A couple of weeks ago, Anthropic announced a $1.5 billion venture with Blackstone, Goldman Sachs, and Hellman & Friedman to embed its engineers directly inside customer organizations. OpenAI is reportedly raising more than $4 billion from TPG, Brookfield, Advent, and Bain for a similar effort, valued at $10 billion. Two of the most valuable AI companies on the planet have decided that selling models is not enough. They now need to be in the room, helping companies actually use them.
Both moves quietly admit something the industry has been dancing around for two years. The hard part of AI is not the model. The hard part is the organization. Microsoft’s latest Work Trend Index puts numbers on it: organizational factors (culture, manager behavior, talent practices) account for more than 2x the real-world impact of AI compared to individual capability. Only 19% of companies sit at what Microsoft calls the “frontier,” where individual AI fluency and organizational readiness both run high. Half are still emergent. The rest are stuck somewhere in between.
If you run a mid-sized B2B manufacturer in Taiwan, the most interesting line in Anthropic’s announcement was the one about their target market. Their existing Claude Partner Network (Accenture, Deloitte, PwC) handles the global enterprises. The new venture will work with mid-sized companies. That is a deliberate carve-out, and it tells you something. Even Anthropic sees mid-market as a distinct problem that needs a distinct model.
So here is the question worth asking.
Does that model translate to Taiwan’s mid-sized B2B sector?
The Anthropic Model and What It Assumes
Anthropic’s announcement describes their typical engagement clearly. A small team starts with the customer to identify where Claude can have the biggest impact. Applied AI engineers from Anthropic then work alongside the customer’s engineering team to build custom solutions and “support customers over the long term.”
Two assumptions are baked into that model.
The first is that the client has an engineering organization mature enough to pair with embedded AI engineers. Anthropic’s design assumes engineer-to-engineer collaboration as the operating principle.
The second is that the engagement is long. Diagnosis exists to seed a continuous embedded relationship. This is the same playbook Palantir pioneered with its forward-deployed engineers, and it is now becoming standard practice across AI deployment consulting.
Neither assumption fits the typical Taiwanese mid-sized OEM. Engineering in these companies usually means process engineering and manufacturing engineering, not the kind of internal software organization that can absorb an embedded applied AI engineer. And the Lao Ban culture, pragmatic, capital-disciplined, and rightly skeptical of long agency commitments, does not buy multi-year embedded transformation engagements on faith.
This is not a critique of Anthropic’s design. Their model is well-suited to mid-sized US software-adjacent firms. It is just not the right shape for the Taiwanese manufacturing floor.
What the Local Equivalent Needs to Look Like
If diagnosis before deployment is the right idea, and we agree that it is, the local version has to start from a different operating model.
It cannot be engineer-to-engineer, because most clients do not have the engineering counterpart. It has to be operations-to-operations, working with the people who actually know where decisions get stuck, where the brand promise breaks down at the seams, and where workflow ownership is unclear.
It cannot be designed to maximize engagement length. Taiwanese B2B owners do not want to be embedded with. They want a clear diagnosis, a tangible quick win, and a roadmap they can act on. Whether they act on it with Geber, with another partner, or in-house is their decision, not ours.
This is the shape of the Geber Intelligence Scan, and the design is deliberate. The scan looks at a small set of operational workflows, identifies a candidate for a quick win, delivers that win, and leaves the client with a roadmap that travels with them. The roadmap is theirs. If the next phase makes sense with Geber, we earn it on the work, not on lock-in.
Why the Team Matters More Than the Methodology
There is a temptation, when describing this kind of work, to reach for a framework. A four-step diagram, a maturity model, a methodology with a name. That is the consulting industry default, and it is mostly theatre. What actually determines whether an Intelligence Scan produces something useful is not the framework. It is who is in the room.
The Scan is built around three different kinds of attention, and they need to sit on the same call to work.
Jonathan leads the applied AI and operational engineering side. He is the one who decides what is feasible to build inside a four-week window, what should be left to an off-the-shelf tool, and what is not worth doing at all. The reason that matters: most Taiwanese mid-sized manufacturers have been pitched AI by people with something to sell. Jonathan’s job is to be the person in the room who tells you when you do not need to buy anything.
Lulu Ding runs the technical side. Five-plus years as an SAP specialist across insurance, cloud, and semiconductors mean she has spent her career inside the kind of enterprise systems most Taiwanese mid-sized manufacturers actually run on. She is also a full-stack developer, which matters because the Scan does not stop at recommendations. When the quick win is buildable, Lulu is the one who wires it into what the client already has, rather than handing over a parallel system no one will end up maintaining.
Ofelia Yang runs the operations side. She’s optimized global marketing and channel workflows moving 80,000+ units a year across international markets, so she knows what scaling actually feels like at the operational layer, not at the slide-deck layer. Her job on the Scan is to make sure the candidate workflow we identify is the one where a quick win actually compounds, not the one that just looks best in a report.
That triangle is not accidental. It is what separates a Scan from a slide deck. Anthropic’s engineers, brilliant as they are, are paid to build. The Geber team is paid to ask whether anything needs to be built in the first place. For a Taiwanese manufacturer trying to figure out where AI fits inside an already-running business, that is the more useful question.
The Bigger Pattern
OpenAI and Anthropic getting into consulting is not a small story. It is the strongest signal yet that AI value will be unlocked at the level of organizational design, not at the level of model selection. Microsoft’s data, the partner network buildout, the deployment company structures, all of it points the same way. Implementation is the bottleneck.
But the model the labs are building is shaped by their customer base. For Taiwan’s mid-sized manufacturers, the work has to look different. Less embedded engineering, more operational intelligence. Less long-term lock-in, more portable roadmaps. Less faith in capability overhang, more attention to the everyday workflows where capability actually meets decision.
The Lao Bans who get this right will not be the ones who hire the biggest name. They will be the ones who insist on diagnosis first, results fast, and a roadmap they own.
That is what the Intelligence Scan is built to do, and it is the work the team is built to deliver.
