page problem solved before you've finished your coffee.

What the report also found is that the remaining 40% doesn't shrink no matter how good the model gets. It's not extra polish tacked onto the end — it's a different category of decision entirely: which data point actually changes the client's mind, whether this audience trusts a bold claim or needs it hedged, why the deck that worked for the board won't work for the sales floor. A model trained on millions of generic slide decks has no way to know that this specific VP hates pie charts, or that this client's last presenter got heckled for over-promising numbers. That's not a training-data gap that scales away with a bigger model. It's missing context that only exists in a human relationship with the work.

The part of the job that's growing, not shrinking

One thing I keep noticing in agency conversations: the designers getting more work in 2026, not less, aren't the ones ignoring AI tools out of principle. They're the ones running two or three of them in parallel, throwing out 80% of what comes back, and being ruthless about the 20% that's actually usable. That's a different skill than "I can use Illustrator." It's closer to editorial judgment — knowing what to keep, what to kill, and why a technically correct layout can still be the wrong one for this room.

This shows up most clearly in custom, brand-specific design work — the kind built around a company's actual voice, colors, and story rather than a generic template. That work resists automation for a boring, structural reason: an AI tool has no memory of your last three pitches, no sense of what made the CFO wince last quarter, no read on the room. A designer who's sat through those meetings does.

Where the job genuinely does shrink

I wouldn't sugarcoat the other half of this. Commodity deck production — the kind where someone needs a competent 20-slide internal update by Thursday and doesn't much care whose hand touched it — is exactly the work AI tools now do adequately, and cheaply. If that was the bulk of your workload, the next five years are going to feel like a contraction, not a shift. This is also the segment where pre-built templates already did most of the disruption before AI showed up; AI just finished the job templates started, by removing even the "pick and customize" step.

The trap here is assuming the whole profession moves the same direction as its most commoditized segment. It doesn't. Law didn't disappear when contract-review software got good — it concentrated around the disputes software couldn't resolve. Presentation design looks to be heading the same way: fewer people doing routine formatting, a smaller but sturdier group doing the judgment-heavy work that routine formatting was never really about.

What actually changes for someone doing this work today

If you're weighing whether to keep building a career here, the practical shift isn't "learn to prompt better," though that helps. It's learning to treat AI output the way an editor treats a first draft — fast to generate, never trusted at face value.

  • Get faster at spotting where a generated layout is wrong for the specific audience, not just wrong technically.
  • Build actual subject-matter familiarity in a few industries — finance, healthcare, SaaS — so you can catch claims that sound fine but are wrong for that audience.
  • Stop competing on turnaround speed for generic decks; that race is already lost to software.
  • Get comfortable pushing back on a client's first instinct, since that's precisely the judgment call no tool will make for them.

None of that is comfortable advice, because it means the easy, high-volume version of this job really is going away. But browse a catalog of ready-made templates for five minutes and you'll notice something: even a well-built template still needs someone to decide which one fits, what to cut, and what the deck is actually trying to argue. That decision hasn't been automated. It's just gotten harder to fake.

So, back to that eleven-minute deck from the opening — the one with the drifting blues and the contradictory story on slide 22. Nobody caught it before it went to the client. That's the job now: not building the deck from nothing, but being the person in the room who would have caught it.