Case Studies
Building an AI operating system for research, content and commercial workflows
From isolated AI experiments to a shared system for people and agents
Sector
Automotive
Client
European multi-brand automotive group
Challenge
Fragmented research, content and commercial workflows
Solution
Modular AI operating system connecting specialised agents, shared knowledge, business tools and human approval
01
The challenge
AI experiments were multiplying. The operating model around them was not.
The automotive group operates across several brands, commercial initiatives and teams.
Every day, those teams process information from competitors, partners, customers, social channels, internal knowledge and market developments.
The opportunities for AI were clear.
So was the risk of solving every problem separately.
Research was still largely manual. Relevant market signals took time to reach the right people. Content moved through several tools and approval steps. Commercial teams repeatedly researched the same companies and topics. Each new AI use case risked becoming another isolated workflow with its own prompts, tools and knowledge.
The client wanted to move beyond individual AI experiments and create a shared foundation that could support many workflows over time.
02
What needed to change
The goal was not simply to automate more tasks.
The client needed a system where AI and people could work together across brands and business functions.
That meant creating a foundation that could:
reuse knowledge across workflows
coordinate specialised AI agents
connect to existing business tools
retain context between tasks
keep people responsible for creative and external outputs
support different models rather than depend on one provider
add new workflows without rebuilding the architecture
make AI activity easier to monitor and improve
The target was a modular operating system capable of growing towards 20+ specialised agents as new use cases were introduced.
03
What we built
A shared AI and human operating system.
WeAreBrain designed a modular application layer connecting AI agents, company knowledge, existing tools and human decision points.
Instead of building a separate application for every use case, the architecture provides reusable building blocks that different workflows can share.
At the centre is an orchestration layer responsible for understanding tasks, routing work and coordinating the agents and tools needed to complete them.
Around it sit specialised agents for activities such as:
Research
Market monitoring
Content creation
Commercial preparation
Image and video production
Knowledge retrieval
A shared memory layer gives those workflows access to persistent organisational context rather than starting from an empty prompt every time.
Human approval remains part of the system wherever judgement, brand voice or external publication matters.
04
How the system works
1. Orchestration
A central orchestration layer receives requests and determines which agents, tools and information are needed.
This makes it possible to coordinate multi-step workflows rather than treating every AI interaction as an isolated request.
2. Shared knowledge and memory
The system combines structured data with semantic retrieval across company knowledge.
This allows different agents to work with shared context, including:
Brand information
Previous research
Documents and files
Commercial information
Workflow state
Earlier outputs
Knowledge becomes reusable across the system.
3. Specialised agents
Different agents handle different parts of the work.
A workflow might involve one agent gathering market information, another evaluating relevance, another retrieving internal knowledge and another preparing content.
The architecture allows additional agents to be introduced as new needs emerge.
4. Existing business tools
The operating system connects AI workflows to the tools people already use.
The architecture was designed around integrations such as:
CRM
Productivity tools
Communication and approval channels
Social publishing platforms
Company databases
External APIs
Teams do not need to move all their work into a new AI application before they can benefit from it.
5. Human approval
Automation stops where human judgement matters.
People remain responsible for areas such as:
Creative direction
Brand tone
Commercial judgement
Sensitive decisions
External publication
The system prepares and coordinates the work.
People make the final call.
05
Workflow example 1
From market signal to approved content
A market development, competitor move or partner announcement is identified.
The system then:
01. Monitors
Research agents collect relevant signals from approved external sources.
02. Evaluates
The information is assessed for relevance to the business and individual brands.
03. Adds context
The workflow retrieves relevant company knowledge, previous material and brand context.
04. Creates
Content agents prepare suitable draft material.
05. Reviews
A person reviews the output and adjusts it where needed.
06. Publishes
Approved content moves into the existing publishing workflow.
One signal can therefore move from discovery to useful output without teams manually transferring information between tools at every stage.
06
Workflow example 2
From commercial trigger to account briefing
The same architecture can support commercial teams.
When a company, prospect or market opportunity needs investigation, the workflow can:
Research the company
Identify relevant developments
Retrieve existing CRM context
Summarise useful information
Generate talking points
Suggest opportunity hypotheses
Prepare a meeting briefing
Surface relevant content ideas
Instead of repeatedly starting research from scratch, teams can build on shared organisational knowledge.
07
The role of SINAS
SINAS provides the agentic infrastructure behind the operating model.
Its role is to connect the different parts of the system:
Agents
Specialised AI workers for defined tasks.
Orchestration
Routing and coordinating multi-step workflows.
Models
Supporting different AI models based on the task.
Tools
Connecting agents to APIs, databases and business applications.
Memory
Maintaining context and reusable knowledge.
Execution
Running defined functions and workflow steps.
Human approval
Creating points where automation pauses for review.
This allows new workflows to reuse the same infrastructure instead of becoming separate AI applications.
08
The result
A foundation for scaling AI across the organisation.
The main outcome was not another individual AI tool.
It was a reusable architecture for introducing AI across different brands, teams and workflows.
The system creates a foundation for:
Faster intelligence to action
Research can move more directly into content, commercial preparation and other business workflows.
Reusable organisational knowledge
Agents and workflows can work with shared context rather than repeatedly rebuilding the same information.
Consistent human control
Approval is designed into workflows instead of being added manually after automation has happened.
Model flexibility
The architecture can work with different AI models rather than tying the operating model to a single provider.
Easier expansion
New agents and workflows can be introduced on top of the existing foundation.
Better visibility
Shared orchestration, workflow state and logging make AI activity easier to inspect and improve.
09
Why this matters
Many organisations are now moving through the same transition.
The first phase of enterprise AI was about giving individuals access to models.
The next is about connecting AI to organisational knowledge, tools and workflows.
That requires more than prompts.
It requires an operating model for how agents, software and people work together.
This project created that foundation.
Product
SINAS
Agentic infrastructure for connecting models, agents, tools, data and human approval.
AI services used
AI Agents
Designing reusable multi-agent workflows that can research, create, coordinate and escalate.
AI Strategy
Moving from a collection of individual use cases towards a shared AI operating model.
Data Infrastructure
Creating persistent knowledge and context that can be reused across agents and workflows.





Moving from AI pilots to shared infrastructure?
If different teams are building separate AI workflows, the next step may not be another use case. It may be a common foundation they can all build on.