Unsure which stack, scope, or guardrails fit your context? Talk to RIDE, our advisory AI agent: it maps your use case, recommends a stack (Azure, AWS, Google, or open-source), and drafts a plan in minutes.
A technology-agnostic guide to designing, delivering, and operating AI agents across Microsoft Azure, AWS, Google Cloud, or open-source. The same editorial discipline and people-centered posture behind GISTM.ai, TSM.ai, RIDE, and Muiraquitãna.
We do not sell stacks. We deliver agents that work. The stack is a means, not an end. What makes an agent reliable is editorial discipline, guardrails, curation cycles, and accountable humans. Whether on Microsoft 365 + Copilot Studio, Azure AI Foundry, AWS Bedrock, Google Vertex AI, or a private open-source deployment, the method remains the same.
Data Riders works where your data already lives, under your identity and governance perimeter, with your compliance team engaged from the start.
A concise, objective map of the four families we build on most often, and where each one excels.
Best fit when you already run on Microsoft 365, Teams, and Entra ID. Native grounding in SharePoint, Dataverse, and Fabric. Governance via Purview.
Ideal when your pipeline is in S3, Glue, and Redshift. Model choice across Anthropic, Meta, and Amazon. Fine-grained IAM, VPC, and KMS controls.
Perfect for stacks with BigQuery and multimodal workflows (documents + images + video). Strong grounding in Workspace and Drive.
Llama, Mistral, and Qwen on your infrastructure, for air-gapped, regulated, or cost-sensitive scenarios. Complete model and data sovereignty.
If your company already runs on Microsoft 365, Teams, SharePoint, and Entra ID, the shortest path to an agent in production is Microsoft Copilot Studio with Azure AI Foundry. Here is how Data Riders designs, delivers, and governs agents on this stack in practice.
What is Copilot Studio?
It is Microsoft's low-code/pro-code platform for building enterprise agents that converse within Microsoft 365, across Teams, Outlook, SharePoint, Word, or on external websites. It connects natively to Microsoft Graph, Dataverse, Power Platform, and models hosted in Azure OpenAI via Azure AI Foundry.
Instead of building a chatbot from scratch, you inherit identity (Entra ID), permissions (SharePoint Sites.Selected), governance (Microsoft Purview), DLP (Data Loss Prevention), and audit logging (Microsoft 365 audit log), everything your IT team already manages.
Why this matters for M365 teams
- Time to production: reduced from months to weeks, without rebuilding SSO, user profiles, or permission trails.
- Inherited governance: Purview policies, sensitivity labels, and DLP apply to the agent automatically.
- Frictionless adoption: the agent lives where users already work (Teams, Outlook, SharePoint), requiring no new tools or logins.
- Predictable cost: message-based licensing or consumption-based billing via Azure, integrated into your Enterprise Agreement.
- Data sovereignty: processing happens within the client's Microsoft tenant, with no data leaving the M365 perimeter.
Scenarios and use cases where Copilot Studio shines
Seven real-world mining and governance scenarios we deliver with Microsoft Copilot Studio.
SharePoint agent
Scenario: governance teams spend hours searching for policies, procedures, and technical reports across hundreds of SharePoint libraries.
Solution: Copilot Studio agent grounded in SharePoint Sites.Selected, responding with direct citations to the document, page, and section. Inherited permissions ensure users only see what they are already authorized to access.
Copilot inside Teams
Scenario: dam engineers need to consult GISTM, ANM 95, and internal protocols during meetings and inspections.
Solution: agent published as an app in Teams with voice and text capabilities, integrated into calendars, OneDrive, and channels. Mandatory citations on every technical response.
Intelligent triage in Outlook
Scenario: the compliance inbox receives hundreds of evidence requests weekly, creating a risk of missed regulatory deadlines.
Solution: Copilot agent + Power Automate classifies emails, applies Purview sensitivity labels, suggests responses based on policy, and escalates sensitive cases to human staff.
GISTM / TSM compliance
Scenario: GISTM and TSM audits require documentary traceability and curation, with the evidence base hosted on the client's SharePoint.
Solution: Copilot Studio agent with a RAG pipeline over curated evidence, integrated Power BI dashboards, and automated report exports for auditors.
Operational agent (Dataverse)
Scenario: operational data, work orders, inspections, and water KPIs reside in Dataverse and Power Apps.
Solution: Copilot agent that reads from and writes to Dataverse via native connectors, triggers Power Automate flows, and returns responses with inline Power BI charts.
Onboarding & Training
Scenario: new employees need to learn internal policies, the code of ethics, and safety procedures, but knowledge is scattered.
Solution: onboarding agent in Teams/Viva featuring interactive quizzes, progress tracking, and direct citations of official documents. Multilingual (PT/EN/ES).
Concrete benefits for M365 teams
First agent in production in 4 to 8 weeks, backed by full editorial scope and guardrails.
Entra ID, MFA, Conditional Access, Purview DLP, and sensitivity labels, without rebuilding anything.
Data does not leave the tenant. Azure OpenAI runs in your chosen region (Brasil Sul, EU, US) with optional EU Data Boundary.
Application Insights, Log Analytics, and Power BI: usage, quality, and cost metrics directly in the dashboards your CIO already uses.
What Data Riders delivers on the Microsoft stack
We do not sell licenses. We deliver the agent: complete, governed, and adopted.
Editorial discovery
Charter defining tone of voice, scope, refusal criteria, and operational rituals, tailored to the client's Purview policies and SharePoint taxonomy.
Tenant deployment
Copilot Studio agent deployed directly within the client's tenant, connected to SharePoint (Sites.Selected), Dataverse, Power Platform, and Azure OpenAI.
Red-teaming & evals
Scenario evaluations in PT/EN, prompt injection testing, regression tests on every knowledge base update, and a signed quality report.
Internal team enablement
Training for the client's M365 team to operate, curate, and scale the agent, eliminating long-term dependency on Data Riders.
Ready for a Copilot Studio agent?
We bring the method. You bring your tenant, your data, and your compliance team. In 4 to 8 weeks, an agent lives in Teams, Outlook, or SharePoint, governed, cited, and auditable.
The same playbook we use to build GISTM.ai, TSM.ai, RIDE, and Muiraquitãna, regardless of which cloud or model hosts the agent.
1. Editorial design
Voice document, scope, out-of-scope patterns, ritual and refusal rules. Written in clear language with the responsible human in the room, not an AI-generated prompt.
2. Knowledge curation
The knowledge base is curated, not scraped. Primary sources, versioning, provenance, and a curatorial inbox for updates. RAG over documents you truly trust.
3. Stack selection
We choose the stack after the editorial work, never before. The question is always: where does your data live, who owns the identity, and what does your compliance need to approve?
4. Safety engineering
Refusal patterns for medical/legal/financial topics, PII handling, defenses against prompt injection, grounded-only responses, human in the loop for sensitive actions.
5. Red-team & evals
Scenario-based evaluations in the user's language, red-team prompts, regression tests with every knowledge base update. No silent drift.
6. Operate & curate
Weekly curatorial cadence, privacy-preserving telemetry, named responsible owner within your organization. An agent is a living product, not a one-off project.
Live agents we operate today, each co-designed with the end client or institutional partner.
GISTM.ai
Continuous compliance with the 77 GISTM requirements. RAG over audit evidence, executive dashboard.
TSM.ai
Readiness, gap analysis, and action roadmap under MAC TSM protocols, audit-oriented.
TCM AI Assessor
AI pre-audit for The Copper Mark certification. Assesses the 33 RRA criteria with verdicts, evidence, gaps, and recommendations. Independent, demo mode, no login.
RIDE Agent
Structured discovery interview that delivers an executive Work Plan. Multilingual, no login.
AquaTwin Agent
Water balance digital twin for mining: visual modeling, document ingestion, Kalman fusion, and auditable Excel export.
Muiraquitãna
Territorial AI with voice and visual modes, community-first. Built with MuCA, launching in April 2026.
That is our favorite kind of conversation. Bring the problem; we bring the stack.
