MANUFACTURING · DATA PLATFORM

One governed context layer for every agent on a manufacturer's data platform

Procurement, finance and support agents each carried their own copy of supplier contracts, terms and rules, so a change to payment terms reached them at different times. Wexa served all of them from one governed context layer through a single MCP tool.

Download this case as PDF Book a demo

manual edits to propagate a contract-term change
to propagate it to every agent
tokens per answer (107,698 → 41,560)
cross-agent answers consistent before and after the change

CLIENT: Manufacturer running AI agents on a leading cloud data platform (with the platform vendor)
REGION: US
WHEN: 2026
ENGAGEMENT: Partner build
WEXA COMPONENTS: Wexa · data-platform connector · context graph · MCP get_context · per-agent policies

01Case 07

The challenge

Context defined three times by three teams
Rules such as "support never sees unit prices" enforced per agent
No single trace of what context an agent used for an answer

What Wexa built

Platform tables and contract PDFs linked in the context graph through the data-platform connector
One MCP tool, get_context, registered by every agent; each agent has its own key and clearances
Field masking and object hiding enforced on every read; only the active contract version served
A trace per answer: which row and clause, which version, which policy allowed or refused each field
A fourth (quality) agent added live in under 3 seconds

02Case 07

How it runs on Wexa

FigureHow this use case runs on Wexa: client systems connected, ontology discovered, context graph in the Wexa memory layer, agents and process flows, and governance. Highlighted components are central to this use case.

03Case 07

Product in use

FigureLive answers from two agents with the trace drawer showing evidence rows, contract clauses and the policy applied (product screen).

04Case 07

Results

Measure	Before / without Wexa	With Wexa
Engineering effort to define and keep each agent team's context in sync (3 teams)	3 FTE	0.5 FTE
Annual cost of the process	$450K	$75K
Time to implement the AI solution	6 months	2 weeks
AI build cost (one-time)	$480K	$15K
AI operating cost per year (models, infrastructure, upkeep, platform)	$312K	$171K
Total AI cost in year 1	$792K	$186K

With Wexa

net annual value to the client after Wexa's own operating cost

faster to implement

lower AI operating cost

less AI spend in year 1

Head-to-head against the platform's native setup on the same agents and questions: median answer 11.4 s vs 15.5 s
Governance acceptance tests 72 / 72; a specific trace found among 873 in 2.2 s
Context call latency p50 0.71 s, p95 1.52 s