More All Things Product episodes

What AI Changes thumbnail

What AI Changes

Published 3 Nov 2026

Duration: 00:23:56

AI exposes existing organizational flaws rather than replacing essential processes like testing, planning, and customer discovery, while SaaS failures stem from poor strategy and excessive valuation rather than AI competition.

Episode Description

Everyone's shouting that AI changes everything. Teresa Torres wrote a LinkedIn "screed" arguing the opposite - and this episode is the debate she and...

Overview

The discussion opens with how AI-accelerated development is surfacing pre-existing organizational weaknesses. Teams report that "the cost of developing something went down so much" but "we still have this massive bottleneck called testing," with work "piling up at the end of our task force." The speakers question whether these organizations "understood what good looks like in the past already," suggesting AI is "exposing what was already broken" rather than creating new problems.

The conversation then examines whether frameworks like Scrum are obsolete. While the speakers agree that story-by-story estimation is "pure fiction from the beginning," they maintain that "you still will need to be asked to put things on time horizons." They note that agent coordination - using planning documents, loops, and graphs - mirrors the same rituals teams use to build software, meaning "the number of agents involved is going up and the throughput of work is going up."

Finally, the speakers push back on claims that customer discovery is unnecessary, arguing that "your customer is not brand new. They're still a human." They also challenge the narrative that AI is killing SaaS, pointing to Airtable's performance decline and Miro's "$17.3 billion valuation" against "$600 million a year in revenue" as evidence that "people at SAS companies making stupid decisions" and excessive venture capital, not AI, drive these outcomes.

What If

  • What if you audit your testing infrastructure before scaling AI-assisted code generation?
    • Move: Inventory your current automated test coverage and code review workflow, then use agents to write the missing automated tests before increasing AI-generated code output.
    • Why Now?: The source describes teams where "the cost of developing something went down so much" but work was "piling up at the end of our task force" because testing was already weak. The speakers note agents are "really good at writing automated testing."
    • Expected Upside: You could propose that fixing the testing gap first prevents the bottleneck from worsening as throughput increases, rather than discovering it after a pile-up has already formed.
  • What if you restructure your product into composable building blocks for customer self-customization?
    • Move: Identify the core data structures, workflows, and templates in your product and expose them as mix-and-match components, following the pattern the source attributes to Notion's "building blocks" architecture.
    • Why Now?: The speakers identify hyper-personalization as a genuine AI-driven shift, noting "SaaS providers figure out how to offer hyper personalization because AI makes it just not possible" to ignore. The source also argues the "one winner takes the market" VC thesis is wrong, opening space for targeted solutions.
    • Expected Upside: You might position your product as a customizable platform rather than a fixed solution, potentially reducing churn from customers who need tailored workflows and differentiating from competitors offering one-size-fits-all tools.
  • What if you replace granular estimation with weekly time-horizon alignment in your workflow?
    • Move: Drop story-by-story point estimates and instead define a weekly work chunk with a clear completion goal, as the source describes: "this week we're working on this chunk of work and our goal is to complete it."
    • Why Now?: The source argues estimation story-by-story is "pure fiction from the beginning" while alignment on time horizons "is always needed." AI accelerates development, making rigid per-task estimates even less useful as a planning tool.
    • Expected Upside: You could reduce planning overhead while maintaining the coordination the source says persists, freeing time for the actual building and review work that the source identifies as the real bottleneck.

Takeaway

  • Audit your automated testing and code review pipeline before increasing AI-generated code volume; the source shows teams where faster development created untested code pile-ups because testing was already weak.
  • Maintain planning, alignment, and review rituals even as you adopt AI tools; the source argues "you still will need to be asked to put things on time horizons" regardless of framework.
  • Continue customer discovery using story-based interviewing focused on past experiences; the source clarifies "your customer is not brand new. They're still a human. They still have past lived experiences."
  • Design your product architecture around composable building blocks to enable hyper-personalization, following the pattern the source attributes to Notion.
  • Replace story-by-story estimation with weekly time-horizon alignment; the source calls per-task estimation "pure fiction" while affirming that aligning on "this week we're working on this chunk of work" remains necessary.

Recent Episodes of All Things Product

17 Nov 2026 The Sharing Gap

Product leaders are encouraged to share their real experiences and specific examples on social media to foster genuine learning and community growth.

10 Nov 2026 Discovery & The 4 Risks

Teresa Torres questions whether continuous discovery has product-market fit for large organizations due to systemic obstacles.

27 Oct 2026 Defining What Good Looks Like

"Explores AI product development challenges, emphasizing measurable outcomes, customer-centric evaluation, cross-functional collaboration, and real-world impact over technical metrics."

13 Oct 2026 Training Expectations

"Continuous product discovery training requires habit-building and practice, not just short sessions, as teams often underestimate the depth of skill development needed for mastery."

6 Oct 2026 Recorded Conversations

"AI and technology's role in recording conversations raises ethical, psychological, and social challenges, including trust, evolving opinions, memory inaccuracies, and the impact of transcripts on relationships and accountability."

More All Things Product episodes