The conversation around enterprise AI has understandably concentrated on intelligence:
- Which model gives the best answer?
- Which one reasons most effectively?
- Which assistant produces the strongest summary?
Those questions matter. But an AI assistant can produce an excellent answer and still leave the user doing too much work to get there.
That distinction becomes clearer when AI is used inside real business workflows.
Consider a relatively simple task: summarise the decisions made across several meetings and turn them into a project update.
The AI may be perfectly capable of completing the task.
But, before it can start, somebody may need to find the relevant transcripts, locate a supporting document, download files, move between applications, upload everything into the AI tool and explain which information matters.
The response itself may take seconds. The workflow surrounding it can take considerably longer.
That’s why an independent enableUC benchmark of Zoom AI, ChatGPT, Microsoft Copilot, Claude and Google Gemini looked beyond model output alone. It evaluated ten practical knowledge-work scenarios across response relevance, time to insight, user experience and cost.
The quality results were remarkably close
Zoom achieved an average accuracy score of 4.7 out of five. ChatGPT, Gemini and Copilot each scored 4.6, while Claude scored 4.5. Completeness followed a similarly narrow pattern.
The much larger gap appeared in time to insight.
Across all ten scenarios, Zoom completed the tasks in 100.2 seconds of measured response time. Once the benchmark’s standard allowance for retrieving required source material was applied to the other tools, Gemini reached 346.05 seconds, ChatGPT 410.3 seconds, Claude 495.6 seconds and Copilot 529.8 seconds.
The point is not that other AI platforms are incapable. The research found that all of them generally produced strong answers.
Zoom’s advantage came largely from context
The benchmark describes Zoom AI’s ability to work with relevant information from meetings, documents, chats and email through its first-party capabilities and built-in connectors. Zoom My Notes can also capture meeting context from Zoom, third-party meeting platforms and in-person conversations.
Less time was therefore spent assembling the information required to ask the question.
That has wider implications for how businesses evaluate AI
Employees already spend significant portions of their working day navigating between communications, documents and business applications. Adding another isolated AI interface simply introduces another destination into that journey.
The operational value of AI increases when it sits closer to the work.
That might mean being able to ask about a meeting without finding the transcript first. It could mean creating an update from information already available across several conversations.
In customer service, the same principle applies when an AI assistant can access the customer, order or account information required to complete a task rather than forcing an agent to search for it manually.
Context is what turns general intelligence into useful assistance
There is also a cost consideration. Businesses increasingly have employees using several AI tools with overlapping functionality. If one platform already supports common summarisation, drafting, meeting and knowledge tasks, additional subscriptions can create duplicated spend alongside the workflow complexity.
The benchmark itself cautions that results will vary according to configuration, permissions and product versions, and it should be read as a comparison of practical scenarios rather than a universal model ranking.
Its broader lesson is nevertheless valuable.
When businesses assess AI, the quality of the answer is only part of the equation. They should also measure how much work it takes to give the AI enough context to produce that answer.
Because every minute spent finding, exporting and re-uploading information is productivity the model itself cannot give back.
Want to understand where AI could remove friction from your existing workflows?
