How to make IQ agent insights sound like your team

IQ agents can draft the performance commentary and insights that usually eat up your reporting time. On default settings, they write accurate but generic notes because they do not yet know your client or your voice. Giving an agent three things: client context, a defined tone, and a review step before anything reaches the client, is what turns its output into commentary that reads as if your team wrote it.

An agent only knows what you give it. Client context lives in the agent’s Context panel, which has two tabs that work differently. Knowledge is the files you upload for the agent to read. Memory is the standing facts the agent records as you work, and it keeps them in mind on every run.

Step 1. In the left navigation, open IQ Agents and select the agent you want to tune.

Step 2. Open Agent settings, then open the Context panel.

Step 3. Upload the material that defines the client: their brand or style guide, a few recent reports, and examples of commentary your team wrote by hand. The agent reads these when they are relevant to the task. If the agent has web access, you can also point it at the client’s website, and it reads that itself.

Step 4. Teach it the standing facts. In a conversation, tell the agent what to keep in mind, for example, which KPIs the client cares about, a naming convention, or an account quirk. The agent saves it to Memory, which it carries into every run from then on, with nothing to publish.

Step 5. Check what it kept. On the Memory tab, you can review every fact the agent has saved and delete anything wrong or out of date. You add to Memory by telling the agent, not by typing here.

Tone is a standing rule about how the agent writes, so you set it the same way you teach it a fact: you tell it, and it keeps the rule in Memory. Memory is read on every run, so the rule shapes whatever the agent writes into a report or a text widget thereafter.

Step 1. In a conversation with the agent, describe how the commentary should read, for example: short, direct sentences, lead with the result and then the cause, name the metric that moved and why, and no marketing jargon.

Step 2. Give it examples to match. Paste one or two short pieces of handwritten commentary into the conversation and ask the agent to write in that style. A concrete example anchors the voice far better than a description of it. For a longer reference, such as a full style guide, upload it to the Knowledge tab instead.

Step 3. The agent saves the rule to Memory, and it applies from the next run with nothing to publish. Review or remove it any time on the Memory tab.

Have the agent hand you its commentary in the chat before any of it goes into the report. You read and shape it there, and only the version you approve gets added. This keeps a person between the agent and the client, which is the point of an agent that works alongside you, not instead of you.

Step 1. In the IQ agent conversation, tell it to write the commentary in the chat for your review and to hold off on adding it to the report until you say so.

Step 2. Refine it with the agent. Read the draft, reply with the changes you want, and go back and forth until it reads the way your team would write it.

Step 3. Approve it into the report. When you are happy, tell the agent to add the approved text to the report. You can also ask it to remember this way of working, so it drafts in the chat and waits for your approval on every run.

Step 4. For an agent that sends or publishes on a schedule or trigger, add an approval gate. Tell it to require your approval before it sends anything. The run then pauses at an Approval needed prompt, which also collects under Activity, and you can route it to a colleague with Teammate who gets your approvals.