VIA Research
What users say about AI-native product development.
Voices from blogs, forums, studies and the press. Each quote is verbatim and links to its source. These are the people who build products with AI today, and they are not customers of VIA.
The bottleneck moved from building to deciding
Generation got cheap. Review, approval and alignment did not.
People on my team spend about 10% of their time writing code and 90% doing design and analysis.
The bottleneck in most teams isn't execution anymore. Reviews, approvals, and alignment decisions can't keep up with the flood of new work.
Last week I spent 3 hours reviewing code that took 20 minutes to write.
Speed without understanding costs trust
Teams ship what they cannot explain, and the asymmetry shows inside the team.
We find tensions between efficiency-driven prototyping ("intending the right design") and reflection ("designing the right intention"), introducing new asymmetries in trust, responsibility, and social stigma within teams.
We've replaced code we wrote and didn't fully understand with code AI wrote and we definitely don't understand.
I think senior engineers find it very difficult to give up on line-by-line review
The system forgets what it decided
Each session starts fresh. Decisions, structure and intent do not carry over.
AI is optimizing for local correctness, not systemic soundness.
Your agent doesn't remember what it decided yesterday. It doesn't know why it structured the auth module that way.
Discovery became continuous, judgment stayed with people
Prototypes and interviews run in hours. Knowing what matters still takes a person.
Discovery is continuous, not episodic.
A PM who relies entirely on AI-synthesized interviews misses the hesitations, emotional cues, and off-script comments that change a product direction.
This technique is now widely adopted by development teams, including product managers and designers.
The unit of work increasingly becomes an opportunity that is continuously evaluated, compared, and reprioritized.
Working software is not fit software
The code runs. The data is open, the rules are unmet, and nobody can say who directed what.
The most common compliance gap in enterprise AI deployments is that AI accesses regulated data under a service account or API key, and no log records which individual directed the access.
Unacceptable and should never be possible.
The canonical failure is an open database behind a client-side key.
AI coding tools do not understand regulatory context. They optimize for working code — not compliant code.
Nobody can show the return
Spend is measured. Value is not.
84% of finance leaders have not been able to measure the ROI of AI initiatives.
But those tools show only what a company spent, not business outcomes.
Every AI use case should have a clear purpose and ensure that purpose has ROI.
How we read these voices
Each quote names a problem that AI-native product development has to solve: who decides, who understands, what the system remembers, what fits for production, and what the work returns. VIA is built to answer them. See what VIA does.
We add voices as we find them and remove any the source no longer supports. Numbers are cited as their source states them and are not ours.