Article reader Listen + reading controls
Article reader
Preparing the reader…
Reading settings
Copilots will rewire the handoff¶
Microsoft has introduced generative artificial intelligence capabilities across Dynamics 365, bringing “copilot” functions into sales, customer service, marketing, and supply-chain work. The examples emphasize drafting emails, summarizing interactions, creating content, and surfacing information inside the applications where people already work.
The most important design question is not how much text a copilot can generate. It is what happens to the handoff.
Work is held together by handoffs¶
Enterprise processes look orderly in diagrams. In practice, they depend on people transferring incomplete context: a salesperson briefs an account team, a service agent escalates a case, a planner explains a disruption, or a manager reviews a draft before it reaches a customer.
Those moments carry more than data. They carry judgment about what matters, what remains uncertain, which relationship is sensitive, and what the next person should notice first. Much of that knowledge is tacit.
Generative artificial intelligence (AI) can improve a handoff by gathering records and producing a coherent summary. It can also flatten the handoff by making incomplete data sound complete. A polished paragraph may obscure the difference between a documented fact, an inferred pattern, and the experienced employee's concern.
The copilot changes who prepares the work¶
When a system drafts an email or summarizes a customer interaction, the human role shifts from author to reviewer. That sounds like a small efficiency change. Human-factors research suggests otherwise.
Bainbridge's “Ironies of Automation” describes a recurring problem: automation handles routine work while leaving people responsible for monitoring and recovering from the difficult cases. The remaining human task can require more expertise even as repeated practice declines.
For copilots, the reviewer may need to detect a subtle omission in a fluent draft without having reconstructed the case personally. If the interface rewards quick acceptance, the organization can gain speed while losing situation awareness.
The answer is not to require people to redo all the work. It is to design the review around the uncertainty that matters.
Preserve provenance and friction selectively¶
A good enterprise copilot should make it easy to distinguish three layers:
- Source facts: records, messages, transactions, and policies the system retrieved.
- Generated synthesis: the model's summary, draft, or recommendation.
- Human commitment: the approval, modification, escalation, or decision made by the accountable person.
Those layers should remain connected. A reviewer ought to reach the underlying source for an important claim, see what information was unavailable, and record a meaningful correction. That correction should improve the workflow's knowledge—not disappear after the message is sent.
This is where knowledge management and product design meet. Carlile's work on knowledge boundaries shows that transferring knowledge across functions becomes harder as novelty and differences in interpretation increase. A generated summary may help with simple transfer, but complex cases still require people to translate and negotiate meaning.
Measure the health of the handoff¶
Productivity metrics alone will miss the effect. A team can produce more messages while escalating the wrong cases or losing critical context. Organizations should track:
- acceptance and editing patterns by task type;
- corrections that change a decision rather than only wording;
- downstream rework and reopened cases;
- whether users can identify the source behind important claims; and
- whether junior staff continue learning the domain rather than only reviewing prose.
Begin with a bounded handoff where both sides can help define success. Let the receiving team describe what information actually supports action. Let experienced practitioners identify the cues that are absent from structured records. Build those requirements into the copilot and its evaluation.
Copilots will make many enterprise interactions faster. Their lasting value will depend on whether they strengthen the transfer of understanding, not merely the production of language. The handoff is where work changes owners—and where accountability must remain unmistakably human.
Sources and research trail¶
- Microsoft, “Introducing Microsoft Dynamics 365 Copilot” (March 6, 2023).
- Microsoft Dynamics 365, “Bringing Next-Generation AI to Every Line of Business” (March 6, 2023).
- Bainbridge, “Ironies of Automation” (1983).
- Carlile, “A Pragmatic View of Knowledge and Boundaries” (2002).
- Endsley, “Toward a Theory of Situation Awareness in Dynamic Systems” (1995).