Platforms

DXC cut IT resolution time 95% on free LangGraph code as AWS and Google priced the layers around it

The production architecture DXC published shows the agent orchestration layer is free open-source code, and AWS and Google priced the grader, search and scoring around it the same week. That split decides whose margin holds: DXC's, Amazon's and Alphabet's.

Richard TangRichard Tangthree-minute read
Share
Raul Fernandez, chief executive of DXC Technology, in a studio portrait
1 / 6Slide 1 of 6
Raul Fernandez, chief executive of DXC Technology, which published the production architecture of its OASIS agent harness on November 11.

DXC Technology−3.08% — DXC Technology, down 3.08 percent today published the production architecture of OASIS, the agent harness it runs for banks, hospitals and governments, and the detail that matters is what sits underneath: the orchestration runtime is LangGraph, the open-source agent-orchestration framework from LangChain$25M — LangChain, private, latest valuation $25M. DXC cut average resolution time by 95% across 56 measured use cases in the months after general availability in March 1.

On November 5 we argued the big platforms are commoditising the layer LangChain charges for; what is new is that DXC has now published a generally available, regulated product in which the orchestration code is exactly that free layer, and the same week two hyperscalers priced the paid layers around it 8.

DXC sells the harness, and the runtime inside it costs nothing

DXC split the system in two. Amazon+1.06% — Amazon, up 1.06 percent today Bedrock is the substrate: a governed gateway to Claude, Nova, Llama and open-weight models, with guardrails and private networking applied uniformly 1. Everything on top is DXC's: orchestration on LangGraph, sixty years of runbooks as context, and four authority tiers where every persona starts read-only and nothing touches a production server without recorded approval 1.

DXC says it designed the whole stack against coupling to any single provider, framework or deployment target 1. That governance harness is where an engineer's day actually changes: from firefighting a failed server to approving a diagnosis the agent already grounded in evidence.

The paid layers sit around the free runtime, and they got priced this week

No vendor can meter the orchestration runtime itself, because it ships as free code; the metering sits in the grading, search and scoring layers around it. AWS open-sourced the Universal Agent Evaluation Framework, a grader that scores agents built on LangGraph, LangChain, Strands or Amazon Bedrock AgentCore, AWS's agent hosting service, hardened so an agent cannot talk its way to a better grade 2.

Its judge treats everything an agent produces as untrusted input and demands a verbatim quoted span as evidence before accepting any score 3. The same day, AWS put a standalone web-search API on AgentCore at $7 per 1,000 queries 4. Google shipped its Decisions API the same week, scoring up to 1,000 rubric questions in one pass at $0.10 per million input tokens 5. On this evidence, the metered layers, grading, search and scoring, sit with AWS and Google+2.40% — Google, up 2.40 percent today, and no vendor can charge twice for the runtime.

Buyers want movability, and DXC's revenue needs the margin

The buyer side agrees on the runtime. Orange's enterprise unit keeps its agents on LangChain's open framework so the components can move to Google Vertex or on-premises, and it measures AI value in speed and revenue protection, not headcount 6. That preference is why DXC could ship LangGraph inside a regulated product: the code is auditable by a regulator and rebuildable on another cloud.

DXC's economics, though, depend on it. Revenue from its SEC filings has fallen year on year for twelve straight quarters, so a 95% resolution-time cut is margin defense on a shrinking base 7.

DXC's revenue fell year on year in all of the last 12 quarters

OASIS's productivity gains land on a revenue base that shrinks every year

2.9B3B3.1B3.2B3.3B3.4BQ1 '24Q3 '24Q1 '25Q3 '25Q1 '26Q3 '263.03B
Data
Quarterly revenue
Q4 '233.4B
Q1 '243.39B
Q2 '243.24B
Q3 '243.24B
Q4 '243.23B
Q1 '253.17B
Q2 '253.16B
Q3 '253.16B
Q4 '253.19B
Q1 '263.13B
Q2 '263B
Q3 '263.03B
Quarterly revenue in USD billions, DXC Technology SEC filings; year-on-year change negative in all 12 quarters7

Price the bookings on free code

DXC closes the gap that matters, turning model capability into governed work on production systems, and deserves the reward for it. Its defensible asset is the runbooks and the governance record, not the orchestration layer. Price its bookings on free code and an AWS substrate, because that is what the architecture publishes. Watch the next DXC quarterly print: whether a 95% resolution-time cut shows up in bookings tells you if the harness is worth paying for once the runtime is free.

Get Platforms by email, free

The cloud, data and developer software AI is built on. In our daily email, every weekday.

Loading GOOGL

Reader comments

0 comments

    More in Platforms

    All in this topic →

    Platforms

    Cloudflare hires a Splunk sales chief to price AI agent traffic as shares trade at 169 times earnings

    Cloudflare hired a Splunk-grade sales chief right after its fastest growth in four years. His real job: turning the AI agent traffic already on its network into revenue the market at 169x earnings is paying for in advance.

    Platforms

    LiteLLM's '7x faster' claim is missing from its own benchmark data; Stripe pays $8 billion for OpenRouter

    A line-by-line check of LiteLLM's own benchmark files finds no 7x gap, while Google sells routing decisions at $0.10 per million tokens and Stripe pays $8 billion for a router. The moat is governance, and it is bleeding.

    Platforms

    Amazon blocks Meta's AI agent, Walmart opens a gateway, and Sierra's $400 million book wins either way

    Amazon blocked Meta's Muse in September and eBay has now restricted outside agents, while Walmart opened a gateway. Sierra co-writes the open protocol for agent access and also sells outcome-priced agents to the enterprises that may gate it, so its $400 million ARR book leans toward whichever way retailers decide.

    Platforms

    Twilio lands its first eight-figure voice-AI deal, then says 99% of the market is untapped

    The company says voice AI is 1% of its minutes and the money is ahead, but the largest insider seller is a director tied to Sachem Head, not the CEO. Whose money moves: anyone holding a stock priced 49.7% above its own fair-value estimate.

    Platforms

    Zeta Global's 46% growth was 30% organic, and its Q4 guide slows to 22% to 24%

    In its November 10 Q3 filing, the marketing-cloud company's growth excluding the Marigold acquisition and election revenue was 30%, and its own Q4 guide slows that to 22% to 24%.

    Platforms

    Percona shipped a free MongoDB AI search twin; BTIG's $417 target on MongoDB rests on Atlas

    MongoDB put AI embeddings inside its paid Atlas cloud on November 9; Percona had already shipped the same search stack self-hosted and free. A $417 Buy from the brokerage BTIG and MongoDB's December 1 earnings will price the difference.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.