A Company Brain for Your Legacy Systems

The knowledge is there. Finding it is the problem.

Every company accumulates systems. An ERP that was best-in-class when it was chosen. A CRM that the sales team relies on. A ticketing system that support lives inside. A wiki that started with good intentions and now sits half-maintained. Shared drives with folder structures that only two people understand, one of whom left the company three years ago.

Each of these systems holds a piece of the company's institutional knowledge. Contract terms live in the CRM. Product specifications sit in the wiki. Client communication history is split between email, the ticketing system, and a project management tool. Financial data is locked in the ERP. Meeting notes from important decisions are scattered across shared drives in formats ranging from Word documents to scanned PDFs.

The systems work well enough for their intended purpose. The ERP handles invoicing. The CRM tracks deals. The problem is that nobody can ask a question that spans more than one of them. "What did we agree with this client?" requires checking the CRM for the deal record, the shared drive for the contract, the ticketing system for any amendments, and possibly the wiki for internal notes. "Where's the spec for that feature we shipped last quarter?" might involve the project management tool, the wiki, the shared drive, and someone's email. These are not unusual questions. They are the kind of thing people ask every day, and the answer to each one is a manual scavenger hunt.

Connecting what you already have

Fabric does not replace your legacy systems. It sits alongside them. Through its connections framework, Fabric links to the tools your organisation already uses: CRMs, ERPs, project management platforms, cloud storage, communication tools, wikis, and more. For systems with custom or proprietary APIs, Fabric supports the MCP protocol, which means that if a system has an API, Fabric can connect to it.

Once connected, Fabric indexes the content across all of these sources and makes it available through a single semantic search layer. This is not keyword matching. It is retrieval that understands what you are looking for, even when the exact words differ between systems. You can read more about how this works in our explanation of retrieval-augmented generation.

Two things make this practical for legacy environments. First, Fabric can build connectors to new data sources, so you are not limited to a catalogue of pre-built integrations. If your organisation runs a custom-built inventory system, a sector-specific compliance platform, or a twenty-year-old database with a basic API, Fabric can connect to it. Second, these connections sync continuously. The data flowing into Fabric is not a one-time export that goes stale. It reflects the current state of each source system, which means the AI assistant's answers and the search results stay current as the underlying data changes. For legacy systems where manual exports and CSV dumps have been the only way to get data out, continuous sync is a significant shift.

The result is that all of the knowledge locked inside your legacy systems becomes searchable from one place. You do not need to know which system holds the answer. You do not need to remember the exact file name or folder. You search once, and Fabric pulls results from every connected source.

Asking questions in plain language

Search is useful, but sometimes you do not want a list of documents. You want an answer. Fabric's AI assistant lets you ask questions in natural language across all of your connected sources. "What did we agree with Acme Corp about payment terms?" will pull the relevant contract clauses, CRM notes, and any related correspondence into a single response, with citations back to the original sources.

This is particularly valuable for teams who interact with systems they did not set up and do not fully understand. A new customer service team member does not need to learn the quirks of a fifteen-year-old ticketing system. They can ask the assistant and get the answer, sourced from wherever it lives. A product manager can ask about a feature's history and get a synthesised view from the project management tool, the wiki, and the shared drive, without opening any of them.

Documentation that writes itself

One of the most common casualties of system sprawl is documentation. When information is spread across many tools, keeping a single source of truth up to date becomes impractical. People stop trying. The wiki falls behind. Process documents go stale. New starters are told to "just ask Sarah" because Sarah has been there long enough to know where everything is.

Fabric's self-writing documentation addresses this directly. These are documents that pull from your connected sources and keep themselves current. A client overview document can synthesise data from the CRM, the latest support tickets, contract terms from the shared drive, and project status from the project management tool. When the underlying data changes, the document reflects it. This is not a one-time export. It is a living layer on top of your existing systems.

This capability is especially useful for onboarding. Instead of asking new team members to learn five different systems in their first week, you can give them a set of self-writing documents that provide context drawn from all of those systems. The documents stay current, so they remain useful long after the onboarding period ends.

Multi-step lookups with agents

Some questions require more than a single search. "Which clients have open support tickets and are also up for contract renewal this quarter?" involves checking the ticketing system, cross-referencing with the CRM, and filtering by date. A person doing this manually would need to export data from two systems and compare them in a spreadsheet.

Fabric's agents can perform these multi-step lookups automatically. You describe what you need, and the agent works through the steps: querying one system, using those results to query another, and assembling the answer. This is where the value of connecting legacy systems becomes most apparent. The data in your ERP, your CRM, and your ticketing system is already rich and detailed. The difficulty has always been in combining it. Agents remove that difficulty.

For consultancies working with client data spread across multiple platforms, this means faster research and more thorough analysis. For sales teams, it means a more complete view of each account, as described in our page on sales knowledge.

Making legacy systems useful again

There is a persistent temptation to solve system sprawl by replacing everything with a single platform. In practice, this rarely works. Migration projects are expensive, disruptive, and slow. They often introduce new problems while solving old ones. And some legacy systems are deeply embedded in workflows. Your ERP may be fifteen years old, but it handles invoicing reliably, and the finance team knows it inside out.

The more practical approach is to keep the systems that work and add an intelligence layer on top. This is what Fabric provides. Your ERP continues to handle invoicing. Your CRM continues to track deals. Your wiki continues to hold process documentation, however outdated. Fabric connects to all of them and makes the collective knowledge accessible. As we explored in our post on the usability layer your business systems are missing, the problem with most enterprise software is not capability but accessibility. The data is there. It is just hard to get to.

This approach also addresses a common concern about AI adoption in organisations with established systems. There is no need to overhaul your infrastructure. You can read more about how Fabric works within existing security and governance frameworks on our AI without the risk page.

Building institutional memory

Over time, Fabric becomes something more than a search tool. It becomes a record of your organisation's decisions, agreements, and reasoning. When you connect your communication tools, project management platforms, and document stores, you create a searchable institutional memory. Why did we choose this vendor? What were the arguments for and against launching that product line? What did the board say about expansion into that market?

These questions are answerable if the information exists somewhere in your systems. The problem has always been finding it. Fabric's AI-searchable knowledge base makes that institutional memory retrievable. And because the search is semantic, you can ask the question in your own words, not in the exact phrasing used when the information was first recorded. For teams that want to maintain a structured record of key choices, the decision log capability offers a more deliberate approach to preserving this knowledge.

Your ERP has the answers. So does your CRM, your wiki, and your shared drives. As we wrote in a recent post, the issue has never been a lack of data. It has been the absence of a way to ask questions across all of it at once. Fabric provides that layer, and it does so without asking you to throw away the systems you have spent years building around.

Frequently asked questions

Does Fabric replace our existing systems?

No. Fabric connects to your existing systems and adds a unified search and AI layer on top. Your ERP, CRM, wiki, ticketing system, and shared drives all continue to operate as they do today. Fabric reads from them and makes their content searchable and queryable from a single interface.

What systems can Fabric connect to?

Fabric has built-in connections to dozens of common business tools, including CRMs, ERPs, project management platforms, cloud storage providers, communication tools, and wikis. For systems without a built-in connection, Fabric can build new connectors through the MCP protocol, integrating with any system that has an API. Once connected, data syncs continuously rather than requiring manual exports or one-time imports.

How long does it take to set up?

Connecting a system typically takes minutes, not weeks. Once a connection is established, Fabric begins indexing the content. The time required for indexing depends on the volume of data, but most organisations can start searching across their connected sources within hours of setup.

Is our data secure?

Fabric is designed to work within existing security and governance frameworks. Data permissions from your source systems are respected, so users only see content they are authorised to access. You can read more about Fabric's approach to security and compliance on the AI without the risk page.

Can Fabric handle very old or unusual file formats?

Fabric can index a wide range of document formats, including older file types commonly found on legacy shared drives. For systems with proprietary data formats, the MCP protocol allows custom integration that can extract and index the relevant content.

How does the AI assistant know which source to use?

It does not need to know in advance. When you ask a question, the assistant searches across all connected sources simultaneously and assembles the answer from whichever sources contain relevant information. You receive a response with citations pointing back to the original documents or records.

Do we need IT involvement to set up connections?

For most standard connections, an administrator can set them up without IT support. For systems with custom APIs or specific security requirements, some IT involvement may be needed to configure the MCP integration and ensure compliance with your organisation's policies.

Can different teams have different views of the same data?

Yes. Fabric respects the access permissions of your source systems and allows you to configure which connections and content are available to which teams. A sales team might see CRM and contract data, while an engineering team might see the wiki and project management data, all from the same Fabric workspace.

What happens if we eventually replace one of our legacy systems?

You disconnect the old system and connect the new one. Because Fabric continuously syncs with connected sources, the new system's data flows in as soon as the connection is live. Migrations can happen gradually without losing access to your unified knowledge base during the transition.

Can Fabric work with on-premises systems?

Yes. Through the MCP protocol and API access, Fabric can connect to on-premises systems provided they have an accessible API endpoint. This is common for legacy ERPs and internal databases that have not been moved to the cloud.

Related pages

• Connections

• Search

• AI Assistant

• Agents

• Docs That Write Themselves

• AI Without the Risk

• Onboarding

• Sales Knowledge

• How to Create a Shared Knowledge Base

• What Is Retrieval-Augmented Generation