GRAIL · Function papers
What does an AI-augmented product organisation look like?
By Johan Grönstedt · Last reviewed
Evidence stays live. Judgment stays with the person. This free nine-page position paper sets out the operating rhythm, the division of work and the beliefs behind it.
An AI-augmented product organisation keeps customer evidence, product behavior and commercial context as a continuous operating layer around every product decision. The roadmap becomes a living record of bets, evidence and choices, while judgment and accountability stay with the person.
The operating rhythm
A day in The product function's 2028 rhythm
Product teams are surrounded by evidence but still make decisions from periodic summaries, partial memories and whoever argues best. When customer evidence, product behavior and commercial context become continuous, the roadmap stops being a presentation of intentions. It becomes a living record of bets, evidence and choices.
The rhythm begins with approved evidence, brings the relevant context together, and gives product managers the contradictions worth investigating. Drafts follow those choices, while each decision carries its assumptions and the evidence that could reverse it.
LISTEN
Approved interviews, support conversations, reviews and sales calls are processed.
INTERPRET
Relevant research, usage, commercial exposure and support history come together.
DECIDE
Product managers investigate contradictions and choose customers to call.
DELIVER
Draft specifications, internal notes and feedback tags are proposed.
LEARN
Decisions record assumptions and the evidence that would reverse them.
The division of work
What runs, and what stays with the person
The agent layer prepares the material around a decision. It can process approved sources, connect findings to source material, monitor public and internal evidence, assemble alternatives, draft requirements and check publication material against release status. It also answers governed questions, flags anomalies and proposes explanations with the underlying data.
The person controls the judgment that follows. Researchers control sampling and interpretation. Product managers conduct important conversations, investigate outliers, own recommendations and decide what action follows. Product leadership judges strategic significance and owns capital allocation.
Review and approval remain explicit. Product, design, legal and engineering review scope, tradeoffs and readiness. Analysts validate taxonomy, causal claims and experiment design. Owners approve positioning, claims, publication and customer commitments. The pattern is consistent: evidence and drafts stay current around the choice, while the person remains accountable for the choice itself.
| Process | What the agent layer does | What stays with the person |
|---|---|---|
| Discovery and customer research | Prepares interview guides from existing evidence, transcribes approved sessions, proposes themes and links each finding to source material | Researchers control sampling and interpretation; product managers conduct important conversations and investigate outliers |
| Roadmap and prioritisation | Assembles alternative rankings, missing evidence, dependencies, strategic fit and rejected assumptions | The product manager owns the recommendation; product leadership owns capital allocation |
| Requirements and product specifications | Drafts problem statements, product requirements, stories, edge cases and open questions from approved discovery material | Product, design, legal and engineering review the scope, tradeoffs and readiness of the result |
| Launch, documentation and enablement | Produces audience-specific drafts from the approved specification and decision record, then checks them against release status | Owners approve positioning, claims, readiness, publication and customer commitments |
| Product analytics and feedback loops | Answers governed questions, creates analyses, flags anomalies and proposes explanations with underlying data | Analysts validate taxonomy, causal claims and experiment design; product managers decide what action follows |
The position
Six things GRAIL believes
These beliefs concern how a company learns, chooses and commits. Every connected system inherits them, regardless of the product tool in use.
The product memory matters more than the first draft.
A fast specification is useful, but it is not the structural change.
Continuous discovery is a coverage problem before it is a synthesis problem.
Combining interviews, tickets, sales calls, reviews and behavior can reveal patterns no single channel holds.
The roadmap is a capital-allocation record, not a list of promises.
Each bet should carry its evidence, commercial exposure, delivery demand and stopping assumptions.
Documents should follow decisions, not become substitutes for them.
Product requirements, release notes and portfolio packs can be prepared from approved evidence and kept current across systems.
Augment the interpretation. Keep the accountability human.
Agents can retrieve evidence, compare histories, expose contradictions and prepare alternatives.
Connect customer evidence before delivery activity defines the truth.
A delivery system is attractive because its fields are structured and its connection is straightforward.
Get the paper
Read the full position on how evidence, decisions and human accountability fit together in the product organisation.