how-to

What KPIs Should You Track for an AI Marketing Agent Program?

The Orbitable Team·AI Agent Practice·19 May 2026·6 min read

Track two categories: business-outcome KPIs your marketing function already reports (pipeline sourced, cost per qualified lead, campaign conversion) and operational KPIs unique to running an agent program (credits consumed per deliverable, first-pass approval rate, cycle time from request to approval). The business-outcome numbers tell you if marketing is working; the operational numbers tell you if the agent program specifically is the reason, because they measure how much each output cost to produce and how much of it survived stakeholder review without being sent back for rework.

Why generic productivity metrics miss the point

Most advice on measuring AI marketing tools defaults to "time saved" or "content volume produced". Both are guesses, because they assume someone is manually logging hours against a project tracker. An agent program that meters its own work removes the guesswork. In Orbitable, credits meter every action a specialist agent takes, and the Dock (the review surface where a client or internal stakeholder sees what is waiting, marks it up with pins, sends it back, or approves it into a Library) records every approval and every rework cycle. That gives you two signals a spreadsheet of estimated hours cannot: the exact cost of a deliverable, and whether it was right the first time.

Credits per deliverable

Credits meter the work agents do; reviewing work never costs credits, so this number reflects production cost only, not review overhead. Calculate it as total credits spent by a squad or the whole fleet, divided by the number of finished deliverables, over a set period (weekly or monthly is enough resolution).

Watch the trend, not the absolute figure. A rising credits-per-deliverable number usually means more revision cycles are going into each output, either because briefs are vague or because the quality bar has moved up. A falling number is what you want to see over time, and it has a specific mechanical cause in Orbitable: every agent shares one world model per customer (ICP, brand voice, products, competitors, uploaded knowledge base). Once that context is established, agents are not re-deriving brand voice or competitor positioning from scratch on every job. Work compounds. If credits per deliverable is not falling after the first month or two, the world model is probably incomplete or out of date, and that is worth fixing before you fix anything else.

First-pass approval rate

This is the number the Dock gives you natively. A deliverable lands in the Dock, the stakeholder either approves it into the Library or sends it back with pins marking specific issues. First-pass approval rate is the percentage of deliverables approved without a single round of pins.

It matters because it separates two failure modes that look identical from the outside: an agent producing weak work, versus a brief or world model that sets the agent up to miss. A low first-pass rate concentrated in one squad points at the agent or the brief template. A low rate spread across every squad points at the world model, most often stale ICP or brand voice information that every agent is inheriting.

Cycle time from request to approval

Requests raised in the Dock carry SLA targets. Cycle time is the gap between a request being raised and it being approved. Track it against the SLA target as a simple percentage: what share of requests closed inside their target this week. This is the metric that tells finance and leadership whether the program is actually faster than the process it replaced, which is usually the first question anyone outside marketing asks.

Traditional KPIs you still need

Operational signals prove the program is running efficiently. They do not on their own prove marketing is working. Pair them with the KPIs you already report:

Pipeline sourced from campaigns the agents produced, pulled through your HubSpot or Attio integration
Cost per qualified lead, calculated the way you already calculate it, with credits spend added as a cost line
Conversion rate on agent-produced assets versus your prior baseline, if you have one to compare against

Do not replace these with credits and approval rate. Use them together. A program can have an excellent first-pass approval rate and still be producing content that does not convert, which is a targeting or positioning problem, not a production problem.

Comparing the two metric types

Dimension	Traditional marketing KPI	Orbitable operational signal
What it measures	Whether the output performs in market	Whether the output was produced efficiently and accepted first time
Data source	CRM and campaign platforms (HubSpot, Attio)	Credits ledger and Dock approval history
Typical cadence	Monthly or quarterly	Weekly
What a bad number tells you	Positioning, targeting, or channel problem	Brief quality, world model gaps, or agent-fit problem
Who usually owns it	Marketing leadership	Whoever runs the agent program day to day

Building a simple weekly view

Autopilot proposes a weekly plan on paid plans: one focus, three to five missions, each with a why-now, a carry-over list, and an explicit not-doing list. It proposes only; approval happens in the Dock. Use that weekly mission list as your reporting unit. For each mission, log credits spent, whether it was approved first pass or sent back, and how long it took from request to approval. Roll that up monthly against pipeline sourced and cost per qualified lead. You now have a dashboard built entirely from data the program already generates, with nothing estimated.

FAQ

What is credits per deliverable and why does it matter?

Credits per deliverable is total credits consumed by agents divided by the number of finished deliverables in a period. It matters because it is the only exact, non-estimated production cost figure you have, and its trend tells you whether the shared world model is compounding value across jobs or whether context is being rebuilt every time.

How is first-pass approval rate calculated in Orbitable?

It is the percentage of deliverables approved in the Dock without being sent back with pins for rework. A low rate concentrated in one squad usually points at that agent or brief template; a low rate across every squad usually points at gaps in the shared world model.

Do traditional marketing KPIs still matter if I track credits and approval rate?

Yes. Operational metrics prove the program runs efficiently; they do not prove the output converts. Track pipeline sourced, cost per qualified lead, and conversion rate alongside credits per deliverable and first-pass approval rate.

Does reviewing a deliverable cost credits?

No. Reviewing work never costs credits in Orbitable, so first-pass approval rate is a clean measure of production quality that is not distorted by review activity.

Which plan includes the Dock for tracking these metrics?

The Dock is included on the Agency and Enterprise plans, and is available as a $49/mo add-on on the Founder and Team plans. Autopilot's weekly plan, which pairs well with this reporting, is available on all paid plans.