A Guide to Figure Out Where Ai is Helpful in Our Work? - AIxC: 76
+ Open Source Explorations for Paid Subscribers
Hola and bienvenidos to #76! Starting off with a poll today as I’m figuring out how to best add value to everyone here.
Today, going to try and keep it short and focused on an important question - where is Ai helpful in our work?
I’ll work to keep it brief (ok just finished it and didn’t keep it brief),
Hector & AIxC team
Where is Ai Helpful in Our Work?
This is a question that I get often. It’s also something that is alluded to in online comments on a regular basis. It usually takes the form of comments like this:
Ai is constantly changing my intent. It’s useless.
Ai is like a slot machine, you never know what you’re going to get. You can’t depend on it.
Ai is only good for up front ideation.
The reality is that understanding how helpful Ai can be for our work requires a structured exercise.
STEP 01
Map Out The Current Process
Let’s think about ROI with Ai efforts. Your investment is going to be time. Your return goals are what needs to be better defined.
I first map out the existing process.
Skills/Tools
What are the skills/tools at play and what do they accomplish? How do they marry up against my intent and how predictable are the results, ie a high fidelity cad rendering is close to a 100% intent-to-output, highly predictable results.
Tasks
Next, what are the tasks in my process? You can focus on just the design/creative tasks, but you can also expand beyond those to all facets of your day-to-day work.
Really think about how your day-to-day works. Think about the intent-to-output requirements - a predictability ratio of sorts.
This is the baseline. We need this to better understand if Ai provides a beneficial ROI.
STEP 02
Test Ai
Next you need to gain an understanding of where Ai is today and how it tracks against your intent-to-output requirements / predictability ratio.
The good news is that we already know some of the answers.
One, there is no 100% high intent / high predictability solution with Ai. If you want that, just CAD up a model.
Two, we know there are Ai solutions that get you 10%, 30%, 60%, etc.
Not so good news, getting a handle on this requires a serious investment of resources. There’s really no way around this.
You either invest your time or you leverage the experience others have, folks like us here. Regardless, it’s an effort.
It’s an effort in understanding how close to our intent/requirements we can get with Ai.
What do the different tools accomplish?
What do different inputs accomplish?
Different workflows?
Different strategies?
Eventually you’ll start understanding that text-to-image with short prompts gives you low predictability. Sketch-to-render a bit higher. A mix of sketch-to-render to retexture costs you some percentage points. Etc. Etc.
I think of it like preparing to build a puzzle by breaking up the puzzle pieces by shape.
STEP 03
Map Ai to Existing Workflows
Once you have your puzzle pieces laid out, you can start putting the puzzle together.
You’ll identify opportunities in your existing processes where you may only need 20% predictability. And you’ll know which combination of Ai tools and workflows meet that requirement.
You map Ai strategies to matching task opportunities.
You start seeing that Ai is not aligned with high fidelity product visuals of existing products for your website, but it is well aligned with photography guidance for your next product photoshoot.
The puzzle starts coming together.
Again, it requires effort, but when expanded to your full day-to-day workload, creative and admin, you’ll find a lot of opportunities to get some gains.
If you’re part of an organization, you multiply those gains.
Incremental gains that will add up.
04 / FOR PAID SUBSCRIBERS
Relighting Efforts with Open Source Tools
For paid subscribers we have new content in the lab focused on some relighting efforts we did with open source tools as well as more SREF code sheets.
Sign up below for access to the AIxC R&D Lab board and the detailed breakdown. Link and password behind the paywall.