For CIOs, CDOs & IT Leaders
Your data doesn't move. We build intelligence on top of where it already lives.
OpsGrid runs on your own infrastructure and reads Business Central with your BC credentials and permission model. Its decision log sits in a database on that infrastructure, used by your company only. QHub Wiki runs entirely on your own server. Both products are designed for organisations where data sovereignty is non-negotiable.
Data architecture
What OpsGrid reads, where it reads it, and what it never touches.
How OpsGrid accesses Business Central
OpsGrid connects to your BC tenant using the BC API and the service account credentials you provide. It reads from the operational tables you configure, nothing more. Data is retrieved at query time to generate a decision or answer, then released. There is no persistent data copy, no ETL, and no external warehouse.
Write operations are strictly bounded. OpsGrid can only write to the specific action types you explicitly enable, and every write requires a human approval event in Teams before anything is committed to BC.
Architecture at a glance
- Connects via BC API using credentials you provision
- Reads from configured tables only: purchase, inventory, production, sales
- Read-only by default; writes require explicit configuration and human approval
- No data extracted to an external store: retrieval is query-time only
- Teams integration via webhook: no additional infrastructure
- All actions logged to an audit channel you specify
Integration model
Three integration points. All of them use infrastructure you already have.
OpsGrid is built for the BC schema and permission model, not a generic middleware layer. It connects to Teams using standard webhook infrastructure.
BC-native integration
OpsGrid is built for the BC schema, not a generic ERP abstraction. It uses BC's API, respects your existing permission model, and works with the table structure BC already uses. No schema mapping, no custom connectors, no middleware.
Teams via standard webhook
Decision cards and approval requests are delivered via Teams webhook. No new Teams apps, no additional licences, no changes to your Teams tenant configuration. Uses the notification and approval mechanisms Teams already provides.
Self-hosted, open source
QHub Wiki is fully self-hosted: WikiJS, Qdrant vector store, and the LLM connector all run on your own server. Apache 2.0 licence. Auditable codebase. No vendor-managed infrastructure involved in document retrieval or query answering.
What you control
Every configuration decision is yours.
OpsGrid is auditable at every layer. You define the scope, the permission boundaries, the routing logic, and what gets logged.
BC credentials and permissions
You provision the service account OpsGrid uses. You set the BC permissions. OpsGrid operates only within the scope you define; it cannot escalate its own access.
Which decision types surface and to whom
You configure which operational events trigger a decision card, which Teams channels receive them, and which users or roles see each type. No decisions surface outside the channels you specify.
Decision owner by decision type
You name one owner per decision type. Business Central keeps its own approval limits, approver chain and delegation, and OpsGrid leaves them alone. Backup owners and escalation are on the roadmap.
Audit log: every action, every outcome
Every decision surfaced, every approval or rejection, every BC write, logged to the audit channel you designate. Exportable. Retainable under your own data retention policies.
Compliance and sovereignty
Built for environments with strict data governance.
Four commitments to your data governance team
For organisations where a CIO or CDO must sign off on any AI deployment, these are the guarantees that matter.
Human-in-the-loop by design
$0 in BC writes without explicit human approval. The approval event is a prerequisite, not a courtesy: the write pathway is blocked until a named user approves in Teams.
LLM sees retrieved chunks only
The language model receives only the data retrieved for one query, through your own API key. It has no standing access to Business Central. What the provider keeps follows its API terms; current deployments use Anthropic.
Customer-hosted today
QHub runs entirely on your infrastructure. So does OpsGrid: one engine per customer, on your servers or your cloud tenant, never shared. The LLM call goes to the provider you choose. A managed option and a self-hosted LLM are on the roadmap.
Open source and auditable
QHub is Apache 2.0. The full codebase is auditable by your security team. No proprietary black boxes in the document retrieval or query answering pipeline.
Common questions
What CIOs and IT leaders ask us first.
Does any data leave our Business Central environment?
Some does. OpsGrid runs on your own infrastructure and reads Business Central with your BC API credentials. Each decision is recorded in a database on that same infrastructure, used by your company only. To draft a recommendation or answer, OpsGrid sends the relevant BC data to the LLM provider whose API key you supply. That provider's terms decide what it keeps. Every current deployment uses Anthropic. A self-hosted LLM option is on the roadmap.
What Business Central permissions does OpsGrid require?
OpsGrid uses your existing BC permission model. Read access covers the operational tables you configure: purchase, inventory, production, and sales are the standard set. Write access is scoped only to the specific action types you enable and requires explicit human approval before any BC write occurs.
Can we run everything on our own servers?
Yes, apart from the LLM. QHub Wiki is fully self-hosted: WikiJS, Qdrant vector store, and the LLM connector run on your own server. OpsGrid runs the same way today: one engine per customer, on your infrastructure, with the decision log in a database there. It is not SaaS. The one outbound call goes to the LLM provider whose API key you supply; a self-hosted LLM option is on the roadmap. A managed service, where IQ hosts the engine, is also on the roadmap, and there too you would supply your own LLM key.
What LLMs does OpsGrid use and do they see our data?
You choose the LLM, and you supply its API key. OpsGrid sends that provider only the BC data a given recommendation or answer needs. The provider's API terms decide what it keeps and for how long, so read them before you pick one. Every current OpsGrid deployment uses Anthropic. A self-hosted LLM option is on the roadmap; until it ships, the provider whose key you supply processes each query.
How does OpsGrid integrate with our existing Microsoft Teams setup?
OpsGrid connects to your existing Teams tenant using a standard webhook. No new Teams infrastructure is required. You specify which channels receive which decision types during configuration. The integration uses Teams' existing notification and approval mechanisms.
What happens to our data and configuration when we stop using OpsGrid?
OpsGrid keeps no copy of your BC data beyond the decision log, and that log sits in a database on your own infrastructure. Revoking the service account and removing the Teams webhook ends all data access. IQ holds no customer data to delete, because OpsGrid never ran on our infrastructure.
From our insights
See all insights →Talk to us about your architecture requirements.
We walk through your BC environment, data governance constraints, and Teams setup, and map exactly what deployment looks like for your organisation.
Request an architecture overview →