CLOSE
megamenu-tech
CLOSE
service-image

Company

CLOSE
CLOSE
CLOSE
Embassy Case Study

Embassy

A multi-agent enterprise assistant on Claude Sonnet 4.6 and Haiku 4.5

Embassy employees used to move between SAP, GreytHR, ClickUp, ERP reporting and SharePoint to get through a working day. Reckonsys built a multi-agent AI platform on Claude that puts all of it behind one role-aware conversation.

One question, the right agent
Asked in the conversation
Does this vendor contract auto-renew, and how many leave days do I have left?
Legal
Contract clause analysis
HR
GreytHR
Finance
SAP · ERP reporting
Projects
ClickUp
Recruiting
Pipeline & screening
Documents
SharePoint
Two agents answer one question. Legal reads the renewal clause with Claude Sonnet 4.6; HR returns the balance from GreytHR. 9+ agents and 15+ internal tools sit behind the same conversation.
Client
EmbassyEmbassy Group
Industry
Real Estate
Anthropic product
Claude API
Models in production
Claude Sonnet 4.6Claude Haiku 4.5
Deployment
ProductionActive since April 2026
01

The customer

Embassy Group is one of India's largest real estate developers. Its employees worked across a wide estate of business systems - SAP for finance, GreytHR for people, ClickUp for project work, ERP reporting, SharePoint for documents - each with its own interface, its own login and its own idea of how a question should be phrased.

Everyday tasks meant knowing which system held the answer before you could ask for it. The cost was not any single system; it was the switching between them.

02

The challenge

  • Put legal, HR, project management, finance, recruiting and document workflows behind one conversational interface.
  • Route each request to the right agent, and to more than one when a question spans departments.
  • Keep every response scoped to what the person asking is entitled to see.
  • Build it multi-tenant, so additional Embassy business units can be onboarded without a rebuild.
03

A platform of agents, not one assistant

Rather than a single general assistant, Reckonsys built a multi-agent platform: each business function is its own agent with its own tools, and a router decides which of them a question needs.

  • Legal
    Contract clause analysis and document review - the highest-judgment work on the platform, and where Claude does the heaviest lifting.
  • HR
    Leave, employee data and people workflows, integrated with GreytHR.
  • Finance
    Reporting and financial queries against SAP and the ERP reporting layer.
  • Project management
    Task and project state, integrated with ClickUp.
  • Recruiting
    Candidate pipeline and screening workflows.
  • Documents
    Retrieval across the SharePoint estate.

Nine or more agents and fifteen or more internal tools have shipped across those functions. Because the platform is multi-tenant, further Embassy business units - OfficeParks and Embark among them - can be onboarded independently rather than forked.

04

How Claude is used in production

Two models run in production, chosen by workload. Claude Sonnet 4.6 handles the high-judgment work: contract clause analysis, document reasoning, and the routing decisions that determine which agent or agents a question belongs to. Claude Haiku 4.5 takes the higher-frequency, latency-sensitive calls where a fast answer matters more than deep reasoning.

Both are accessed through Azure AI Foundry and orchestrated with LiteLLM, which gives Embassy a single provider-abstracted interface across the platform and keeps model selection a routing decision rather than an application rewrite.

Contract clause analysis is the clearest illustration of why model choice matters here. Asking whether an agreement auto-renews is not a lookup - it means reading the renewal language, the notice provisions and whatever qualifies them, then saying plainly what it amounts to. That is judgement under constraint, and it is the work Sonnet is carrying.

05

What shipped

9+
Distinct agents in production
15+
Internal tools integrated
5
Business systems unified
2
Claude models, routed by workload

Across legal, HR, task management, finance and recruiting, integrating SAP, GreytHR, ClickUp, ERP reporting and SharePoint behind a single conversational surface.

06

The outcome

Employees ask one question in one place instead of learning five systems. A request that spans departments - a contract question and a leave balance in the same breath - is answered by the agents that own each part, without the person needing to know the platform has agents at all.

The multi-tenant architecture means the same platform extends to further Embassy business units as they come on.

Stack

Technologies

Claude Sonnet 4.6 · Claude Haiku 4.5 · Azure AI Foundry · LiteLLM orchestration · multi-agent routing with internal tool calling · SAP · GreytHR · ClickUp · ERP reporting · SharePoint · multi-tenant architecture

Contact us

Let's put your whole company behind one conversation.

Get in touch

Modal_img.max-3000x1500

Discover Next-Generation AI Solutions for Your Business!

Let's collaborate to turn your business challenges into AI-powered success stories.

Get Started