GuideAgency services

Scope a monthly AI service retainer

Turn source updates, conversation reviews and tested workflow changes into a defined monthly service. Set a delivery allowance and use a client report to explain the work.

In this guide

What you’ll finish with

A monthly service scope with named tasks, limits, ownership and a reporting method.

Before you begin

  • A deployed or launch-ready client workspace.
  • The client’s source and follow-up owners.
  • Your delivery capacity and support hours.

List the recurring work

Review what the client’s agent needs after installation. Typical tasks include checking a conversation sample, reviewing request quality, updating approved knowledge, testing changed instructions and preparing a client update.

Distinguish maintenance from expansion. Adding a new channel, a new enquiry workflow or a new customer journey may need its own scope rather than being absorbed into a general monthly promise.

WorkKuga location
Review answers and customer questionsWebsite Chat > Conversations
Update approved sourcesKnowledge
Improve enquiry summariesActions > Collect enquiries
Check plan and volumeSubscription & Usage
Review available analysisAI Insights
Share client accessClient Portal

Set a workable delivery allowance

Choose a review frequency and define the changes included. Estimate the time required to investigate, implement, test and communicate each update. Leave capacity for issues that need more than a quick edit.

Write down the boundary between included support and separately scoped work. State your service hours and response expectations without implying that the agent itself supplies a staffed support team.

Example retainer scope to adapt

Included: one monthly conversation and request review; one client report; approved source updates and workflow adjustments within the agreed monthly delivery allowance. Client responsibilities: maintain approved information, provide decisions on changes and follow up enquiries. Separate scope: additional channels, new enquiry journeys, new external destinations or a substantial change to the service. Support: the stated service hours and agreed response process.

Create a review record

In KugaClient workspace › Website Chat › Conversations

Record the period and sample you reviewed. Group findings by the actual cause: missing knowledge, outdated content, unclear instructions, booking setup, request context or follow-up ownership.

For each change, keep the relevant conversation reference privately in your working record, note the setting changed and repeat the affected test. Avoid putting visitors’ personal details into a broad client summary.

  1. Record the period and conversations reviewed.
  2. Identify the source or workflow behind each finding.
  3. Agree and make the relevant change.
  4. Repeat the affected test and record its result.
  • The review period and sample are recorded.
  • Each finding has supporting evidence.
  • The proposed change has an owner.
  • The affected path is retested after the change.

Prepare the client update

In KugaClient workspace › AI Insights

Use the relevant available Insights report when its dates and coverage match the review. Keep the internal agency analysis separate from the client-facing report. If suitable analysis is not available, prepare the report from a documented manual review.

Explain what happened, what you changed and what needs the client’s decision. Use the client’s own records for bookings or sales outcomes. Do not treat chat activity as proof of revenue.

Review the report’s coverage before using it. The agency and client views serve different readers.

Review scope as the service grows

Compare actual delivery time, usage and support with the retainer assumptions. If the client needs more features or volume, review the client plan and the agency service separately.

Use the record to explain an adjustment. A stronger retainer proposal names the additional work and its purpose, rather than offering an undefined premium level.

Pick a client you know.

Show them their AI.

Use their website to build a demo. Put your name on it and let them try it.