The HUMAN_INPUT marker appearing in this article is a placeholder left by the AI writing skill to indicate “a human should fill in the confirmed value here later” (format example: <!-- HUMAN_INPUT: enter the value -->).

Business Portfolio Design — How to Choose Between Outsourcing and Productization on an AI-Native Organizational Foundation

Before wrestling with “outsourcing or productization” as a binary choice, there is a prior question: what does the organization doing the work actually have?

Marketing, sales, internal knowledge management, engineering, sales assistance — if you first build AI agent infrastructure for each of these operational domains, both outsourcing and product development can run in parallel as business options on top of that foundation. Start outsourcing halfway through and let the organization get pulled by resource demands, and you lose the opportunity to build this foundation altogether.

Yakumo made this sequence — build the AI-native organizational foundation first, then design the business portfolio on top — its founding decision from day one. This article organizes that framework, treating bateson Booking Engine as one example of a business choice deployed on top of the foundation. The full decision framework is covered in the parent pillar “AI-Native Organizational Foundation and Business Portfolio — Phase Strategy as Seen in bateson”.

Build the Organization Before Choosing the Business — The Sequencing Decision

Before debating “outsourcing or productization,” there is one more upstream question: does your organization currently have what it takes to execute that business?

An organization centered on AI agents means having agent infrastructure in place for each operational domain — marketing, sales, internal knowledge management, engineering, and sales assistance. Start outsourcing without this foundation, and all of the organization’s resources get pulled entirely into outsourced work. Handling routine tasks like inquiry responses, proposal writing, and project management manually while simultaneously thinking through the next business from scratch is structurally untenable.

Yakumo’s founding decision was “do the foundation work first.” Even when outsourcing opportunities were available, we deliberately held them off and built the infrastructure to operate as an AI-native organization first. This is not opportunity loss — it is upfront investment to expand business options later. Start outsourcing halfway and let resources get consumed, and you lose the opportunity to build this foundation altogether.

Components of an AI-Native Organizational Foundation

The phrase “AI-native organizational foundation” may sound abstract, but in practice it means re-framing the question for each operational domain: “Who handles this work?” “What parts can an AI agent replace?” “Where is the boundary between routine and non-routine?” Working through these questions for each domain produces the following set of foundational infrastructure.

Once this infrastructure is in place, routine work scales even as outsourcing volume grows, and time for product development becomes available. Trying to run outsourcing and product development in parallel without this foundation means people are split halfway between both, inadequately.

Yakumo’s current AI agent infrastructure uses four directors, one per operational domain. Each of the four domains — sales, marketing, system development, and web development — has an agent responsible for policy decisions, beneath which sit sub-agents and deterministic skills. Pre-order proposals and estimates, content generation and SEO, internal product implementation, and corporate site design are each separated as responsibilities under their respective directors.

Business Portfolio on Top of the Foundation — Outsourcing and Products as Equal Choices

Once the foundation is in place, “outsourcing or product” is no longer a binary choice. Both become business options sitting on the same underlying infrastructure.

The advantage of outsourcing (delivering outputs to individual clients) is immediate revenue and customer learning. You gain deep insight into clients’ operational challenges, and that knowledge feeds back into improving the foundation. The advantage of product (a business that continuously licenses usage rights) is scalability and medium-term asset building. Once a system is built, marginal cost does not change as the customer base grows.

These two do not compete. As long as both sit on the same foundation, the customer understanding gained through outsourcing refines product design, and the maturity of the product strengthens the persuasiveness of outsourcing proposals. It is not a one-way causality of “accumulate through outsourcing, then productize” — it is a structure where both run in parallel sharing the same foundation.

The substance of the decision is really a question of allocation: “which business do we put more effort into right now?” Being able to change that allocation is precisely what the foundation enables. Without the foundation, no matter which you focus on, people continue to be consumed by routine work.

Yakumo’s business structure is built around its own products. The design is to generalize the SaaS originally built for internal use to serve external customers, with AI agents handling its operation — an AI-BPO model that “turns the work itself into a service.” Custom development and consulting sit outside that core, positioned to deepen understanding of customer operations and feed learnings back into the internal foundation.

Decision Criteria for Making Parallel Operation Work

When running outsourcing and product development in parallel on top of the foundation, four criteria distinguish “parallel operation that works” from “parallel operation in name only.”

Foundation generalizability: Is the work done on outsourcing projects feeding back into improvements to the foundation? If you are only ever chasing customer-specific responses without those learnings being reflected in the foundation, parallel operation is just outsourcing growing larger.

Outsourcing as validation field: Are outsourcing projects functioning as a validation field for the problems your product is trying to solve? Whether you can use outsourcing as a channel to learn what is actually problematic in customers’ real environments determines the directional accuracy of your product.

Self-funding path: Can outsourcing revenue self-fund product development? Whether you have a path to grow the product without relying on external capital directly determines your freedom in decision-making.

Decision-maker in place: Is there a clearly designated person responsible for prioritizing product decisions? Outsourcing comes with concrete requests from clients, but a product only moves forward if someone keeps making decisions about it. A structure that prevents product decision-making from stalling during peak outsourcing periods needs to be in place first.

The bateson Case — How Outsourcing Work Becomes Multi-Tenant on the Foundation

bateson (the booking engine Yakumo implemented in src/lib/bateson/) serves as a concrete example of this “business development on top of the foundation.”

Phase A is the stage of “implement as a booking engine for Yakumo’s own use.” The decisions at this stage are about structure, not feature count. Separating customer-specific implementations, designing for tenant isolation to be added later, allowing external vendor swap-out, and enabling external injection of brand elements — building these into Phase A decisions is equivalent to “keeping the door open for future multi-tenanting.”

Phase B is “offering to external customers (multi-tenanting).” If the structural decisions from Phase A are maintained, this transition can be implemented as an extension rather than a refactor. The criterion for making this transition is “is the next external customer candidate concretely visible?” Investing in Phase B without clear prospects carries high risk.

Phase C / D are horizontal expansion to multiple tenants and functional integration as a sales funnel. The reason bateson is designed not just as a standalone booking tool but to create “an experience of touching Yakumo’s product from the moment of inquiry” is that this context was built into the implementation from Phase A.

As of 2026-05, bateson remains in Phase A. The src/lib/bateson/ implementation operates only within Yakumo’s own inquiry flow, with no deployment to external customers and no public packaging. The decision is not “deferral” but rather “the next external customer candidate is not yet concretely visible” — and because the Phase A structural decisions (full tenantId parameterization, adapter separation, cancelToken hashing) are maintained, the door to Phase B remains open and waiting.

Organizational Operations During Parallel Running

The core of organizational design for effective parallel running is “designating the product decision-maker first.” Outsourcing comes with concrete requests from clients. A product does not move forward unless someone continues to make decisions. If outsourcing keeps getting priority during busy periods, the product accumulates as “things to do in the future” and time passes without it being touched.

A practical way to prevent this is to “formalize a weekly minimum guaranteed time.” Codify it as a rule — “regardless of outsourcing status, a fixed number of hours per week go to product development” — and maintain operations with no exceptions. In weeks when this time is not protected, rather than rolling it over to the next week, record why it was not protected. If the cause is “a structural problem with the outsourcing engagement,” the criteria for accepting work need to change.

The other criterion is “the standard for accepting outsourcing work.” Accepting indiscriminately — projects that do not feed back into foundation improvements, projects that do not serve as validation fields for the product — eliminates the benefit of parallel running. The habit of asking “how does this project contribute to parallel running?” at the point of accepting work is what keeps the medium-term business structure healthy.

As of 2026-05, no serious priority conflicts have been experienced. Even during periods when outsourcing work came in, we have continued operating with product development (the internal foundation and bateson) as the higher priority, and there have been no situations where outsourcing busyness caused product decision-making to stall. This is largely because the decision to “prioritize product” was set in advance, meaning the structure does not require making that judgment again each time volume increases.

Decision Checklist

Before asking “how do we choose between outsourcing and product,” here are the prerequisite conditions to check.

Conclusion

The most upstream question in business selection is not “outsourcing or product” but “is the organizational foundation capable of executing that business already in place?” Without the foundation, either business choice results in non-scalable execution dependent on human labor. With the foundation, both outsourcing and product can be designed as a parallelizable business portfolio sitting on the same base infrastructure.

bateson is a concrete example of “Phase expansion premised on the foundation being in place.”

As of 2026-05, bateson’s Phase A is complete. src/lib/bateson/ has the functionality needed as a booking engine implemented, with all four principles reflected: separation of customer-specific implementations, adapter design allowing tenant isolation to be added later, swap-out capability for external vendors including Google Calendar and Gmail, and template injection of brand elements. It is in live operation within Yakumo’s inquiry flow.

If the Phase A structural decisions are maintained, the door to Phase B is open. Confirm that the door is open, then make the Phase B decision when an external customer candidate becomes visible — if this sequence breaks down, multi-tenanting becomes a “rebuild” with significant cost.

That said, Yakumo itself has not formalized an explicit rule like “X hours per week” as of 2026-05. The reality is that operations work with the decision to place product development at the top of the priority stack, so a situation requiring quantitative rules has not yet arisen. The expectation is that time-based rules will first become necessary when the proportion of outsourcing rises, or when Phase B triggers become visible and the decision-making load increases.

The full picture of the framework — design the parallel-running organization first, then choose the business — is covered in “AI-Native Organizational Foundation and Business Portfolio — Phase Strategy as Seen in bateson”. For the technical implementation of the engine, see “Design Strategy for Building a Booking Engine In-House”.