What the Company Brain is, how it is built, how it is governed and how it reaches any AI. Explainers, architecture and cases, all in one place.
Explainers, architecture decisions and cases, published by the Strattum team.
Arthur Guttilla · · 5 min Context quality Company Brain: when the CRM and the ERP report different values The CRM holds the negotiated amount, the ERP holds the invoiced one, and both are correct in their own context. What has to be defined for an agent to know which of them answers the question. Arthur Guttilla · · 5 min Governance and trust Company Brain: the context an agent needs before it acts A wrong answer is corrected by whoever reads it. A wrong action becomes a record and steers the decisions after it. What an agent has to have settled before it executes. Arthur Guttilla · · 4 min Fundamentals Company Brain: AI answers carrying each customer’s data A standard company text covers tone of voice and product naming, and never reaches the contract of the customer who just called. What changes once the general rules point at each individual record. Arthur Guttilla · · 5 min Fundamentals The anatomy of AI Mouth, eyes, ears and hands have already been bought. The brain is the part still missing, and it is what makes the other three get things right. Arthur Guttilla · · 7 min Architecture Data that never becomes training: what changes in integration design With no copy into the model, the integration retrieves at query time. The impact on latency, caching policy and data that changes often. Arthur Guttilla · · 5 min Architecture The graph engine decision: cost and capacity at volume Two graph engines under the same 300M-element target: load time, RAM, disk, latency and the AWS bill. The cutover line between in-memory and hybrid. 
Allan Fraga · · 12 min Context quality Switching models rarely fixes it: what changes and what stays the same A pricier model writes better, but it still answers from the same context it always had. The ten-minute test that tells a model problem apart from a data problem. Arthur Guttilla · · 5 min Architecture Business ontology: how AI understands your company's vocabulary The same company becomes three different customers to the model if it exists three times, in the ERP, the CRM and the helpdesk. The layer that resolves entities, verbalizes the relationship and gives the data meaning. 
Matias Schimuneck · · 5 min Architecture Company Brain: shared memory across your AI agents Each agent keeps its own memory and learns on its own what the others already know. The graph, the ontology and the semantic layer that make that learning common, served over MCP. Arthur Guttilla · · 4 min Governance and trust BYOC or hosted: where your company's AI context layer runs BYOC or vendor cloud: what changes in data, network and cost across the two deploy models for an AI context layer. Understand the concept. 
Marcos Ribeiro · · 5 min Fundamentals The context layer will follow the same path as the data warehouse Infrastructure components follow a known pattern: special project, competitive edge, standard item. The context layer is halfway through it, and the data warehouse already showed where it ends. Arthur Guttilla · · 5 min Fundamentals Data lake or Company Brain A data lake solves storage. Connectors, ontology, memory and governance solve context. Why gathering data is not the same as giving AI context. Arthur Guttilla · · 9 min Context quality The sources left out of the pipeline that carry the business rule Contracts, internal policy, team spreadsheets and support history rarely make it into the analytics pipeline. That is where the rule that never became a table lives. Arthur Guttilla · · 5 min Governance and trust Where your data lives when the company uses AI What leaves the environment on each query, what the provider retains and which topologies exist. The map that unblocks AI projects in regulated industries. Arthur Guttilla · · 5 min Context quality Why your company's AI nails the easy question and misses the one that matters Summarizing an email works, prioritizing a customer does not. The three signals that separate a lack of context from a model limitation. Arthur Guttilla · · 5 min Architecture Source citation and audit are an architecture requirement, not a UI one A control built into the screen stops holding at the first consumer that arrives over the API. Why citation and logging need to be born in the layer that retrieves the data. Arthur Guttilla · · 5 min Fundamentals What context engineering is Given a finite context window, what is the best combination of information for the task at hand. Source selection, compression and ordering, and long-term memory. Arthur Guttilla · · 4 min Fundamentals What is a Company Brain? Giving an AI access to your systems does not make it understand your company. What enterprise context is: what the company knows, how the work actually happens and what is allowed. 
Rafael Viana · · 11 min Nothing matches those filters. Clear the selection to see everything.
