GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
At a glance
Global Advisors replaced a legacy content environment with a modern, self-managed knowledge and collaboration platform. The immediate requirement was safer, more usable document management. The more consequential objective was to create a governed content layer that could later support search, retrieval, contextual chat and agent-assisted work without duplicating the document estate or weakening access controls.
This was not a file-copy exercise. It required choices about information architecture, identity, permissions, coexistence, metadata, migration sequencing, user adoption, integration boundaries and the role of the content platform in a wider AI operating model.
Maturity: active operational use, with continuing improvement of retrieval, metadata and management controls.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
Challenge
The legacy estate held useful institutional material but reflected an earlier operating model. Content storage, collaboration, internal applications and search had grown as separate concerns. Documents were difficult to treat as reusable knowledge objects, and adding AI directly to each application would have created parallel indexes, inconsistent permissions and multiple versions of the truth.
The business problem was therefore larger than replacing an ageing document repository. Global Advisors needed to:
- preserve continuity for users and existing content;
- establish a durable source of truth for documents and collaboration;
- make identity and access decisions usable beyond the user interface;
- create integration seams for future retrieval and workflow services;
- support incremental migration rather than a risky single cutover; and
- ensure that AI-derived experiences remained subordinate to source permissions and records.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
The architectural response
We treated the new platform as the authoritative collaboration and document experience, not as the place where every future capability had to run. That distinction allowed the user-facing environment to remain stable while search, retrieval, model access and workflow automation evolved behind explicit interfaces.
The target pattern separated five responsibilities:
- Content authority. The collaboration platform remained authoritative for files, ownership, sharing and user-visible document state.
- Identity and permissions. Access context had to travel with downstream requests. Retrieval could not be allowed to infer entitlement merely because a document had been indexed.
- Retrieval and enrichment. Indexing, chunking, embedding, ranking and contextual answer generation were treated as external capabilities with their own scaling and lifecycle needs.
- Workflow integration. Events and service interfaces provided controlled hand-off points for business processes, agents and background automation.
- Presentation. Users continued to work in familiar document and collaboration surfaces while AI capability could be added without forcing a new destination for every task.
This architecture also protected optionality. A useful business platform may outlive several generations of AI infrastructure. Stable contracts at the boundary reduce the cost of replacing either side.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
What we did
The work combined platform migration with operating-model redesign:
- assessed the legacy content estate, user patterns and dependencies rather than treating all stored material as equivalent;
- established the successor platform and its identity, sharing and collaboration model;
- migrated in stages so that live work could continue and difficult content could be dealt with explicitly;
- preserved coexistence where immediate replacement would have increased operational risk;
- designed the content layer to expose stable integration points for search, retrieval and workflow automation;
- separated the document system of record from computationally intensive AI services;
- built downstream retrieval patterns that maintained user or collection scope;
- added retry, deduplication and backpressure behaviour for large ingestion runs;
- made metadata quality, exclusions and re-indexing policy explicit operating concerns; and
- documented what was operational, what remained transitional and what should be retired later.
Migration also exposed adjacent platform questions. Communication services, internal application pages and document collaboration share identity, governance and continuity concerns even when they are different systems. Renewing one core layer created an opportunity to simplify roles elsewhere in the estate rather than perpetuating legacy boundaries.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
Difficult problems we had to solve
Permissions are a data problem, not a front-end feature
A retrieval service can return an excellent answer and still be wrong if the caller should not have seen the source. The system therefore had to translate the collaboration platform's user context into explicit downstream scope. Where the source contract did not carry rich group or tag semantics, we treated that as a real design gap rather than assuming the retrieval layer could reconstruct policy safely.
Large ingestion loads behave differently from interactive use
Bulk indexing can saturate embedding workers, queues and databases while ordinary document use still looks healthy. We introduced bounded concurrency, retryable busy responses, timeouts, skip windows and content-hash deduplication. These were not performance embellishments; they were necessary to keep scanning recoverable and to prevent background ingestion from degrading interactive work.
A migration can create a new monolith
Moving everything into one modern platform does not automatically improve architecture. We resisted placing retrieval logic, heavy processing and business orchestration directly inside the content application. The platform owns the experience and records; specialist services own specialist workloads.
AI readiness depends on content discipline
Poor filenames, uncontrolled duplicates, ambiguous ownership and weak metadata become more damaging when AI systems amplify them. The migration reinforced that knowledge architecture must be improved alongside technology. Retrieval cannot compensate indefinitely for an unmanaged source estate.
Controls and assurance
The operating design used several layers of control:
- source permissions remained authoritative;
- downstream retrieval was restricted to caller-relevant collections or scopes;
- excluded content types and unsuitable material could be kept out of ingestion;
- document hashes and cache state reduced unnecessary reprocessing;
- retry and backpressure preserved recoverability under load;
- application, retrieval and model layers could be observed and changed independently; and
- transitional dependencies were recorded rather than hidden behind a false claim of complete migration.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
Results
The migration created a modern operational content and collaboration layer, but its larger result was architectural. Documents could now participate in governed search and AI-assisted interaction without making the AI system the owner of the records.
The firm gained:
- a cleaner platform boundary around content, collaboration and sharing;
- a practical route from documents to permission-aware retrieval;
- reduced dependence on one application's embedded search or AI implementation;
- a reusable integration pattern for future user surfaces and agents;
- stronger understanding of ingestion economics, concurrency and metadata quality; and
- an incremental migration model that delivered value before every legacy dependency was removed.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
What we learned
AI readiness is not a feature switch. It is the result of clear authority, usable metadata, portable permissions and stable interfaces. The most important platform decision was not which retrieval technology to adopt; it was to prevent document authority, AI computation and user experience from collapsing into one inseparable system.
We also learned that migration value compounds. The full strategic value of a renewed content platform became visible only when later retrieval, chat and agent workflows began using it as a governed source layer.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
Why this matters for leaders
Organisations evaluating AI-enabled document platforms should ask more than whether a product offers summarisation or chat. They should ask:
- Where does content authority sit?
- How are source permissions enforced after indexing?
- Can retrieval and model infrastructure change independently?
- How will bulk ingestion be isolated from interactive work?
- What coexistence period is realistic?
- Which metadata and ownership problems must be solved before migration?
Those questions determine whether the platform becomes a durable knowledge foundation or simply another application with an AI button.
GA AI Case Study – Migrating to an AI-ready knowledge and collaboration platform
Note
Global Advisors does not perform technical AI implementation or systems integration for clients. However, we have worked on architecting and implementing AI at a deep level in our own business since the beginning of 2024. This allows us to provide grounded AI strategic and architectural advice based on a deep hands-on knowledge of AI. We work with clients to build strategies, business and operating models to win in an AI enabled world. We help them make architectural and partner choices for implementation and work with them to change their businesses in response.
