There are many AI instruments to assist construct displays or slide decks. We’ve experimented with lots of them and located some are fairly good, others are generic or fairly terrible.
Often, we come again to doing a lot of the work manually.
But once we not too long ago used OpenAI’s GPT-6 Astra (and its pc use capabilities) to construct a presentation deck instantly inside Keynote, the outcome was fairly unbelievable. (Computer use means the mannequin can work inside an app, click on by means of it, and take actions in your pc.)
AI bought roughly 80% of the best way there on the primary construct inside Keynote. It dealt with sizing, alignment, formatting, points that it didn’t at all times get proper in earlier AI experiments or took hours of human time to do.
What remained, the 20% wanted to complete the undertaking, required inventive judgment from a human: the slides that required a particular contact or the narrative that wanted skilled enhancing.
It’s vital to recollect: AI accelerates the construct. Humans elevate the outcome.
The objective is spend extra time on the pondering behind the presentation and the script, and fewer time on formatting and nitty-gritty fixes on particular person slides. Tools corresponding to Astra are getting us nearer to that.
A Step-by-Step Approach
For our use case experiment, we began with a full define, a bunch of analysis, and a few preliminary narratives and scripts that have been all human-created and illustrated the general content material and route. Then we adopted these steps:
Step 1: Define your aesthetic first
A transparent aesthetic is required to keep away from a generic-looking presentation. We gave it steerage on logos, branding, and examples of most well-liked slides. It took loads of backwards and forwards to get the look proper. Spending this a lot time, each time, doesn’t make sense, so it is vital to create a reusable ability or template when you get the aesthetic proper.
We suggest establishing a reusable immediate, doc or ability to make use of subsequent time.
Step 2: Give AI construction, not a clean slate
AI performs dramatically higher once you carry organized materials to the session, not a imprecise transient. For a presentation, which means having your define, key factors or full script mapped out earlier than you ask AI to construct something. We used a script.
Step 3: Let AI deal with the primary construct
With your aesthetic outlined and your content material structured, watch what Astra does, the place you possibly can actually observe it working inside Keynote. Note what it bought proper and what it missed. It’s fascinating to look at.
Step 4: Apply your judgment to the ultimate 20%
This is the place the human contact is available in. Review the draft towards your model requirements, your viewers’s expectations, and your personal inventive instincts. Identify what wants a human contact and spend your time there.
Step 5: Capture what you be taught for subsequent time.
The actual long-term worth of this course of is not the asset you simply constructed however what you realized about working with AI to construct it. Pay consideration to what made the most important distinction. Over time, this turns into an institutional functionality. Teams that doc their AI workflows get dramatically higher outcomes than groups that begin from scratch each time.
You can even apply your studying to different apps that Astra can work inside.
One Big Caveat
Employing computer-use brokers corresponding to GPT-6 Astra to click on by means of desktop apps — on this case, Keynote — burns by means of loads of tokens, making it doubtlessly costly. Be conscious that utilizing these capabilities can result in further prices, so it may not be proper to make use of this usually within the present token-pricing setting. But it’s a good experiment to run and perceive AI’s advancing capabilities, in addition to reveal what’s doable inside your different apps.
This publish attracts on the AI Use Case Spotlight section of Episode 239 of The Artificial Intelligence Show, hosted by Paul Roetzer and Mike Kaput. To hearken to the complete Episode 239 of The Artificial Intelligence podcast, go to: https://podcast.smarterx.ai/shownotes/239
