# What will you build next?

Last updated: 2026-10-14
Canonical source: https://confidence.spotify.com/blog/what-will-you-build-next
Owner: Spotify AB
Authors: Johan Rydberg (Product Director)

> If this file and the page at the canonical URL disagree, the page is authoritative.

As AI makes building cheap, the competitive advantage shifts to the quality of product decisions. Confidence is expanding into a system for the full product loop.

The answer to that question used to come down to team size, roadmap, and engineering capacity. Good ideas were shelved all the time, simply because there wasn't enough time to build them.

Execution was the bottleneck that decided what had a real chance to matter. That has changed.

## Context to match speed

AI has fundamentally changed how products get built. You can draft a PRD in minutes, ship a prototype in an afternoon, and stand up an experiment with a fraction of the effort it once took. Building is no longer the hard part.

When building took months, judgment had time to catch up, and ideas were debated, stress-tested, and refined long before anything reached users. Now an idea can go from a conversation to a working prototype in an afternoon, and a slick demo can make it look ready to ship before anyone has asked whether it should. The decisions have to keep pace with the building, which means making them sooner and far more often.

As execution gets faster, the question shifts from *can we build this?* to *should we?*

Why do we think it will work? What do we already know that should change our minds? As AI takes on more of the building, the competitive advantage is the quality of decisions. Anyone can burn tokens generating ideas and prototypes. The hard part is sourcing the right signals, making the decisions that actually move the needle for users, and proving it, so the work is judged by outcomes, not by how much was built.

## Rigor at every stage

Picking the right things to build matters more than ever, so that's where people will spend most of their time. And even if building is no longer time-consuming, we still only want to ship the good stuff, so we need better tools to apply judgment.

Ideas need to be generated and refined based on prior learning, user behavior, and strategic context. They should be quickly shaped into prototypes and tested against reality, with the level of validation matched to what the idea actually needs, from lightweight user feedback for small changes to rigorous experimentation for the ones with greater stakes.

Once something ships, the rigor should continue. Changes that don't work need to be identified early and discarded or iterated on. Systems need to continuously observe user behavior to identify where they struggle, where they succeed, and what new opportunities emerge. AI increasingly assists at every stage of this process, shaping ideas, running experiments, and surfacing results.

We used to think of product development in phases: discovery and delivery. Now it runs as a loop. Prior learning systematically feeds into the next iteration. If delivery speed was once the advantage, what matters now is how effectively you move through the product loop: how well you refine, design, and discard ideas, and whether the learning from one cycle meaningfully shapes the next.

Going forward, product teams will spend most of their time inside this loop; continuously refining ideas, testing them, and acting on what they learn, rather than passing work between separate stages.

## The product builder emerges

As the product loop accelerates and AI increases individual impact, the boundaries between product management, design, analysis, research, and engineering become less rigid. In their place, a new default is emerging: the product builder.

A product builder operates across the full loop. They generate ideas, orchestrate AI agents to build prototypes and production code, interpret signals, and make decisions. One person can now explore more ideas, run more tests, and ship more changes than entire teams could before.

This creates a step change in capability, but also in complexity. There are more ideas to evaluate, more signals to interpret, more decisions to make. The surface area of the role expands, and so does the risk of getting things wrong. Moving faster also means there is a need to make faster decisions, drawing on more and more context.

## Products are becoming partners

Products are no longer tools that amplify your ability to get work done. A product has become a collaborator that helps you explore ideas, surfaces relevant context, and proactively points out opportunities.

For lone product builders, this shift is especially important because the volume of signals is too high to rely on manual exploration alone. You need systems that investigate on your behalf, while taking your direction and judgment into account at every step.

## A new system for the product building loop

Building has never been faster, but the context behind product decisions hasn't kept up. There are strong systems for execution, tools that help us ship, measure, and operate at scale, but no equivalent system for the full product loop.

The signals that should inform decisions are often scattered: product analytics in one place, experimentation in another, qualitative insights somewhere else. Support conversations, app reviews, learnings from session replays, and sales feedback rarely surface when an idea is forming, so much user frustration remains hidden in the data. And as the volume of signals keeps growing, starting with an idea and then hunting for supporting evidence doesn't scale.

Confidence was built to change that.

Confidence started as Spotify's internal experimentation engine, running the A/B tests behind our own product for over a decade. It's been available to product teams outside Spotify for a while now as an experimentation platform, giving them the evidence to prove whether an idea worked.

We've now begun a new journey that expands Confidence into a system that pulls in the full range of signals, support conversations, sales calls, session replays, and connects them to every stage of the product loop, with your judgment guiding it at every step.

It's also your sparring partner, supporting idea generation, giving feedback, and helping you stress-test your reasoning. You can connect it to Slack, where it builds context from the channels you add it to and answers questions as your work unfolds.

Confidence continuously grounds ideas in real user behavior, shapes them into testable hypotheses, and validates them with the level of rigor each question demands. Every turn of the loop generates new context that becomes shared memory, informing the next set of decisions.

To enable this, we are adding the following capabilities to Confidence:

* Product Analytics - [Read more](/blog/product-analytics)
* Session Replay - [Read more](/blog/session-recordings)
* SlackBot - [Read more](/blog/introducing-confidence-slack-bot)
* Confidence Cloud - [Read more](/blog/introducing-confidence-cloud)

On top of all these capabilities sits the [Confidence Loop](/blog/confidence-loop). The [Confidence agent](/blog/introducing-confidence-agent) continuously sifts through signals to make sense of them, identifying issues to fix and opportunities to pursue, and suggesting next steps for the product builder to act on.

For the most obvious issues, the Confidence Loop goes further, autonomously preparing the suggested fix and opening a PR for you to review. But nothing ships until you approve it. Your judgment stays in the loop at every step.

When signals turn into good decisions instead of guesswork, product development feels different. You can move fast without moving blindly. You can explore more ideas, but with better grounding. You spend less time on things that clearly don't matter, and more on things that might.

**You can build almost anything now. The real question is what's worth building and why. Confidence helps you find the answer and build the right thing for you.**