Working with the resolution plan
How auto-processing determines the level
Prerequisites, inputs and examples for Off, With approval and Automatic
Tip
For email and letter tickets in the resolution plan, auto-processing decides for the whole plan whether a service agent handles it (Off), whether it is prepared and released with one click (With approval), or whether enneo carries it out itself and sends the reply (Automatic). Several inputs each contribute a level – the lowest one wins.
The settings themselves are described in Types of release. This page explains how enneo derives the level of a single plan from them. How to review and confirm a plan by hand is covered in Review and confirm a resolution plan.
Prerequisites
Before any level matters, enneo checks:
- Auto-processing on is switched on.
- The ticket is open and not spam.
- The channel is email, letter or system, and the message is inbound.
- Exactly one plan that can be released is attached to the ticket.
- If the ticket is assigned to a contract, the customer is legitimated – with one exception for follow-up questions, see Customers who are not yet legitimated.
If a prerequisite is missing, the ticket is not processed automatically and a service agent handles it as usual.
The inputs
Every input that applies to the plan contributes a level – or stops the plan. Inputs that do not apply do not count.
| Input | When it counts | What it contributes |
|---|---|---|
| 1. AI completion type | Always | If the AI escalated the ticket to a service agent, the plan stops: Off. Otherwise the remaining inputs decide. |
| 2. Steps | For every step that is carried out | The level set under Level per AI agent case for the matching case. If the case has no level, the plan stops: Off. Skipped steps do not count. |
| 3. Question to the customer | The reply asks the customer something – also because an unreviewed step turns into a question | The level from Replies asking the customer a question – or, if the plan's smart AI agent has an entry there, its Reply with follow-up question level from Different levels per Smart AI agent. |
| 4. Neither step nor question | A plain answer or closing without a reply | With approval – or, if the plan's smart AI agent has an entry, its Respond without asking back level from Different levels per Smart AI agent. |
Inputs 3 and 4 are decided by the smart AI agent that created the plan – not by the AI agents of the individual steps.
The lowest level wins
The levels are ordered from low to high: Off, With approval, Automatic. The plan gets the lowest level of all contributions. A single stop is enough for Off.
So a setting can only hold a plan back, never raise it. If a step is at Automatic but the question is at With approval, a plan with a question waits for approval.
Different levels per Smart AI agent
Under Settings → AI Customization → Auto-processing, in the Rules for every ticket section, you can set how the replies of individual smart AI agents are released. An entry consists of:
- Smart AI Agent: the AI agent that creates the plan.
- Reply with follow-up question: level for this AI agent's replies that ask the customer something. Replaces Replies asking the customer a question for it.
- Respond without asking back: level for this AI agent's replies that ask nothing and carry out no step, including closing without a reply. Replaces the fixed With approval for it.
A smart AI agent without an entry follows the rules for all AI agents. The setting starts empty, so nothing changes until an entry is added. Typical use: an AI agent that asks non-legitimated customers for missing details should send its question on its own, while the questions of all other AI agents still need approval.
Note
An entry does not override the other inputs: if the plan carries out steps, their levels still count, and the lowest wins.
What the level means in the ticket view
| Level | What happens |
|---|---|
| Off | A service agent handles the ticket. The plan shows why it is not processed automatically. |
| With approval | The plan is fully prepared. A click on "Process now" carries it out and sends the reply. |
| Automatic | enneo carries out the plan and sends the reply – immediately or after the delay, which is counted in SLA service hours. |
Note
Automatic is only scheduled if every open plan of the ticket is at Automatic. "Process now" ignores the delay.
Customers who are not yet legitimated
If the ticket is assigned to a contract but the customer is only recognised and not legitimated (legitimation level 10 to 19, see Customer Validation), the following applies in addition to the level:
- Only follow-up questions run automatically. A service agent handles a plain answer or closing without a reply – even if a higher level is set for the AI agent under Respond without asking back.
- No steps for legitimated customers. Steps of AI agents that require a customer (Only for authenticated customers) are never carried out automatically. The plan is then at Off.
- No stored data in the question. If the reply contains customer or contract data the customer did not write themselves, enneo does not send it, and the ticket goes to a service agent. This check of the reply text is a safeguard, not a substitute for a suitably configured AI agent.
- Limited attempts. Once the customer has used up the attempts to be identified automatically, the ticket goes to a service agent.
What the customer is missing: The smart AI agent that creates the plan sees in the customer identification result which details the legitimation rules still require – one bullet per failed matching group, naming the details only, never stored values:
Missing for legitimation:
• Contract number or Customer number
• 2 out of Delivery address, First name, Last nameDetails the customer cannot supply in a reply – for example a sender email address that does not match the customer profile – are marked "cannot be met by a reply". Service agents see the same list in the warning on the contract.
Only data the AI agent may read: A smart AI agent only gets the customer and contract data if the built-in customer data tool is selected for it. An AI agent that should only ask non-legitimated customers for details does not get this tool – then it cannot mention stored data.
A typical case:
- A customer emails their meter reading and only mentions their contract number. The legitimation rules also require the meter number, so the customer stays at legitimation level 10.
- A smart AI agent set up for this asks them for exactly that detail.
- If Reply with follow-up question is set to Automatic for this AI agent under Different levels per Smart AI agent, the question goes out without a service agent.
- When the customer replies with the meter number, enneo checks the legitimation again. Once the customer reaches level 20, their request is processed normally under the rules on this page.
An example agent for this: Legitimation agent.
Examples
Assumed settings: Replies asking the customer a question = With approval; under Different levels per Smart AI agent, Reply with follow-up question = Automatic for the "Legitimation agent"; case "Meter reading recorded" = Automatic; case "Statement recorded" = With approval.
| Situation | Contributions | Level |
|---|---|---|
| Legitimated customer, one step "Record meter reading" | Step: Automatic | Automatic |
| Legitimated customer, meter reading and statement | Step: Automatic, step: With approval | With approval |
| Legitimated customer, meter reading, the reply also asks for the reading date | Step: Automatic, question: With approval | With approval |
| Legitimated customer, plain tariff information without a step | without question: With approval | With approval |
| Non-legitimated customer, the "Legitimation agent" asks for the meter number | the agent's question: Automatic | Automatic |
| Non-legitimated customer, the Base agent asks for the meter number | question: With approval | With approval |
| Non-legitimated customer, the reply is a plain answer | only follow-up questions run automatically | Service agent handles it |
| Non-legitimated customer, a step would change contract data | step requires a legitimated customer | Off |
| Any customer, the case of a step has no level | stop: AI agent does not execute automatically | Off |
| Any customer, the AI escalates to a service agent | stop: completion type | Off |