Whitepaper

Ten agentic squads are not scale

BRQ's proposal for reusing capabilities across squads, tools and platforms.

Download the whitepaper

Free. Fill in your details and get instant access. The whitepaper is in Portuguese.

Author Fabio Gobbi CTO of Architecture and Generative AI at BRQ

View on LinkedIn for Fabio Gobbi (opens in a new tab)

Diagnosis

What changes between the second and the tenth squad

Some squads already work well with agents. With ten, the problem changes in nature.

  1. Each squad redoes what another has already done

  2. The context that matters never reaches the agent

  3. No one knows what each agent can access

Contents

What you'll find in this whitepaper:

By the tenth squad, what is most worth reusing is governed access to systems and the eval that tells whether the work came out right. Specialist agents should be the exception.

  1. Four hypotheses about the challenges of scaling agentic AI.
  2. A reference architecture based on governed capabilities.
  3. Criteria for quality, cost and access control.
  4. A proposed experiment to compare local and shared solutions.
  5. Conditions for deciding when to expand, adjust or stop the approach.

“The question I would ask a year from now: in the tenth squad, how much of what it uses already existed in the catalog, and how much of that would the model alone have done just as well?”

Fabio Gobbi Fabio Gobbi on LinkedIn (opens in a new tab) CTO of Architecture and Generative AI, BRQ

Download the whitepaper