The customer has signed. Sales marks the opportunity as won. A message arrives in the onboarding channel, and the next Digital Employee begins preparing the account.
Then somebody notices that the start date was conditional.
In this hypothetical healthcare sales case, the agreement covers employees in several regions. Most services are standard. One location requires an additional arrangement, and the commercial team has promised to confirm it before activation. Sales has the detail. Onboarding has a summary saying “contract complete”. Both teams can appear to be making progress while the customer’s expectation sits between them.
The weakness is in the transfer of responsibility.
A Digital Workforce combines people and Digital Employees. Its usefulness depends on whether that combined organisation can carry a commitment through changes of role, system and owner. The handover is where that claim becomes testable.
A message is only the beginning
The receiving role needs to know what has been agreed, which source establishes it, what remains conditional and what it is authorised to do next. It also needs a way to reject an incomplete transfer without losing the case.
Technical interoperability helps information move. The Agent2Agent protocol provides a framework for agents to discover capabilities, exchange work and communicate results. A business still has to specify when an obligation has been accepted and who is accountable while acceptance is pending.
Consider the difference between “onboarding received the payload” and “the onboarding lead accepted responsibility for activation on the agreed conditions”. The first is a delivery event. The second changes how the company must behave.
A useful handover therefore has an explicit acceptance step. The current owner remains accountable until the receiving role accepts the defined scope. There can be collaboration before that point, but there should be no gap in responsibility.
Carry the decision, including its limits
The temptation is to send everything. A long conversation history feels safer than a short summary because the missing detail must be somewhere inside it.
The receiving role then inherits a reconstruction task. It must infer which statement became the decision, which earlier version was superseded and whether a promise was authorised. More context can increase the burden when its status is unclear.
Start with the current commitment and link back to the evidence. Carry unresolved conditions as structured obligations with an owner and due time. Keep the supporting discussion available for investigation without requiring every receiver to interpret it afresh.
In the healthcare example, the agreement is signed, but activation of the additional location remains blocked. Those two facts can coexist. Reducing them to a single “complete” label destroys information the next role needs.
A worked handover record
This proposed record illustrates the transfer between sales and onboarding. The references and events are hypothetical.
| Field | Worked example |
|---|---|
| Case and transfer | Corporate account HC-104; transfer H-02 from sales to onboarding. |
| Commitment | Activate the agreed service package for the listed customer entities. The additional location requires written confirmation before inclusion. |
| Authoritative evidence | Signed agreement version 3; approved offer version 5; commercial exception decision C-17. |
| Receiving scope | Create the account and prepare standard locations. Do not activate the conditional location. |
| Outstanding decision | Sales manager to confirm the regional arrangement before the customer’s agreed confirmation deadline. |
| Current accountable owner | Sales manager until onboarding explicitly accepts H-02. |
| Acceptance condition | Onboarding confirms access to the evidence, the correct entity list and the blocked location. |
| Response deadline | Four business hours, with escalation to the onboarding lead if no response arrives. |
| Duplicate protection | Repeated receipt of H-02 must reference the same case and cannot create a second account or activation request. |
| Recovery route | Rejection returns the case to the sales manager with a reason; existing customer commitments remain visible. |
On a successful transfer, onboarding records acceptance of that scope. Ownership of account preparation changes. Ownership of the unresolved commercial decision remains with the sales manager. A transfer need not pretend that every obligation moves together.
Now test a rejected transfer. Onboarding finds that the signed agreement names a different legal entity from the proposed account record. It rejects H-02 with the reason recorded. Account creation remains blocked. Sales corrects the record and submits H-03, linked to the original case. Onboarding accepts the new version after checking the correction.
The business has experienced a delay, but it has retained a coherent account of the work. That is a better foundation for recovery than two systems each claiming their own task is complete.
Design the uncomfortable paths
A demonstration tends to follow a cooperative sequence. The sender provides a valid package, the receiver responds promptly, and the next action succeeds.
The evaluation should include a missing source, an expired approval, a receiver that is unavailable, and a repeated delivery after a timeout. Test what happens when a customer changes the request after acceptance. Check whether a failed downstream action produces a new decision or merely another attempt at the same mistake.
An acknowledgement can also be lost after the receiver has already acted. Before retrying account creation, the system needs to establish whether an account already exists. A repeated message must not become a repeated commercial consequence.
These cases deserve attention because coordination itself can introduce failure. Anthropic’s August 2026 research describes coordination problems and conflicting objectives in experimental multi-agent settings. Those experiments do not establish an enterprise failure rate. They do challenge the assumption that adding capable agents will automatically produce a more capable organisation.
The relevant test is the shared result under the conditions in which roles must cooperate.
Keep the mechanism proportionate
There is a fair objection: formal handovers can become bureaucracy. A three-field internal research request should not require the ceremony appropriate to activating a contractual service.
Scale the mechanism to the consequence. For a low-risk, reversible task, receipt and a simple acceptance response may be enough. For a customer commitment, transfer the applicable limits and retain the evidence needed to resolve disagreement. The objective is to make the next decision easier.
The same principle applies to people. Requiring a manager to approve every routine transfer can turn that manager into the bottleneck the Digital Workforce was supposed to relieve. Give receiving roles a clear acceptance standard and delegate within it. Escalate failures of that standard and decisions that exceed their authority.
Observe the pattern over time. Frequent rejection for the same missing field suggests a problem in the upstream role. Frequent timeouts may reveal insufficient receiving capacity. Neither problem is solved by making the handover animation smoother.
The customer experiences one company
Outcome1.AI’s Digital Workforce should be judged at the boundaries between roles as well as within them. A role can produce accurate work and still contribute to a poor customer experience if its output cannot be accepted by the next owner.
This extends the Frontier argument about agentic authority. Authority has to remain intelligible as work moves. The next role must inherit the relevant limits, and a human owner must remain accountable for the business commitment.
For the customer, the sales channel and onboarding channel are parts of the same company. The promised service either becomes available on the agreed terms or it does not. A Digital Workforce earns trust when its internal transfers preserve that simple external reality.