# When the enterprise assistant you already own is enough.

> Your company may already pay for tools that can answer its knowledge questions. Identify the actual gap before funding a custom search and AI application.

Canonical URL: https://marmation.com/review/when-copilot-or-glean-is-enough/

By Vikram Roy · 2026-11-12

A new custom assistant can duplicate a capability the business already pays for. Before commissioning it, test the existing platform on real tasks and identify the gap that would justify another system to maintain.

An IT leader is asked to commission an internal AI knowledge assistant. The business already pays for an enterprise productivity platform with built-in search, knowledge connectors and user access. An external developer proposes rebuilding the same experience with a bespoke retrieval stack.

## Include the work a platform already provides

The proposal may be unnecessary. Packaged knowledge tools often provide connectors, administration, policy integration and managed upgrades that custom teams would have to recreate. The question is not whether a custom retrieval-augmented generation (RAG) system, which finds company information for a model to use in answers, can work. It is whether it adds value after operating cost and risk are included.

## Test requirements against the existing service

Evaluate the actual user tasks, source coverage, permission behavior, latency, data policy, action integrations and total licensing costs. Run representative tests in the existing platform. Record which requirements remain unsatisfied, not just which custom features are technically possible.

## Compare ordinary search with a specialized approval task

A business may need an assistant to answer ordinary policy questions, in which case a configured enterprise tool could be sufficient. Another may need a specialized approval interface integrated with multiple internal APIs and document-level exceptions. Custom engineering may be justified in the second case.

## When the gap earns a custom build

Where a packaged product has restrictive licensing, inadequate source handling or a poor UX for the task, building can be rational. But the burden of proof sits with the custom proposal.

## Ask the custom proposal to prove its value

Before buying custom engineering, document the commercial product’s tested limitations and the operational advantages of a bespoke solution. Treat the purchase decision as an architecture problem, not an identity statement.

### References and context

[Microsoft Copilot Studio knowledge sources](<https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/knowledge-sources-overview>); [AWS Bedrock knowledge bases](<https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html>).

## Sources

- [Copilot Studio knowledge sources](<https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/knowledge-sources-overview>) — Microsoft Learn. Existing enterprise platforms already provide usable AI knowledge capabilities. Limit: Capabilities, licensing and the treatment of source permissions depend on the selected product and configuration.
- [Knowledge bases for Amazon Bedrock](<https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html>) — AWS Docs. Managed retrieval and knowledge base infrastructure available, making simple RAG less differentiated. Limit: Features, charges and regional availability depend on the configuration and should be checked for an actual deployment.

## Related reading

- [The choices behind technology investments](https://marmation.com/review/decisions/)
- [Build, buy, extend or integrate? The AI decision that comes before the technology.](https://marmation.com/review/build-buy-extend-or-integrate/)
- [How to evaluate an AI engineering partner.](https://marmation.com/review/choosing-an-ai-engineering-partner/)