FOR AGENCIES RUNNING MANY CLIENTS

AUDITABLE AI VISIBILITY
FOR AGENCY HANDOFFS

Claims that survive the client challenge, the staff change and the offboarding. A prompt register with a history behind it, an approval with a name on it, and a pack the next team can pick up.

A CLAIM IS ONLY AS GOOD AS THE RECORD UNDERNEATH IT
REGISTER · RUN · REVIEW · APPROVAL · CLAIM · EXIT PACK

When this matters

Not every engagement needs this. These four do.

The client pushes back

Somebody senior asks where a number came from. What answers them is a record rather than a recollection.

The brand is risk sensitive

Regulated, listed, or simply careful. Claims that reach a deck need something standing behind them.

The person who ran it leaves

The register, the reasoning and the caveats have to outlive the account lead.

The account transitions

An incoming team, in house or another agency, has to continue the series rather than start a new one.

What auditable means

Three things kept apart, and a record of how each of them changed.

Observation, interpretation and decision

What the engine said. What you concluded from it. What the client agreed to do. A pack that blurs the three cannot be checked.

A prompt register with a history

Not only the questions being asked now, but when they changed and why.

An approval with a name on it

Who signed off which report, and what language it permitted.

A difference between runs

What moved, what held, and what was asked differently.

An exit pack

What leaves with the client if the relationship ends.

The custody chain

Eight stages, each with somebody who owns it.

  1. Brief and scope

    client

    The client says what they want known and which brands count as competitors.

  2. Locked prompts and platforms

    shared

    Agreed together and frozen. Changes from here are versioned rather than quiet.

  3. Run and capture

    agency

    The agency runs the register and keeps the raw answers it returned.

  4. Internal review

    agency

    Somebody other than the person who ran it reads the answers before the client does.

  5. Client approval

    client

    Sign off on the report, and on what may be said publicly off the back of it.

  6. Published claim

    shared

    The deck, the page or the pitch that uses the number.

  7. Supersession

    agency

    When the register or the method changes, the old series is closed rather than quietly rewritten.

  8. Exit pack

    shared

    Everything the next team needs, handed over once.

One client cannot see another

Portfolio work fails on this before it fails on anything else.

Registers do not travel

A prompt set written for one brand belongs to that brand, and reusing it elsewhere makes both studies suspect.

Competitor frames are client specific

Naming one client's competitor inside another client's workspace is the kind of mistake that ends an account.

History stays where it was made

Runs, answers and scores belong to the workspace that produced them.

Staff move, client data does not

In Eclipse a person is assigned to the clients they work on, and each client workspace is isolated from every other.

What actually gets signed off

An approval that covers only the report is not an approval.

  • The prompt register

    The questions, and the date they were frozen.

  • The metric definitions

    What mention, citation and recommendation mean inside this engagement.

  • The report version

    Which run produced it, on which platforms, across which period.

  • The claim language

    What the client may say in a deck, and what they may not.

The exit pack

The test is whether the next team can continue the series without asking you anything.

  • The prompt register and its history

    Every version, with the date each one was frozen.

  • References to every run in the period

    Enough to find the answers again rather than take them on trust.

  • The last approved report

    The one the client signed, not the latest draft.

  • Open caveats

    What the runs could not see, written down rather than remembered.

  • The competitor set used

    Who the brand was measured against, and when that set last changed.

Where Eclipse fits, and where it does not

The record is product. The governance around it is how an agency works.

Profiles hold the frozen register

Buyer questions, competitor brands and platforms, kept per client.

A GEO profile in Eclipse, with per-platform performance across ten engines and the questions it runs

Logs are the run history

Each run by profile, platform, date and status, opening to the full answer the engine gave.

The recent executions table on a GEO profile in Eclipse, each run stamped with its platform, date and sentiment

Workspaces are the isolation

One per client, with staff assigned only to the clients they work on.

One client workspace in Eclipse, holding that client’s own positioning, brand, proof and trusted sources and nothing from any other

Approval and claim language are yours

Eclipse keeps the evidence and the history. Who signs off, and what a deck is allowed to say, stays agency process.

Common questions

TAKE ONE CLIENT THROUGH IT

Bring an account you already run. We will set up the register, the workspace and the first run on the call.