A New, Better Way to Build
AI should raise the quality bar for software teams, not lower it.
Matt Teixeira
4 minutes read
"How will roles change because of LLMs and AI agents?"
Everyone is asking this.
Fair enough. They already have changed - at least in tech. Most of us use AI every day now. To write. Design. Code. Think. Summarize. Ship.
But a strange belief has become popular on X, LinkedIn, and in almost every AI podcast I listen to:
Because AI makes teams faster, shipping unfinished work is now acceptable. Moving fast at the expense of quality seems to be the right approach to take.
I disagree.
Not with speed - speed is one of Deck's three operating pillars. Speed matters.
But speed is not permission to lower the bar.
I said this in a talk a few months ago, but I would much rather ship something twice as fast that’s polished and solves a real problem than ship an average experience 5x faster.
AI should raise the quality bar. Not lower it.
I wrote down some ideas that I operate with in mind when I think about Deck.
These ideas have one thing in common: building better software faster, with a tool-centric view of how product teams should work - and how we’re enabling that at Deck.
One Product Operating System
The core bet behind Deck.
I don't think product teams should need ten different tools to close the Build-Measure-Learn loop.
They should be able to do it in one place. Well.
From consolidated customer feedback, to prioritizing that feedback, to connecting it with company strategy, to creating initiatives and projects, to linking everything with analytics.
That should all live in one organized system of engagement.
Because scattered context makes teams slower and worse.
If you care about quality and speed, you need a system that helps you understand what matters, decide what to build, execute on it, and measure whether it worked.
Fast, Actionable Feedback
Customer feedback has, for so long, has been put aside.
Not because teams don't care.
Mostly because it is painful.
It’s hard to collect. Hard to synthesize. Hard to organize. Hard to connect to strategy. Hard to turn into something people can actually act on.
So it just sits somewhere.
In a spreadsheet. In a Slack channel. In a Notion doc. In an interview recording no one will open again.
Your customer told you what mattered, but the system around the feedback made it too expensive to use.
At Deck, we believe the opposite should be true.
It should be easy to synthesize customer feedback. Easy to organize it. Easy to understand which patterns matter. Easy to decide what to do next.
The point is not to collect more feedback.
The point is to act on the right feedback faster.
The Graveyard of Documents
The graveyard of documents is the name I give for knowledge management tools like Confluence and Notion.
They're needed and often useful, but the reality is that 95% of the documents written by me, by you, and by our coworkers are never read, shared, or really acted on by any of us.
We write documents, share them, and we're lucky if our coworkers read even 20% of it. The likelihood is that one week after the document is created, it won't be accessed ever again.
You know what I'm talking about.
Therefore, we will build Deck with the intention of influencing our customers to stop contributing to the graveyard of documents and only really write when you need to.
Less, but better.
Closing the Feedback Loop
Let's go through how 99% of companies operate:
-
An executive or a middle manager has an idea about something to be built.
-
They ask the product team to build it.
-
The product team goes ahead and builds it.
-
The solution is released.
And that’s that.
No one really monitors whether the release has actually contributed to something. No one asks their customers whether they saw a change or that the product became more valuable as a result of this change.
All that really happens is a tap on the back for having released, another idea, and another solution to be executed on Monday morning.
It simply cannot be that way.
The whole point of being a product operating system is to make that loop visible and closable.
A customer says something. A pattern emerges. The team prioritizes it. The team connects it to strategy. The team identifies possible solutions. The team builds one. The team ships it. Then the team checks whether it worked.
That is the loop.
Many claim to do it, but few actually do it.
We want to enable that to be the norm.
Giving Space for Creativity and Business Thinking
Many have spoken on which skills really matter now that AI is eating the world.
It’s hard to really pinpoint what the answer(s) is (are), but I can guarantee that creativity and business thinking will become much more important.
LLMs are trained in a specific set of data. Being creative enables you to ship work that won’t ever be identified by the LLM - because it doesn’t know how to produce and go through little iterations as well as a human can. It doesn’t operate as well with agency and self-iteration loops.
For a similar reason, business thinking will prevail: commercial smarts matter, but nothing like a business plan failing when facing the real world. LLMs don’t face the real world well.
That means at Deck, we’ll work to take care of the rest, while you take care of these 2.