Oilfield Chemical Selection with Microsoft Fabric and Copilot Studio

Smartbridge built a conversational AI solution that brings product guidance, regional sales history, and laboratory results together in one governed experience, helping an oilfield chemical services company make faster, more consistent, and evidence-backed chemical recommendations.

For oilfield chemical service providers, selecting the right production chemical can require significant technical expertise. Field conditions vary, product portfolios can include thousands of options, and the best recommendation often depends on a combination of product guidance, regional experience, laboratory testing, and years of institutional knowledge.

The Challenge: Product Selection Depended on Fragmented Data and Expert Knowledge

When a field engineer or salesperson needed to identify a chemical for a specific situation, such as CO2 corrosion, scale buildup, or bacteria control, the answer could require research across several separate sources.

The company’s product selection guide lived in Excel, regional sales history was stored in the ERP environment, and years of laboratory and field-sample results were housed in a separate lab database.

But even those sources did not tell the full story. A significant amount of product-selection knowledge came from experienced specialists who had developed expertise over years of field and laboratory work which created a bottleneck.

Formal recommendations often required senior specialists to review surveys, laboratory work, historical information, and other technical evidence before preparing a multi-page recommendation report. The quality of the process was high, but it was difficult to scale. Field and sales personnel frequently had to wait for expert availability before moving forward.

The client wanted to make more of that knowledge accessible without removing professional human judgment from the process.

From Reactive Product Selection to Evidence-Backed Recommendations

The goal was to give employees a reliable starting point much earlier in the chemical-selection process.

Instead of manually researching three to five separate knowledge sources and reviewing information line by line, users can now ask the chemical selection tool solution a question in natural language.

The agent returns a ranked list of candidate chemicals based on the category and region. Users can then refine the recommendation conversationally as more information becomes available.

Behind that simple experience is a unified data and AI architecture designed to bring several forms of evidence into the same decision.

Building the Data Foundation in Microsoft Fabric

Microsoft Fabric serves as the data foundation for the chemical selection tool. Smartbridge worked with the client to ingest, prepare, and organize the information needed by the AI agent in OneLake. This created a governed data environment supporting ingestion, warehousing, analytics, and AI without requiring a separate platform for each workload.

The product selection guide flows from SharePoint through Power Automate and Azure Blob Storage into a Fabric pipeline. Fabric notebooks transform each category into structured, prompt-ready knowledge for the agent.

Regional sales history is prepared in Fabric Data Warehouse, while laboratory and field results are curated into Fabric gold-layer tables.

Instead of asking employees to navigate those systems individually, the solution brings their evidence together behind one conversational interface.

Adding AI Orchestration with Copilot Studio

Microsoft Copilot Studio provides the conversational layer.

An orchestrator agent evaluates each question and routes it to the appropriate knowledge lane through connected sub-agents. This allows the experience to draw from different types of information while keeping each source focused on the role it plays in the recommendation.

Smartbridge also developed a custom scoring service using a Python Azure Function called through a Power Automate custom connector.

The scoring engine calculates a composite recommendation score using regional product popularity and laboratory performance. The result is a ranked recommendation grounded in the client’s own operational evidence.

The agent is available through Microsoft Teams and Microsoft 365, putting the experience inside applications employees already use which was an important factor for our client.

Explainability Built into Every Recommendation

For a technical process like chemical selection, providing an answer is not enough. Engineers need to understand why a product was recommended. That’s why our chemical selection tool was designed to show its work.

Each recommendation includes the underlying information and data used to reach the answer, giving users a way to trace the recommendation back to its supporting evidence.

Just as important, the solution recognizes when AI should not make the decision alone.

Safety-sensitive situations and scenarios outside established parameters trigger an expert-consultation path to Technical Services. The agent supports professional judgment rather than attempting to replace it. Chemical composition information also remains protected and is not disclosed through the agent.

Early Operational Impact

Because the deployment is still relatively new, long-term business results are continuing to be measured. The initial operating model, however, changes the product-selection experience significantly.

A process that previously required employees to research three to five separate knowledge sources can now begin through a single interface grounded in three unified data sources containing thousands of records and products.

Recommendations that could take days when specialist research and reporting were required can now begin within minutes.

The solution is also expected to allow roughly 50% of routine product-selection questions to be self-served by field, sales, and laboratory employees before senior specialists need to become involved. This gives technical experts more time to focus on complex situations where their experience adds the most value.

Creating a More Scalable Product-Selection Process

The chemical selection tool gives the client a new way to put years of internal data and expertise to work.

Instead of leaving product guidance in spreadsheets, historical evidence in separate databases, and critical knowledge with a small group of specialists, the solution connects those resources through Microsoft Fabric and makes them accessible through conversational AI.

Employees get faster access to evidence-backed recommendations., Technical Services remains involved when professional judgment is required, and the organization gains a more consistent approach to chemical selection across field, sales, and laboratory teams.

This is a practical example of how Microsoft Fabric and Copilot Studio can turn fragmented industrial data into an AI-assisted workflow that supports better decisions without sacrificing transparency or control.

What’s Next

This deployment represents the first phase of a broader roadmap.

Future phases can expand the number of supported chemical categories and laboratory test types, introduce more detailed formation-level geography, and extend expert-consultation coverage to additional scenarios.

As the solution grows, the same governed Microsoft data foundation can support a broader range of chemical-selection decisions while continuing to keep technical expertise and human judgment at the center of the process.

Turn Your Operational Data Into Action

Your organization’s most valuable AI use cases may already be hiding inside spreadsheets, databases, historical records, and the knowledge of your most experienced employees.

Smartbridge helps energy and oilfield services organizations bring those sources together with Microsoft Fabric, Copilot Studio, and AI to create practical tools built around real operational workflows.