Client Report AI Summaries: AgencyAnalytics, Whatagraph, DashThis

The short answer: for an agency sending ten clients a monthly report, AgencyAnalytics puts an AI summary on every report and holds each one for approval at $25 per client, so 10 x $25 = $250 a month on monthly billing (AgencyAnalytics pricing, checked on 26 October 2026); DashThis Professional covers 10 dashboards for $164 a month, with Claude writing the recap (DashThis pricing); Whatagraph starts at $812 a month billed annually and suits a roster big enough to need its agents (Whatagraph pricing). What it saves is writing time, not reading time. For example, if commentary takes 40 minutes per client by hand and reviewing a draft takes 15, ten clients go from 10 x 40 = 400 minutes to 10 x 15 = 150, and the account managers get 250 minutes back each month.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

The three tools agree that a person should approve the text. They disagree on where that approval sits, and that decides more than any feature list: one holds the whole report at the moment it would send, one gates the agent before it writes and the report before it leaves, one gates the writing and then mails whatever the dashboard shows on the day.

Before the first AI summary reaches a client:

1. Price the roster the way each vendor counts it. AgencyAnalytics charges per client, DashThis per dashboard and source, Whatagraph per credit, and its pricing FAQ says each source costs one. For example, ten clients with GA4, Google Ads and Meta each is 10 x 3 = 30 sources: inside DashThis Professional's 40 and inside the 50 credits Whatagraph's Max plan starts from (Whatagraph pricing).
2. Mine the commentary you already wrote. Take the last three months of notes sent to each client. They give the agent two things it cannot see in the data: the client's goals and what happened that month, such as a budget move or a site launch.
3. Write the rule for what a summary may claim. Figures only from widgets on the page, for the page's own period. No causes unless the account manager supplies them. No total of conversions across ad platforms.
4. Count this month's cycle by hand. Minutes of commentary per client, and the questions clients sent back about the numbers. Those two are the baseline.

Three places to put the approval

AgencyAnalytics: the gate is at send time. The AI Summary widget summarises a page, a dashboard section or the whole report, and reads only the widgets on it, Google Sheets widgets included; it cannot read uploaded PDFs or documents (AI Summary). So budgets and targets go into a Sheet widget if the summary is meant to mention them. With Require Approval on the schedule, each report waits at its send time until someone opens it from the Inbox, edits what needs editing and clicks Approve and Send. In the Reports table, the Awaiting Approval tag counts every pending report and lists them for review or rejection (scheduling article), which is the queue an account lead works through on the first of the month.

One interaction needs care. The AI Summary FAQ says toolbar edits are reset whenever the summary auto-refreshes, and that on a scheduled report auto-refresh runs for whoever opens the summary first after it is sent. Read together, a figure corrected by hand during review can be regenerated before the client reads it. Put corrections into the Guide your summary prompt, which is reapplied on every refresh, or switch auto-refresh off on that widget before approving.

Whatagraph: two gates. The Insight agent answers questions about a report and drafts a client-ready summary into a text widget (pre-made agents). Before any agent changes a report or widget it shows an impact warning and asks for confirmation, and agents can run on a schedule, for example refreshing a set of reports before a client call (IQ Agents). The second gate is the automation: with I'll review before sending selected, the report does not go out until someone confirms the review, and Do not send reports until issues are reviewed holds it when a source has a problem (automation article). IQ Agents are for admins and managers; editors do not get them.

DashThis: the gate is at write time. Claude, ChatGPT or another connected assistant writes the recap into a comment box, and nothing lands on the dashboard until you approve it; if the box already holds a comment, it shows you that comment and asks before replacing it (AI comments article). The scheduled email then sends the dashboard as it looks on the day. On a fixed-period dashboard a comment carries forward until a later period gets its own, so a skipped month sends last month's story beside this month's numbers. For many dashboards, the same article says an n8n automation can do the writing with a service key that has read and write access.

The account manager's check

The DashThis article puts the limit plainly: the assistant sees the numbers, not what is behind them. The check therefore covers both halves.

• Every figure in the text appears on the page, for the same period. DashThis notes that very long widget tables may be shortened when the connector reads them (MCP limitations), so a top-ten list in a summary can be a top ten of the rows it received.
• Direction words match the comparison. "Up" against the previous period and "up" against last year are different claims; AgencyAnalytics compares a custom range with the same number of days just before it (comparison article).
• Rates are read with their counts. A higher click-through rate on fewer impressions can be a smaller campaign rather than a better one.
• Platform conversions stay separate. Google Ads counts only conversions linked to its own ad interactions in the same account (Google Ads data discrepancies), and Meta credits events inside its own click and view windows (Meta attribution settings). A summary that adds the two has invented a number the client will repeat in a board meeting.
• Causes come from the account manager. "Conversions rose after the new landing page" is a claim about the world. If the summary says it, the person approving it should know it.

Free first

Nothing here needs an AI subscription to start. The account manager writes four lines per client from the finished report: what moved, against what, why if known, and what happens next. On DashThis, every plan includes AI Insights Free with four preset readings, a Summary among them (DashThis pricing), which is a usable first draft at no extra cost. Each of the three offers a trial: 14 days on AgencyAnalytics with no card (AgencyAnalytics pricing), 10 dashboards for 14 days on DashThis, and a 14-day Max trial on Whatagraph (upgrade article). Run two cycles of handwritten notes beside the drafts during the trial, and the comparison is made before anything is bought.

Before switching on AI drafting, read the client contracts. The summary feature is one more system the client's data passes through, and some contracts say which systems may see it.

What ten clients cost

Checked on 26 October 2026 on the pricing pages of AgencyAnalytics, DashThis and Whatagraph, all in US dollars.

Whatagraph sells IQ Agents as teammates that carry work out on their own, and its Activity panel, where agent approvals collect, offers the shortcut to that: the person who started a run can always allow an action so the agent stops asking. Only that person can approve, while anyone who sees the request can deny it (Activity article). Always allow turns the agent into the hands-off teammate the product pages describe, and it removes the one moment a person reads what goes into the client's report, so leave it off for anything that writes text.

Did it work?

Measure three things per cycle: minutes from report ready to report sent, corrections the account manager made per summary, and client replies that question a number. The baseline is the hand-counted cycle from the list above.

Then split the roster for two cycles. Half the clients get drafted summaries, half get handwritten notes, chosen by alternating down the client list rather than by who is easiest. If the drafted half saves minutes without more corrections or more questions from clients, extend it. If corrections stay high on particular clients, the context those clients need is missing from the brief, and that is a fix to the input, not to the approver.

Per-tool walkthroughs are in the how-tos: monthly reports in AgencyAnalytics, blended Google Ads and Meta spend in Whatagraph and paid media across platforms in DashThis. Plans and limits are in the entries for AgencyAnalytics, Whatagraph and DashThis.

FAQ

Which tool lets one person clear every client's approval in one place?

AgencyAnalytics, through the Awaiting Approval tag in the Reports table, which lists all pending report instances to review or reject (scheduling article). Whatagraph gathers agent approvals in Activity; DashThis approvals happen in the chat, one dashboard at a time.

What happens if nobody approves in time?

In AgencyAnalytics the report waits for review, and in Whatagraph a report set to review is not sent until someone confirms it. The DashThis email does not wait for a recap: it shows the dashboard as it looks on the day, with the last recap carried forward on a fixed-period dashboard (AI comments article).

Can the summary mention the client's budget or targets?

In AgencyAnalytics, only if they sit in a widget on the page, and a Google Sheets widget works (AI Summary). In DashThis, put them in the request as context; the article suggests the client's goals, what happened that period and notes from the last meeting.