In April 2026, we hosted the first in what we hope becomes a long series of community roundtables. The format was simple: bring together eleven practitioners across engineering, marketing, and research disciplines. Pose a single question. Let the conversation go where it goes.
The question was the one I'd been sitting with for months:
Is AI actually making engineering teams faster — or just making them feel faster?
The setup wasn't theoretical. AI can reduce fix times from 7.5 hours to 23 minutes. It can compress a week of work into a day. The individual-level productivity gains are not in question. But when we zoom out from individual tasks to how teams actually operate — to projects, to deliverables, to the system as a whole — does the speed compound, or does something else happen?
Each participant shared how they use AI, where it accelerates them, and where it frustrates them. What emerged was more interesting than any single answer.
---
Five Themes
1. The Productivity Is Real — At the Individual Level
Nobody in the room doubted that AI speeds up individual work. The examples were concrete and varied:
- Parsing complex specifications and generating documentation. A week's work compressed to a day. - Debugging and code completion at iteration speeds nobody had a year ago. - Document processing, summarization, and proposal preparation. What used to take a small team now takes one person with a good prompt. - Hardware and data analysis, presentation generation. - Hiring committee write-ups — more interviews per week with AI handling the synthesis.
The speed gains are not theoretical. Everyone in the room is measurably faster at specific tasks. That part of the conversation closed quickly.
It was what came next that made the room quiet down.
2. The Inflation of Value
One participant — a researcher who's spent years thinking about creative output — made an observation that reframed the entire conversation:
"The easier something gets to produce, the less valuable it is."
If AI makes PowerPoint presentations trivial, presentations stop being a differentiator. If long research reports take minutes instead of weeks, the report itself is no longer where the value lives. If code is easy to produce, code alone doesn't distinguish you.
This is inflation. The output that used to signal effort and expertise now signals access to a tool. Which means the question shifts: where are you adding value that isn't easy to replicate?
The room sat with that for a moment. The implication is uncomfortable. If AI flattens the production curve across an industry, the work people built careers around may stop carrying the weight it once did. The answer the room kept circling back to: creativity and constant reinvention. As one participant put it — "if what you're doing has become easy to replicate, it's a matter of time before it goes."
3. Architecture First, AI Second
A senior architect at a major tech company described himself as a heavy AI user. Reverse engineering, parsing specifications, architectural design across systems with billions of files. His conclusion was precise:
"Once I have the right idea or the right architecture, AI saves me a lot of time. Without it, it's naive and hallucinates."
The pattern: AI is a powerful accelerator after the human provides structure. Before that — before you know what you're building and why — it produces confident nonsense. The human's job is the architecture. AI's job is execution within that architecture.
A task that took a week now takes a day. But only because the week of thinking happened first.
I haven't stopped thinking about this one. Every time I see someone disappointed by an AI tool, I now wonder whether the disappointment was about the tool — or about the absence of the architecture the tool needed to be useful.
4. The Soul Question
A marketer in the group brought a different perspective. In their world, two modes of AI use exist side by side.
The logical side — analytics, campaign optimization, data interpretation — has been largely automated. Generalists can now do work that used to require specialists.
The creative side — video, writing, presentations with genuine impact — is where AI hits a wall. Not technically, but structurally:
"AI cannot replicate the human soul. Your creative energy, your vision — it is inherently different than anything AI can create."
Their conclusion: AI-created content isn't necessarily bad. But it's "a completely different category." The audience knows the difference even when they can't articulate it. A commercial made with AI doesn't sit in the same category as something filmed with human vision and intent.
Does AI make us faster at creative work? Their honest answer: "I don't know. Does it help us make things we couldn't have made before? Sure. Is it the same thing? I think it's just a different category entirely."
5. AI Works When You're Developing Your Own Thinking
Toward the end of the conversation, one participant zoomed out and named a pattern running through everyone's stories:
AI is a good use case when you're developing your own thinking — when you're shaping it, training it, evaluating output against your expertise. It works well in that mode.
It fails when you're not the domain expert. When you can't evaluate what AI produces. When you're asking it to do something you don't deeply understand yourself.
This was the throughline of the entire roundtable. Every success story in the room involved someone who knew what good looked like using AI to get there faster. Every frustration involved someone working outside their expertise and getting confident-sounding output they couldn't trust.
It's a clean diagnostic. If AI is working for you, you probably know your domain. If it's not working for you, you're probably trying to skip the part of the work where you build judgment.
---
What Else Came Up
A handful of other observations didn't fit the five-theme structure but deserve mention:
Multi-agent infrastructure. One participant described running infrastructure teams of AI agents — multiple agents working together, with humans reviewing code, testing, and security. The orchestration between agents (not just single-agent tasks) is where they see the real future.
The personal tutor model. Several people described using AI as a private tutor for unfamiliar territory — "level zero, it works. When it gets really complicated, it misunderstands."
The voice problem. One creative writer noted: "It's very important that I don't give it my voice." Their fear: as AI absorbs more of someone's writing, it starts producing things in their voice that they didn't write. The protective move is to keep the personal voice out of the training set.
The decisions stay human. A practical voice in the room: "AI gets you the data. It organizes the data. The decisions — humans make the decisions."
---
The Memory Thread
Toward the end, the conversation turned to something I've been thinking about for a long time: memory. When you work with an AI teammate over time, the importance of memory becomes obvious — the AI needs to inherit your style, remember lessons learned, maintain character across sessions.
You can teach AI your own version of creativity. If you're going to write something, you can teach it how you write. Over time, the AI doesn't just help you faster — it helps you in a way that's recognizably yours.
The cognitive partner model — where AI records, summarizes, and remembers on your behalf — saves significant time. But it's an acquired taste. You have to use it yourself to understand what it gives you.
---
The Unanswered Question
We started with: is AI making teams faster, or just individuals?
After ninety minutes of stories from eleven practitioners across six companies, the honest answer is: we know it makes individuals faster. We don't know if it makes teams faster.
The gap between individual acceleration and team-level coherence went unnamed in the discussion — but every frustration story pointed at it. Specs that drift faster because everyone's writing them faster. Code generated faster but reviewed slower. Decisions made faster, but propagated through the same old channels.
The system between people didn't change just because the individuals got faster. And in some cases, the system is now the bottleneck more than ever.
That's the question we plan to explore in the next roundtable.
---
What's Next
This was the first community roundtable. The format worked: each person sharing their actual experience, then the group finding patterns. No slides. No presentations. Just practitioners comparing notes.
If you'd like to join the next one — or propose a topic — join the mailing list. The roundtables are invite-only and small by design, but the topics are shaped by what the community is actually working on.
Thanks to everyone who showed up.
---
— Khaled Alashmouny Founder, AIDAChip