The 500-Agent Catalogue: NTT DATA Borrows Palantir's Vocabulary and Reframes the SI Business Model

The interesting thing about NTT DATA's Google Cloud announcement is not the headline numbers. It is that a 190,000-person systems integrator (SI: a firm that builds and runs IT and now AI systems for large enterprises) has chosen to describe its delivery model using Palantir's vocabulary, and to describe its output as a catalogue of 500 agents rather than a book of consulting hours. Both of those moves matter more than the certification count.

On 9 June, NTT DATA said it would build a dedicated Gemini Enterprise practice with Google Cloud, train 5,000 certified experts inside it, and co-innovate "up to 500 AI agents" across banking, insurance, manufacturing and retail. The press release explicitly uses the phrase forward-deployed engineers to describe how the work will be delivered.¹ That phrase is not generic. It is the specific term Palantir has spent a decade branding as its defensible delivery model, and seeing it lifted into a global SI press release is the part of this announcement worth dwelling on.

The vocabulary tell. When a category-defining firm's proprietary language shows up unattributed in a competitor's marketing copy, one of two things is happening. Either the term has become generic, in the way "cloud-native" or "DevOps" became generic — drained of moat, useful only as shorthand. Or the competitor is staking a deliberate claim that they can do the same thing. In NTT DATA's case it is plainly the first. Forward-deployed engineering, embedding senior technical staff at the client to build the system in the room, is now the assumed delivery pattern for serious enterprise AI work. It is what KPMG is doing with Microsoft, what IBM is doing with Google Cloud, and what NTT DATA is now doing at scale.

This is not nothing for Palantir, but it is also not fatal. The defensible parts of Palantir's business — Foundry's data ontology, the AIP runtime, the brand premium on national-security and critical-infrastructure work — sit underneath the delivery layer. What erodes is the implicit story that nobody else can deliver AI this way. That story has been quietly true for years and is now quietly false.

The catalogue claim is the bigger story. "Up to 500 AI agents, co-innovated with Google Cloud" is a structurally different promise from "we will staff your transformation." It implies a library of pre-built, configurable artefacts — reusable building blocks that get deployed across clients rather than rebuilt from scratch each time. If NTT DATA actually ships this, the unit of sale changes. You are no longer buying 14,000 consultant-hours at a blended rate. You are buying agents from a catalogue, with implementation around them.

That matters because the SI business model has always been a time-and-materials margin on people. Agents in a catalogue are closer to software than to services — they have versioning, support contracts, performance specs, and ideally a price per deployment rather than a price per hour. The economics of selling the 47th deployment of the same agent are not the economics of selling 47 bespoke engagements.

I want to be careful here. The press release says "up to 500", which is a ceiling, not a commitment. It gives no timeline, no revenue target, and no description of how an "agent" is validated as production-ready versus shelfware. The structural shape of the announcement, vague ceiling, no operational milestones, is identical to the "10,000 trained consultants" claims every large SI made during the 2015–2019 cloud-migration wave. Most of those took five to seven years to translate into real deployable capacity, and several never did.

The 5,000-certified-experts number invites the same scepticism. That figure represents roughly 2.6% of NTT DATA's global workforce.² A certification is a passed exam, not delivered production work. So the dedicated practice is, in structural terms, a thin specialist layer over a much larger base of staff who will continue doing what they were doing before. Which is fine, that is what a practice is, but it is not the transformation the headline implies.

The unit of sale is changing from a consultant-hour to a catalogued agent, and the SIs that miss this are going to look very expensive very quickly.

What the verticals tell you. Banking, insurance, manufacturing, retail. The first two are the interesting picks. Banking and insurance are the densest concentrations of structured knowledge work in the global economy, underwriting, claims adjudication, KYC, trade operations, procurement, and they are the workflows where agents have the clearest line of sight to displacing or compressing human labour. When NTT DATA and Google Cloud name those verticals first, they are telling you where they think the first at-scale agent deployments will land. The labour-substitution story is most advanced where the work is most proceduralised, and that is where the catalogue gets sold first.

The hyperscaler distribution read. Three big SI-hyperscaler announcements inside 48 hours, IBM with Google Cloud, KPMG with Microsoft, NTT DATA with Google Cloud, is not coordination. It is competitive forcing. The hyperscalers cannot reach the enterprise install base fast enough through direct sales, and the adoption gap (the lag between model capability and enterprise deployment) is the binding constraint on their revenue growth, not model quality. They need distribution. The SIs are being recruited into that role, and the open question is whether they retain pricing leverage or get compressed into something close to resellers with implementation services attached.

NTT DATA's answer, judging by the catalogue framing, is to try to retain leverage by owning the agents. If the 500 agents are genuinely NTT DATA IP wrapped around Google Cloud reference architecture, that is a defensible position. If they are Google IP with NTT DATA branding, the SI is the distribution channel and the margin compresses accordingly. The press release does not resolve which it is, and the answer will determine whether this practice is a real business in three years or a line item in a future restructuring slide.

The vocabulary has shifted, the unit of sale is shifting, and the hyperscalers are quietly rewiring the global SI bench into their distribution layer. The 5,000 certifications are the noise. The 500-agent catalogue and the borrowed Palantir vocabulary are the signal.

Footnotes

¶