John Rezabek has a great article, When to Use Control in the Field, on the use of Foundation fieldbus "Control in the Field" (CIF) capability. He concludes:

If you're leading a project, you'd like a simple and straightforward bit of guidance to help your consultant/EPC easily choose the logical applications for CIF or control in the host, such as:

• • Whenever the output of a PID block is configured to be wired directly to the AO block of a valve, CIF shall be the default choice for solving PID. The PID should be executed in the positioner, allowing the operator to run the block and, hence, the valve in "manual," even when all the other devices in the loop are being serviced.
  • HMI faceplates shall use the output of the loop's AI block to display the measured variable, so it remains a live value even when the positioner is being replaced. The device measuring the variable controlled by the PID shall always be on the same segment.
  • The remaining loops—cascade masters, for example—may be configured to run in the host with little loss of performance. Any output being controlled by 4-20 mA—for example, most variable-frequency drives—also are sensible choices for being solved in the host.

Are there other spots you would or would not recommend CIF?