How we’re using AI to support Business Central implementations

AI is changing how we approach Microsoft Dynamics 365 Business Central implementations.

Rather than treating AI as a separate tool, we’re exploring where it can support the work our consultants already do. This includes capturing workshop information, preparing documentation, converting data and supporting development and testing.

Reducing manual and repetitive work through AI allows our consultants to spend more time where they add the most value: solution design, problem solving and understanding client requirements.

Starting with workshops

One of the first areas we’ve been testing and applying AI at Fenwick is during implementation workshops. Workshops involve a lot of information. Consultants are listening to client requirements, documenting decisions, identifying follow-up actions and considering potential solutions at the same time.

We introduced Microsoft 365 Copilot into this process to support our consultants before, during and after workshops. During each session, a facilitator captures a transcript of the conversation. Our consultants continue taking their own notes, including key decisions, requirements and client terminology.

After the workshop, these sources are brought together:

• Consultant notes
• Workshop transcript
• AI-generated summary

The consultant reviews and validates the information before it progresses any further. This provides another way to check what was discussed and identify follow-up items that may have been difficult to capture while facilitating the workshop.

Building the first draft

Once the workshop information has been validated, AI can support the preparation of the first draft. Using Fenwick’s existing document structure as context, Copilot can organise the information into the required format and prepare elements such as decision and action registers.

This reduces some of the manual work involved in creating and formatting documentation from scratch. In our initial application of this workflow, we completed the work approximately 33% faster than estimated.

That time can instead be directed towards areas where consultant expertise matters most, including solution design, enhancement specifications and data conversion. The important distinction is that AI produces the first pass, not the final result. Our consultants still review the document, validate its accuracy and remain responsible for what is delivered to the client.

Applying AI to data conversion

Data conversion is another area where Fenwick is testing the use of AI. Preparing client data for Business Central can involve large spreadsheets that need to be cleaned, mapped and matched before anything is imported.

Much of this work is repeatable and can be guided by explicit rules, making it well suited to AI. For example, an AI agent could complete a first pass of matching customer and contact records across a large dataset, while flagging records it cannot confidently match for consultant review. This allows consultants to focus on exceptions and decisions that require judgement rather than manually working through every record. Fenwick is testing this approach one dataset at a time before considering its broader application.

Keeping people in control

AI can produce convincing results even when the underlying information is incomplete or incorrect. Recent examples of AI-generated reports containing fabricated information demonstrate what can happen when AI-generated content isn’t adequately reviewed.

We’re conscious of these risks. Human review remains an important part of how we use AI, with our consultants validating AI-generated work against their Business Central expertise, client requirements and implementation decisions.

Fenwick also has an internal AI policy that guides the responsible and ethical use of AI across our business. This provides a framework for how our teams use AI while maintaining appropriate oversight and accountability. AI can provide a stronger starting point and reduce manual work, but our consultants remain responsible for the final result.

Moving towards agentic workflows

Fenwick’s current use of Copilot generally involves consultants prompting AI for a specific task and reviewing the response.

Agents provide an opportunity to take this further. Rather than providing the same instructions, templates and context for repeatable tasks, agents can be designed around specific parts of Fenwick’s Business Central implementation methodology.

Workshop documentation

Workshop documentation is one area where agents could support Fenwick’s implementation process. With access to the required document structure and context, an agent could use workshop transcripts and consultant notes to prepare a first draft for review.

A similar approach could be applied to data conversion. Once rules have been tested across enough scenarios, an agent could use them to prepare data and flag exceptions that require consultant review.

This creates a clear division of work: AI can handle volume, while our consultants retain judgement.

Where agents could go next

There are other stages of a Business Central implementation where this approach could support our teams. Enhancement specifications are one example. An agent could use validated workshop notes, transcripts and previous specifications as context to prepare a first draft of a technical specification.

Development could follow a similar model. AI could help establish an initial object structure, naming conventions or other repeatable components before a developer begins the more complex work.

Testing and user acceptance testing (UAT) also present opportunities. An agent with access to the relevant requirements and specifications could prepare initial test scenarios or UAT checklists. Consultants and developers could then review these against the actual solution before they’re used.

Across each scenario, the principle remains the same: AI provides a starting point, while our people validate and own the final output.

Building AI into our methodology

AI and agents have the potential to support more of the Business Central implementation process as these capabilities continue to develop. The next step is moving beyond individual Copilot prompts towards agents designed around repeatable implementation tasks.

Technologies such as Copilot Studio and Business Central’s Model Context Protocol (MCP) create further opportunities for agents to interact with Business Central data and processes directly.

As AI becomes more capable, knowing where and how to apply it becomes increasingly important. The focus remains on using AI where it can improve efficiency, while our consultants continue to retain responsibility for the decisions, solution and final result.