The 5 value-creation moves
Agentic AI is moving the dividing line of the product profession. Yesterday, a profile's value lay in its specialty: mastery of a trade, a tool, a stage in the cycle. Today, it lies in the ability to create value end to end. Producing is no longer what is scarce. Deciding what to produce, and guaranteeing it, is. We are moving from an era of specialists in a single trade to an era of value creators.
This shift shows up in five creation moves, five archetypes (as described by Boris Cherny, creator of Claude Code) that agent augmentation now puts within reach of a single profile:
Imagines new ideas and produces many of them, even if most will never see the light of day.
Quickly turns a prototype into a real product.
Cleans up the interface, simplifies the code and architecture, removes what is unnecessary, optimizes performance.
Continuously improves an existing product until it finds product-market fit.
Keeps a system secure, reliable, fast and performant as it grows.
These five moves are not five jobs: they describe the spectrum of value that an augmented profile can now cover, around two emerging roles.
The agentic breakthrough
An agent does not suggest, it executes: it analyzes data, writes specs, generates code, produces interfaces and stories, tests and deploys. In a few hours, where a team used to need two sprints. The constraint shifts: it is no longer the capacity to build that limits value creation, it is the capacity to make trade-offs.
The breakthrough happens when you move from a prompt to a harness of agents orchestrated by humans. It is a breakthrough in how people work, not just in execution speed. And because this setup generates quickly, it raises two immediate questions: how do you de-risk at the right pace, and how do you avoid generating for the sake of generating (the accelerated feature factory)?
The real risk is systemic. Speeding up delivery throughput without speeding up what comes upstream does not hold: after two or three teams, throughput falls back and scaling becomes pointless. If you do not look at how methods are disrupted across the whole chain, you make that disruption dependent on its weakest link. Reforming the overall flow is a prerequisite for any acceleration.
Who does what in the agentic era
AI does not erase professions, it automates their execution. Expertise remains; what changes is the time spent applying it.
| Function | What remains its job | What AI takes over |
|---|---|---|
| Business | Naming the real problem: why, what for, how much. Setting the OKRs. | Case handling, processing, routine reporting |
| Product | De-risking the opportunity and the solution, making it efficient for tech and design to execute | Producing stories, the backlog and specs |
| Tech | Architecture, technical choices, guaranteeing the development | Code generation, tests, documentation |
| Design | Designing experiences that delight users | Producing interfaces from a defined direction |
The key point: each area of expertise refocuses on its own judgment, while routine execution shifts to agents. The boundary between professions does not disappear: it moves from "who does what" to "who decides what, and when". That is the boundary the emerging roles are there to hold.
Framing excellence
The product-business interface shifts to co-building around a prototype. The core of the product role is no longer producing a backlog, it is aligning stakeholders continuously. A versioned lifecycle and a constantly enriched documentation corpus make it possible to sync almost in real time with each stakeholder: latency is fought and the back-and-forth caused by decisions made too late disappears. The business / design / tech-data triad is no longer endured in sequence; it plays out together, from the very start.
The cadence changes too. Inspired by Shape Up, framing, delivery and run cycles replace the classic sprint. Framing is no longer a two-week affair but takes about a month. That is where most of the risk is removed and where the hypotheses to test and prototype are formulated. The Product Builder's typical week runs on a dual or triple track: framing one topic, following the delivery of another, and reading their product's metrics every day.
PDLC / SDLC symmetry. This framing discipline needs tooling. The flow relies on GO / NO-GO gates and on symmetrical agent harnesses: a product harness (PDLC) mirroring the engineering harness (SDLC), with the Product Builder tooled on a par with the developer. Without this symmetry, the speed gained on the dev side is paid for on the product side: the same frictions as before come back, multiplied by ten or a hundred. Moving to spec-driven development (SDD) revives the demand for a flawless spec: no longer a document patched along the way, but the input that drives all agentic production.
The two emerging roles
Product Builder
They steer the value of a product in a fully agentic chain. They do not feed a backlog of user stories: they orchestrate a system of agents, a harness, that analyze, prototype, generate backlogs and specs, produce code and interfaces, document and test. They are its decision-maker and its guarantor.
What they do not delegate: vision, trade-offs, negotiating with stakeholders who disagree, interpreting an ambiguous market signal, accountability for decisions. Three moves become the core of the job: framing (stating the intent and the specifications), contextualizing (giving agents the right material), verifying (remaining the sole guarantor, able to reject non-compliant output: the move that can least be delegated). Their distinctive skill: aligning continuously and making trade-offs at the right pace. A hybrid profile, more technical than the classic PM (comfortable with LLMs, RAG, orchestration and CI/CD without coding) and more strategic, because when execution is fluid, it is the quality of the intent that makes the difference.
The paradigm does not level the field, it widens the gap: an excellent practitioner sees their impact multiplied tenfold; one who frames and verifies poorly, much less so. And because they steer a system, the Product Builder also acts as a guarantor: they prevent agentic debt (ghost agents, lost contexts, undetected hallucinations) through good documentation and agent versioning hygiene.
Agentic AI Coach
This is the least settled role, and the one Niji considers the most strategic. They work within real teams, on real projects, to make an agentic way of working sustainable. They diagnose maturity (from the occasional assistant to autonomous agents under supervision), set the framework (harnesses, prompt versioning, supervision rules), train teams, spread practices and address agentic debt.
Their field is not an isolated squad but the flow of the whole chain: they support the transformation of ways of working between business and tech, upstream, where the real gains are made. What sets them apart from an Agile coach: they do not frame delivery processes, they frame autonomous action systems, which requires understanding the risks specific to agents. A role that is naturally suited to outsourcing: an organization growing in maturity needs an outside perspective to see its blind spots. They are the ones who sustain the change over time: without this overall framing, a team's local acceleration stays sterile and falls flat as soon as you try to scale.
Niji's conviction
The shift to value creator cannot be decreed: it needs tooling and support. It asks four simultaneous things of the Product Builder:
- More business. Owning the excellence of the spec and the framing, now the limiting factor of value.
- More delivery. Joining the team in real time, in the language and tools of development.
- More AI. Mastering agents and their orchestration, not just using them.
- More steering. Choosing the topics (what to do, and above all what not to do), getting through the gates, steering the flow, controlling cost: the token as a resource.
These requirements only hold if someone puts them in place and maintains them: that is the role of the Agentic AI Coach, who ensures that acceleration benefits the whole chain and not just one isolated link.
Our conviction: in the agentic era, the advantage does not go to those who produce the fastest, but to those who best decide what to build and know how to guarantee it. The profession does not disappear: it rises from producing artifacts to steering value. That is the move from specialist to value creator.
