SUPPORT WRITING GUIDE
How to evaluate a support writing assistant
A manual trial worksheet for Gorgias teams: record wording quality, whether suggestions were used and what would justify continuing with Vellix.
Copy this manual trial worksheet
- Trial setup: name the routine draft types, participating agents, review date and person who will decide whether to continue.
- Decision rule: write the wording improvement you want to see and which recurring problems would make you stop.
- Draft purpose: use a local reference and a short description of the task. Keep customer messages and contact details out of this worksheet.
- Wording observation: note one specific improvement, no useful change or a problem introduced. Check the answer separately from the tone.
- Choice: record whether you kept the original, applied the suggested replacement or wrote a different draft, with a short reason.
- Further correction: if you applied the suggestion, record any later rewrite needed to restore a fact, condition or intended meaning.
- Trial decision: continue, change the scope or stop. Cite the observations behind the decision and the types of reply you actually tried.
Choose the question before starting the trial
Decide what you want a writing assistant to help with. A useful question is specific: do its suggestions make routine next steps easier to understand while keeping the answer intact?
Choose a fixed set of ordinary draft types, a review date and a decision owner. Record each attempt, including suggestions you reject or requests that fail. A trial made up only of your favourite results will leave out reasons the tool may not fit.
Intercom describes answer accuracy and tone as separate review areas. Use that distinction in your trial: a natural tone is worth assessing, but it does not compensate for a changed answer. The worksheet and continuation rules here are editorial suggestions, not Intercom product features.
Record what happened after the suggestion
For each draft, record the purpose, your wording observation, the choice you made and the reason. Keep the categories the same throughout the trial so that the notes are comparable.
A fictional worksheet entry might read: “Purpose: explain a setup step. Observation: clearer opening. Choice: applied. Further correction: restored a condition that the suggestion removed.” This is an example of a manual record, not a Vellix result.
That entry shows why an applied replacement alone does not settle the question. You also need to know whether it preserved the answer and remained useful after review. Gorgias documents writing and sending as separate steps; check the actual outgoing draft before sending it.
Make a narrow continuation decision
At the review date, consider every recorded attempt. Look for the wording improvement you chose and any repeated corrections, failures or reasons agents preferred their originals.
For example, you might decide in advance to continue only if suggestions help with the chosen wording goal, agents choose to use them and the notes show no recurring changes to facts or conditions. Choose the rule for your team before seeing the results.
If results are mixed, name the draft types that appear useful and the ones that need more work. A small manual trial informs a workflow decision. It does not establish time savings, customer outcomes or performance across all support replies.
Sources and editorial note
Examples are fictional. Sources for the platform guidance are linked below.
- Intercom on reviewing answer accuracy separately from tone and using clear writing criteria. Checked 2026-10-20.
- Gorgias documentation on reviewing context, drafting a reply and sending it. Checked 2026-10-20.