Automating Bank Reconciliation with Microsoft Copilot Studio and Financial Solution

Automating Bank Reconciliation with Microsoft Copilot Studio and Financial Solution

How the Financial Solution MCP Server turns bank reconciliation into a guided, conversational workflow — step by step

Bank reconciliation is one of those processes every finance team runs, yet it still eats up hours of manual effort — scrolling through transactions, hunting for matches, and chasing down discrepancies. What if your ERP could walk you through it conversationally, handle the straightforward matches automatically, and only ask for your input when there is a genuine decision to make?

That is exactly what Microsoft Copilot Studio delivers when connected to Financial Solution through the Financial Solution MCP Server. In this post, we walk through the complete five-step flow: configuring the agent, selecting the company and bank account, reviewing automatch results, and confirming ambiguous transactions — all inside a single chat experience.

How the Flow Works

The bank reconciliation agent follows a structured, five-step workflow:

1. Configure the agent with bank reconciliation instructions

2. Start the conversation and select the company

3. Select the bank account to reconcile

4. Review automatch results — auto-skipped, auto-matched, and ambiguous transactions

5. Confirm ambiguous matches and continue in rounds

Each step narrows the scope progressively — company → bank account → reconciliation — so the agent stays focused and the user always knows where they are in the process.

Step 1: Configure the Agent Instructions

The first step happens in Copilot Studio itself. Open your agent and add the bank reconciliation instructions to the Instructions section. These instructions define the complete workflow the agent will follow: how to ask the user for their company, which MCP tools to call against Financial Solution, how to organise match results into categories, and when to escalate to the user for a decision.

Once saved, the agent follows this workflow automatically whenever a user asks to process bank reconciliation. There is no additional coding required — the instructions act as the agent’s playbook.

Figure 1 — Copilot Studio agent overview with bank reconciliation instructions configured

MTC Tip: Write your agent instructions as if you are onboarding a new team member in your finance department. Be specific about the order of operations, what data to present, and what to do when matches are ambiguous. Vague instructions lead to unpredictable agent behaviour.

Step 2: Start the Conversation — Select Company

The user kicks things off by typing something like process bank reconciliation in the chat. The agent does not guess the company. Instead, it calls the `GetCompanies` tool via the Financial Solution MCP Server and presents the full list of available companies for the user to choose from.

This is critical in a multi-entity environment — every downstream step (bank accounts, reconciliations, transactions) is scoped to the selected company. Getting this wrong would mean reconciling against the wrong data set entirely.

Figure 2 — Copilot lists available companies and asks the user to select one

Step 3: Select the Bank Account

After the user selects a company, the agent calls `GetBankAccounts` for that company and presents the available bank accounts. The user picks the account they want to reconcile.

The agent keeps interactions clean and user-friendly — the user only sees business names they recognise, not internal IDs, GUIDs, or system references. This is a deliberate design choice that keeps the conversation natural and reduces the chance of errors.

Figure 3 — Copilot lists bank accounts for the selected company

Step 4: Review Automatch Results

This is where the agent does the heavy lifting. It fetches unreconciled bank transactions and match suggestions from Financial Solution, then organises them into three clear categories:

Auto-Skipped

Transactions with no matching journal entries. These are set aside for this round — the user can revisit them later or investigate manually.

Automatically Matched

Transactions with exactly one match. The agent matches these automatically without requiring user input. This is where the bulk of the time savings come from.

Multiple Matches Found

Transactions with more than one possible journal entry. These need user review because the agent cannot determine the correct match on its own.

The agent presents a compact summary showing what was handled automatically and what needs the user’s attention — no scrolling through hundreds of transactions to find the ones that matter.

Figure 4 — Automatch summary showing skipped, matched, and ambiguous transactions

Figure 5 — Multiple match details with the reply format the user should follow

Step 5: Confirm Matches and Continue

For ambiguous transactions, the user simply replies with their selection. For example:

JE-2593 → JE-1504 match it

The agent maps the displayed reference back to the actual match payload in Financial Solution, posts the match via the MCP Server, and reports the result. No need to open Financial Solution, navigate to the reconciliation page, or manually post entries.

The process works in rounds — the user can continue with the remaining transactions or stop for now and come back later. There is no pressure to solve the entire reconciliation in a single session.

Figure 6 — Match confirmed, agent shows remaining transactions

Why This Approach Works

This flow is effective because of four key design principles:

Progressive scope narrowing — Company → Bank Account → Reconciliation. Each step reduces the problem space, so the agent and the user stay focused.

Automatic handling of clear matches — The agent only involves the user when there is a genuine decision to make. Straightforward one-to-one matches are processed automatically.

Conversational interface — No switching between screens or navigating complex menus. The entire workflow happens inside a single chat thread.

Round-based processing — Users are not forced to complete everything at once. They can process what they can, stop, and return later.

The result is a bank reconciliation process that is faster, more controlled, and fully explainable — the user can see exactly what the agent did and why at every step.

The Technology Behind It

Component	Role
Microsoft Copilot Studio	Hosts the AI agent, manages the conversation flow, and executes the instructions.
Financial Solution MCP Server	Connects Copilot Studio to Financial Solution using the Model Context Protocol (MCP).
Financial Solution	Source of truth for companies, bank accounts, transactions, and journal entries.
MCP Tools	GetCompanies, GetBankAccounts, and transaction matching APIs called by the agent.

The Model Context Protocol (MCP) is an open standard that allows AI agents to interact with external data sources and services. The Financial Solution MCP Server implements this protocol specifically for Financial Solution, giving Copilot Studio agents native access to Financial Solution data and operations.

Getting Started

If your organisation runs Financial Solution and you want to bring AI-assisted automation to your finance workflows, bank reconciliation is an excellent starting point. The process is well-defined, repetitive, and benefits immediately from intelligent matching and user-guided decision-making.

To set this up, you need:

1. Microsoft Copilot Studio — available as part of your Microsoft 365 or Power Platform licensing

2. Financial Solution MCP Server — configured and connected to your Financial Solution environment

3. Agent instructions — tailored to your organisation’s bank reconciliation workflow

At MTC, we help organisations across multiple sectors deploy Copilot Studio agents connected to Financial Solution. If you would like to explore how this could work for your finance team, get in touch with us.

Conclusion

Bank reconciliation does not have to be a tedious, manual grind. With Microsoft Copilot Studio and the Financial Solution MCP Server, you can turn it into a guided, conversational process that handles the routine matches automatically and only asks for human judgement when it genuinely matters.

The five-step flow — configure, select company, select bank account, review automatches, and confirm ambiguous entries — keeps the process structured, transparent, and efficient. And because it works in rounds, your team can fit reconciliation into their workflow without blocking out large chunks of time.

AI is not replacing your finance team — it is giving them a smarter way to work.

Want to see more real-world examples of Copilot Studio in action with Dynamics 365? Check out our complete guide to Microsoft Copilot Studio or contact MTC to discuss your automation goals..