AI Memory Just Got an Upgrade: How Projects + Custom GPTs Build Smarter Workflows
One of the biggest bottlenecks in AI adoption has been its the context gap.
What does that mean? Every conversation you have with ChatGPT starts with amnesia. You might experience some magic after a single prompt. This is what we call a "one-shot prompt."
But that's not how it usually goes. Usually, there's quite a bit of back-and-forth until the model gets what you're after, and things start clicking. And then you start a new chat.
You lose momentum, nuance, and history; all the cognitive volleying and context that make your thinking yours.
That’s what “Projects” quietly fixed.
If you haven’t tried them yet, think of Projects as ChatGPT with memory, collaboration, and shared workspaces built in. They connect the dots between threads, files, and team members so your AI doesn’t just respond once, but remembers the next time you come knocking of answers.
For leaders trying to operationalize AI inside an organization, that shift is massive.
You’re no longer experimenting with a chatbot. You’re building a living, breathing knowledge engine for your team.
And the real bonus? Projects and Custom GPTs now play together beautifully.-
Meet Your New AI Workflow
At our agency, I lean on both Projects and Custom GPTs more than any other AI tool in my quiver.
When I’m building or writing, I treat Custom GPTs like instruments. They’re trained for specific tones and tasks. One helps with narrative structure, another handles data synthesis, a third assists with UX writing and strategy notes. Here's another frame. If I've asked AI more than three times to help me with a specific workflow, I then turn that into a custom GPT so that I can reuse those instructions and save a lot of prompting keystrokes.
But Projects are the studio.
They hold the work, remember the context, and let my team jump in without losing continuity.
Before Projects, every idea lived in isolation. I’d have ten conversations with ten versions of ChatGPT, each one starting at zero. That might work for creative bursts, but it’s terrible for building long-term systems.
Now, I can start a Project for something like:
Client Research Ops: a shared workspace where the AI remembers our data sources, frameworks, and naming conventions
Framework Friday: the backbone of this newsletter, with context spanning past issues, reader insights, and draft ideas
Internal SOP Builder: where we’re documenting and evolving how the agency actually runs
Each one builds its own memory over time. Each can reference uploaded materials and summarize context across threads.
When I drop into a Project, it feels like coming back to a colleague who remembers exactly where we left off.
That’s the shift.
AI isn’t a tool you use. It’s a collaborator you develop with.
Stop Asking Which One’s Better
If you’re just starting to integrate AI into your workflow or scaling it across a team, here’s a simple model I use to decide which to use.
Use a Project when:
1. You’re working across multiple sessions or contributors
2. You need collaboration
3. You’re building knowledge over time
Use a Custom GPT when:
1. You want repeatable, standardized tasks
2. You’re scaling access beyond the project team
3. You want external shareability
In short:
Projects = memory, collaboration, evolution
Custom GPTs = specialization, automation, distribution
They’re actually complementary, not competing.
A Custom GPT runs inside a Project for specialized work, while the Project itself holds the broader memory and context.
Think of it this way:
You might have a Project called “Marketing Strategy.” Inside that Project, you’ve uploaded past campaigns, messaging frameworks, and audience personas.
Now you can bring in a Custom GPT (just @mention its name in the prompt), maybe, your “Brand Voice Coach," to rewrite new campaign copy directly inside that Project. Because the Project already contains all your background materials, the GPT instantly has context on tone, goals, and audience.
You can then switch to another Custom GPT, maybe your “Analytics Synthesizer," to summarize campaign performance in the same workspace, using the same shared memory.
The Project connects it all together.
Each GPT brings its own skillset, but they’re all working from the same page.
That’s the future of AI workflows: memory (Projects) meets specialization (Custom GPTs).
How Teams Are Winning With Both
Here are a few ways I’ve seen leaders and operators use this duo effectively:
Build Persistent Knowledge Loops
Create one Project per domain such as Revenue Ops, HR, CX, Product, etc. and upload your documentation, templates, and notes.
Now when anyone asks the AI a question, it answers from your actual materials, not random internet data.
Share Projects With Teammates
Stop sending prompt screenshots. Give your team shared access to the Project itself. They’ll see the context, history, and relevant files, which keeps outputs aligned and predictable.
Train Specialized GPTs for Each Function
Use Custom GPTs as micro-experts within Projects. For example:
A Tone Coach GPT inside your brand voice Project
A Data Synthesizer GPT inside your research Project
A Recruiting Writer GPT inside your HR automation Project
Build a Company Memory, Not Just a Chat Log
When you rely on threads, you’re collecting ideas. When you rely on Projects, you’re compounding knowledge. That’s a structural advantage most teams haven’t even realized they can build yet.
We’ve reached a turning point now that memory is getting better in length and cross-referencing.
For years, AI tools have been individual productivity boosters: fast, clever, but siloed.
Projects shift that paradigm. They turn AI from a set of disconnected assistants into a coordinated team.
And once you layer in Custom GPTs, you’ve got something close to an internal AI ecosystem.
Your company’s knowledge becomes searchable, expandable, and self-reinforcing.
This is the kind of leverage that can compound through an entire organization.
If you’re serious about using AI strategically, whether you’re in HR, RevOps, or CX, start by naming your first Project today.
Give it a purpose, invite a teammate, and let it start remembering.
The future of AI work won’t belong to people who ask the best prompts.
It will belong to people who build the best memories.
Find your next edge,
Eli
Want help applying this to your product or strategy? We’re ready when you are → Let's get started.
