Agency

A retainer needs evidence, and visitor questions are the cheapest evidence you have.

Agencies lose retainers in the month where nothing visible shipped. An assistant on a client site produces a continuous record of what visitors could not find, which is both a deliverable and an argument for the next piece of work.

Before and after on a agency siteThe same visitor question, handled by the current site and then by an assistant grounded in that site's own content.BEFOREClient asks what the retainerbought this monthAgency shows uptime, smallfixes and an analyticsdashboardNothing in it says whichcontent is failingThe next proposal is aredesign, because that is theonly visible optionAFTEREach client site runs itsown assistant on its owncontentEvery unanswered question islogged by topic per clientThe monthly report names thefive things visitors couldnot findThe next piece of work is aspecific content fix withevidence behind it
  • Private launchOnboarding selected teams now
  • Built by CreoglyphSeven client engagements behind the product

The retainer conversation nobody enjoys

Three months into a retainer the client asks what they are paying for. The honest answer is usually maintenance, small fixes and availability, all of which are real and none of which photograph well.

The work that would justify the retainer, improving pages that are failing, is hard to propose because there is no evidence about which pages are failing. Analytics tells you a page has a high exit rate. It does not tell you the visitor was looking for the returns policy and could not find it.

So the proposal becomes a redesign, which is expensive, slow, and produces the same site again for the same reason, because the underlying question of what visitors actually need was never answered.

What changes

Today

  1. Client asks what the retainer bought this month
  2. Agency shows uptime, small fixes and an analytics dashboard
  3. Nothing in it says which content is failing
  4. The next proposal is a redesign, because that is the only visible option
  5. The redesign reproduces the same gaps

With an assistant

  1. Each client site runs its own assistant on its own content
  2. Every unanswered question is logged by topic per client
  3. The monthly report names the five things visitors could not find
  4. The next piece of work is a specific content fix with evidence behind it
  5. The retainer has a visible output that is not a redesign

Running several client sites without them bleeding into each other

One agent per client

Separate agents, separate sources, separate tone. A client's assistant must never retrieve from another client's content, which is why agent count rather than page count is the limit that matters to an agency.

Train on their content, not yours

The assistant speaks for the client. Source selection is per client and usually smaller than the agency expects, because a focused set answers better than a complete one.

Report per client

Content gaps by topic, refusal rate, and the questions that escalated. That is a monthly deliverable that writes itself.

Turn gaps into scoped work

A ranked list of unanswered questions is a scoped content brief. It is the difference between proposing a redesign and proposing five pages with a reason attached to each.

The three ways a conversation leaves the flowMost conversations end in a grounded answer. The exits are a logged refusal, an escalation to a person, or a capture before continuing.IN THE FLOWQuestionRetrieveGrounded answermost conversationsEXITSRefuse, log gapEscalateCapture, then continue

What to put in the monthly client report

Four things, and none of them require a dashboard the client has to learn.

Conversations answered from existing content

The count, and the share that your current pages already covered. That is the baseline and it should trend up as the content improves.

The top five refused topics

Ranked by frequency. This is a content roadmap that came from the client's own visitors rather than from a keyword tool, which makes it far easier to get signed off.

Escalations, and whether they should have happened

An escalation that a content fix would have prevented is billable work with a clear before and after.

One question quoted verbatim

Not a metric. A real sentence a real visitor typed on their site. It does more for retainer renewal than any chart, because it is the client's own customer saying what is missing.

Where the data goes

Each agent is scoped to the sources you give it. Creobot has no cross client data sharing feature, because it has no mechanism to share anything between agents. Security and data boundaries.

Questions

Removing the Creobot badge is planned for the Pro plan and above. Full white labelling beyond that is not built and is not being promised.

Agents are the limiting factor: one on Free and Starter, two on Pro, five on Scale. An agency running more than five sites is a case we would rather discuss than guess at, and additional agents are on the list of add ons being considered.

No. Each agent retrieves only from its own sources. There is no shared index and no mechanism to move content between agents.

The client, in any arrangement worth having. This is worth settling in your own contract before launch rather than after, and it is one of the questions our security page is explicit about not having fully answered yet.

That is between you and the client. Pricing is per account rather than per seat, so the arithmetic is straightforward, and pricing is agreed with you before anything starts.

Then the assistant will refuse most questions, and the refusal log becomes the case for the content work. That is a legitimate first deliverable rather than a failure.

Scale is the closest fit today. Whether there is a dedicated agency tier is still open.

Signal

Is this your situation?

Access is by waitlist, demo request or public-content pilot. Describe the question your visitors keep asking and we will tell you honestly whether it fits.