For IT directors, systems managers, technical leads and ERP maintenance leads at companies combining Business Central with dv* or third-party extensions. A copilot that assists the internal technical team in diagnostics, light customisation, release notes reading and documentation. It does not replace the partner: it filters and accelerates day-to-day.
The internal technical department at a company running Business Central and a handful of extensions has an agenda too dense for the headcount it usually has. It has to maintain the ERP, handle functional incidents from business area users, do small AL or Power Automate customisations, generate ad-hoc reports when finance or production asks for them, maintain documentation nobody updates, read Microsoft release notes every Wave and manage permissions. And all of that with one, two or three technicians.
The problem is not just volume: much of the load is repetitive. The unusual incident that appears every six months forces the team to rediscover the pattern from zero because there was no time to document the solution properly the first time. The small customisation (add a field, generate a new report, automate a simple flow) requires remembering exact AL syntax or Power Automate shortcuts the technician used months ago and forgot. Querying Microsoft Learn means navigating a huge documentation tree.
The official Microsoft Learn documentation, public forums, Microsoft Partner KB and the technical documentation of each dv* or third-party extension are scattered. A technician loses between thirty and sixty minutes every time they have to diagnose an unusual error: look up the error code in Learn, compare with the forum, contrast with the partner KB, test the solution, validate. Multiplied by dozens of atypical incidents a month, a good portion of the team technical capacity goes into searching, not solving.
The last front is the Wave. Twice a year Microsoft publishes the BC release cycle with hundreds of pages of changes. An internal technical team can hardly read them in full, so they tackle them piecemeal: they review what looks familiar, assume the rest and discover breaking changes on upgrade day. The consequence is the usual: late surprises, urgent ticket to the partner, extra time on the customer account and nerves.
The case is a technical copilot deployed for the customer IT team. It is a conversational assistant combining four capability layers. The first is resolving technical questions on BC and extensions: the technician asks in natural language ("Why does posting fail with error X in module Y?", "How do I configure a multi-level purchase approval workflow?", "Which table stores warehouse movements between locations?") and receives a synthetic answer with citations to Microsoft Learn, to the Davisa knowledge base and to internal precedents from the customer itself.
The second layer is assisted code and configuration generation. When there is a small customisation to do, the copilot proposes AL snippets for new tables, page extensions, report objects, control add-ins; Power Automate flows for typical cases (notification, copy to SharePoint, cross-validation); DAX queries for Power BI reports; PowerShell commands for BC environment administration. The technician reviews, adjusts and applies. The copilot does not execute anything against production by itself: the human barrier is by design.
The third layer is assisted incident diagnosis. When a user reports an error, the technician pastes the log or message into the copilot, which proposes likely causes ordered by probability, based on patterns from Microsoft Learn, the Davisa KB and precedents previously resolved at the same company. For each proposed cause it suggests verification steps. It is a fast triage that reduces diagnosis time before the human verification.
The fourth layer is assisted release notes reading. Each BC Wave (October and April) the copilot automatically processes the official notes, cross-checks them with the customer real configuration (which modules they use, which extensions are active, which in-house customisations have been developed) and identifies which changes affect them, which are priority breaking changes and which deliver new functionality worth leveraging. It delivers a prioritised report so the technical department decides clearly what to review before the upgrade.
The cross-cutting layer is the integration with the customer Azure DevOps. The copilot reads the history of issues, pull requests and internal releases to understand the project context. When a change is documented, the copilot generates an automatic summary from the commit and the PR. That documentation stays in the customer internal KB and feeds future queries, closing the learning loop without anyone having to sit down to write documentation by hand.
| Technical work aspect | Before (no copilot) | After (with AI) |
|---|---|---|
| Average diagnosis time for an unusual incident | Between 30 and 60 minutes searching Microsoft Learn, forums and KB until the pattern is found. | The copilot proposes a likely cause and verification steps in seconds, with source citation. |
| AL code snippet generation | The technician writes from scratch or copies from old projects. Syntax remembered halfway. | The copilot proposes the AL snippet with the typical structure and adapts it to customer context. |
| Searching Microsoft Learn documentation | Open docs.microsoft.com, navigate the tree, read several pages to find the exact data. | Direct question to the copilot: synthetic answer with citation to the Microsoft Learn page. |
| Change documentation | Documented after the fact, badly or not at all. Next incident diagnosed from scratch. | The copilot generates an automatic change summary from the log and the PR. |
| New technician onboarding | Three to six months asking the senior. Curve depends on veteran availability. | The newcomer asks the copilot from day 1 with citations to internal customer precedents. |
| Dependency on the external consultant | Ticket to the partner for queries the team could resolve with the context at hand. | Prior filtering: the team resolves what is assistable, escalates to the partner only what requires it. |
| Reading Wave release notes | Hundreds of pages every six months. Read piecemeal or not at all. Surprises at upgrade. | Prioritised impact report on your real configuration, with recommended actions. |
Audit of the current technical environment: BC version, active extensions, in-house customisations, IT team organisation, incident management tools and existing internal documentation. Identification of the most frequent query types and selection of the priority knowledge corpus.
Copilot deployment in the customer tenant, indexing of relevant Microsoft Learn, Davisa KB, code of authorised extensions and internal precedents from the customer Azure DevOps. Integration with the IT team workflow and user training. Measurement of average diagnosis time against baseline.
Extension to more technical areas, activation of the automatic Wave report, assisted documentation generation from PRs, ongoing support and monthly retraining with new customer precedents. Coordination with the Davisa consultant for critical cases.
Some scenarios do not pay back or are not viable. We say it directly.
Same RAG pattern, but for the whole team: HR, quality, management. Non-technical queries.
BC extensionDavisa analytics layer on BC. Useful when the technical copilot generates complex ad-hoc reports.
AI sectorSector cases where the internal technical team maintains BC with critical production modules (in Spanish).
BC hubPublic documentation of the BC ecosystem, AppSource and the Davisa dv* stack (in Spanish).
MethodThe detail of the three phases (Discovery, Pilot, Scale-up) with deliverables and pricing (in Spanish).
HubFull catalogue of cases, sectors and method of the AI Studio (in Spanish).
It works with any Business Central extension. The knowledge layer includes all of Microsoft Learn (the official BC and AppSource documentation), public community forums, and the Davisa knowledge base, which covers the dv* extensions in detail but also implementation cases on third-party AppSource extensions and in-house customisations. If your company runs a typical mix of base BC plus two or three AppSource extensions plus some proprietary module, the copilot covers the lot.
Only proposes. The copilot suggests AL code snippets, Power Automate configurations, DAX queries, PowerShell commands or troubleshooting steps, but does not execute anything against the production environment. Review, validation and rollout are always done by your team. This restriction is by design: AI is an assistant, not an operator. During scale-up, assisted execution can be enabled in sandbox or development environments, never in production without double human validation.
The private code of your extensions, configurations and internal precedents is indexed with Azure AI Search inside your Azure tenant. The engine retrieves fragments relevant to the query and passes them to the language model as context. The code itself is not used to train third-party models, is not sent to public OpenAI, does not leave the perimeter of your subscription. If you work with GitHub Copilot Business on private repositories, the integration respects its contractual guarantees.
No, it complements them. The copilot resolves the band of frequent queries and quick diagnostics your technical team can handle alone, without opening a ticket. For large customisations, architecture decisions, version upgrades, problems requiring access to the partner Microsoft tenant or critical production cases, the path is still the external consultant. What changes is that when you reach the consultant, you reach them better prepared, with prior context and dismissals done. The consultant billed time drops, not the relationship.
The Business Central Waves (two a year, October and April) bring hundreds of pages of release notes that an internal technician can hardly read in full. The copilot processes those notes automatically, cross-checks them with your current configuration and active extensions, and identifies which changes affect you, which are breaking changes and which deliver new functionality you can leverage. It delivers it as a prioritised report, not a dump. This turns the Wave from a diffuse risk into an actionable technical review list.
The copilot is framed within the current relationship with your usual consultant, who coordinates the indexing of precedents and the integration with your tenant.
Talk to the team →We start with the 5-day discovery. We audit your technical environment, size the copilot and deliver a rollout plan for your internal IT team.
Request AI discovery →