Einstein Case Wrap-Up
Einstein Case Wrap-Up is the Service Cloud feature that suggests closure-time field values when an agent is wrapping up a Case.
Definition
Einstein Case Wrap-Up is the Service Cloud feature that suggests closure-time field values when an agent is wrapping up a Case. The model reads the case description, comments, email threads, chat transcripts, and any related Knowledge attachments, then recommends values for closure fields like Status, Resolution Code, Root Cause, and similar custom picklists. The agent reviews, accepts or edits, and saves. The aim is to keep closure data clean and consistent without forcing the agent to remember every picklist option on every case type.
The feature is the closure-time sibling of Einstein Case Classification (which acts at intake) and shares the same training-data-quality dependency. Clean historical closures produce sharp suggestions; inconsistent historical closures train a model that confidently recommends the wrong values. The biggest single rollout decision is which closure fields to suggest; a focused set of three or four fields outperforms a long list that becomes noise.
Why closure-time AI is the cleanest service-team win nobody talks about
Where Case Wrap-Up lives in setup
Setup, Einstein Service, Case Wrap-Up. The page lists picklist and text fields on Case eligible for closure suggestions. For each, you toggle on, pick mode (suggested only, or pre-fill on close), and configure training data scope. The feature surfaces in the case feed close-out flow as a panel of suggested values next to the close button; the agent accepts, edits, or clears each before saving. Without enabling the panel on the Case record page in Lightning App Builder, the feature trains in the background but never surfaces to agents.
How the model learns from historical closures
The model trains on closed Cases where the target field was populated at close. For Resolution Code, it looks at every historical closed Case where Resolution Code was set, learns which case content correlates with each code value, and uses those correlations to suggest values on new cases at close time. The minimum 1,000 closed cases per value rule applies the same way it does in Case Classification. Free-text fields (Resolution Notes) can be suggested too, but text suggestions are usually too generic to accept without editing.
Closure-time vs intake-time AI
Case Wrap-Up runs at close; Case Classification runs at create. Both predict picklist fields. The difference is signal: at close, the model has the full case history (description, all comments, all email and chat content, attached articles) to read. At intake, only the initial description and subject are available. Closure suggestions are consistently sharper than intake predictions because they have more to work with. Teams running both features see compounding cleanliness in their reports, since both intake and closure fields are populated consistently rather than left blank or filled inconsistently.
Which fields belong in Wrap-Up
Three field categories work well: closure picklists with 3 to 20 well-defined values (Resolution Code, Root Cause, Defect Category), closure boolean flags (Knowledge Article Created, Escalation Required, Customer Followup Needed), and closure date fields where the value can be inferred from case events (Resolution Date, First Response Date). Free-text fields rarely benefit. Required closure fields are eligible but the suggestion does not override the required flag; agents still must accept or fill a value before saving.
Acceptance rate and the cleanup loop
Every accepted or rejected suggestion writes a signal that feeds the next retrain. Acceptance rate per field per value is the diagnostic metric. Production deployments hit 60 to 80 percent acceptance on well-defined closure picklists and 20 to 40 percent on noisy ones. Values with low acceptance are either rarely used (the model has nothing to learn) or have inconsistent historical labels. The cleanup is the same as in Case Classification: audit the 100 most recent closures with the low-accept-rate value, fix mislabeled ones, retrain. Repeat per quarter to keep accuracy from drifting.
Permissions, audit, and what shows in record history
The Einstein Case Wrap-Up permission set grants agents access to suggestions. Without it, the close panel renders without recommendations. Accepted suggestions write to the field as the agent saving the case, not as Einstein, so record history shows the agent name. The audit team should know that Wrap-Up-suggested values do not flag themselves as Einstein-suggested in field history; this matters for compliance teams who want to distinguish human-typed values from AI-suggested ones. Some orgs add a custom checkbox field "Wrap-Up Used" that the close-out flow sets, giving a separate audit signal.
Relationship to Knowledge, Reply Recommendations, and Article Recommendations
Wrap-Up sits inside a broader Einstein for Service stack. Article Recommendations surfaces Knowledge during the case; Reply Recommendations suggests reply text during chat; Wrap-Up suggests closure fields at the end. The three features compound. A case that closes with the right Knowledge article attached, the right reply patterns used, and the right closure fields populated produces clean downstream reporting on a level that any one feature alone cannot. Treat them as a related set during rollout rather than independent features evaluated in isolation.
How to roll out Case Wrap-Up so agents actually accept the suggestions
Wrap-Up's accept rate is bounded by historical label quality. Audit closures, enable on a focused set of fields, pilot for four weeks, expand. Trying to enable every closure field at once produces an overwhelming panel that agents start clearing reflexively, which trains the model to expect rejection.
- Audit recent closure labels
Pull the most recent 500 closed cases per closure picklist value. Confirm each is consistent. Fix obvious mislabels. The audit takes a few days and is the highest-leverage step in the rollout.
- Pick three to five closure fields for the pilot
Resist enabling every closure field. Three to five high-value fields produce a clean close panel; ten or more turns the panel into noise the agent dismisses without reading.
- Enable Wrap-Up in suggested mode for the pilot fields
Setup, Einstein Service, Case Wrap-Up. Toggle on. Mode Suggested. Add the close panel component to the Case record page in Lightning App Builder.
- Train the pilot team on the close panel
Brief the agents on what the panel does, how to accept or edit suggestions, and how their accept signals feed back into the model. Agents who do not know the panel exists do not use it.
- Watch accept rate per field for four weeks
Pull the Insights report weekly. Above 60 percent accept on a field, the field is working. Below 60, audit the historical labels for that value or remove the field from Wrap-Up.
- Expand to the rest of the team and add fields gradually
Roll the pilot fields to the full service team. Add one new field per month, not in batches. Per-field rollout keeps the panel focused and the accept rate measurable.
- Schedule the quarterly label re-audit
Labels drift over quarters. A quarterly re-audit of the bottom-accept-rate values keeps the model accurate and the close panel useful.
The closure fields Wrap-Up suggests. Three to five is the sweet spot; more becomes noise.
Suggested shows the value for agent acceptance; pre-fill populates the field at close time without explicit acceptance. Start with Suggested.
Where the panel renders on the Case record page. Sidebar near the close button is the convention.
Record types, date range, owner filters for the historical closures the model trains on. Useful when older closures used different labels.
Weekly by default; can be triggered immediately after a label audit or policy change.
- Enabling too many closure fields turns the close panel into noise. Three to five is the sweet spot; agents stop reading at ten.
- Inconsistent historical closure labels produce inconsistent suggestions. The audit is the highest-leverage step in the rollout.
- Wrap-Up trains in the background even without the close panel component on the Case page. If suggestions are not appearing, check Lightning App Builder before opening a support case.
- Accepted values write as the saving agent in record history, not as Einstein. Compliance teams who want a separate AI-suggested audit signal should add a custom flag.
- Free-text closure fields rarely benefit from Wrap-Up. Text suggestions are too generic to accept without editing, which defeats the time savings.
Trust & references
Straight from the source - Salesforce's reference material on Einstein Case Wrap-Up.
- Einstein Case Wrap-UpSalesforce Help
- Set Up Einstein for ServiceSalesforce Help
- Einstein Service InsightsSalesforce Help
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What does Einstein Case Wrap-Up do?
Q2. Why is wrap-up an important moment to automate?
Q3. What pairs well with Case Wrap-Up?
Discussion
Loading discussion…