Workflow
If workflow capacity is limiting
Narrow the task, reduce judgment load, strengthen inputs, or move a decision back to a named owner.
Sophizo operating diagnostic / 01
Every company will have access to the same agents. Almost none will know how many one person should run.
AI agent capacity calculator. Model review time, exceptions and agent workload before adding more automation to your team. See governed AI teams.
The model itself isn't the advantage. Every competitor can buy the same one. The advantage goes to whoever builds the system around it: the ratio formula, the phased rollout, and the governance that converts the model into measured output. Six questions. Real math. Your number.
Six questions
one accountable ratio
one accountable ratio
ASC / CONTROL LOOP
Human oversight is the constraintCapacity ÷ pressure
Agent span of control, make the ratio before the ratio makes you
Risk weights follow our R1 to R4 standard: R1 and R2 weigh one, R3 weighs two, R4 weighs four. Passive agents are R1.
After the calculation
The next move depends on the constraint, not the headline ratio.
A safe span is the lower of workflow capacity and protected review time. The diagnosis tells you which lever to pull first.
Review
If review time is limiting
Protect review hours, reduce output volume, sample intelligently, or add a qualified reviewer.
Balanced
If both are limiting
Do not add agents yet. Improve the work design and the review plan together before scaling.
90-day stopping rule: expand only if the chosen workflow holds its agreed evidence baseline for a full operating cycle. If quality, review burden, or business outcome cannot be demonstrated, pause, redesign, or retire it.
STEP 1 / 9
Which role are we sizing?
Passive or active? Count them differently.
Passive agents read, watch and alert. Active agents change records, send messages and move work forward. One person can oversee far more passive agents than active ones, because a passive error is a bad suggestion and an active error is a bad action. Tag each agent passive or active, give it a risk level from R1 to R4, and the calculator weights it for you.
How we assign R1 to R4
The 15-page executive brief
Your ratio is the number. This is the plan leaders can approve.
Get Rebuilding the Company for the Agentic Era, a concise operating brief for turning Agent Span of Control into a 90-day pilot, a defensible business case, and governance that holds when the first agent fails.
- Redesign the work, not just the tool stackSee the five operating decisions that must change together: human work, skills, incentives, authority, and rhythm.
- Pilot without betting the companyTake 3–5 low-risk agents through a 90-day rollout with an exit test and a stopping rule written in advance.
- Bring the CFO a real caseUse a transparent cost-and-value model with break-even, higher-oversight, and pricing sensitivity scenarios.
- Govern before you scaleAssign risk tiers, decision rights, escalation ownership, and one accountable name for every activity.
Immediate download. One work email. No paywall and no nurture sequence.
How many AI agents can one person run?
What ASC measuresAgent Span of Control (ASC) is the number of AI agents one person is able to review and manage well, calculated as base task capacity divided by risk weight times complexity factor.
Why the formula mattersAgent Span of Control is a formula, not a guess: base task capacity divided by risk weight times complexity factor.
What changes the ceilingBase capacity is how many agent outputs one person can review in a day without quality slipping. Risk and complexity pull that number down, since a high-stakes, judgment-heavy task needs more attention per agent than a routine one.
Why one ratio failsThe ratio isn't fixed company-wide. A sales rep reviewing low-risk research agents has a different ceiling than a CS manager reviewing renewal-risk scores that touch churn decisions. Calculate it per role, not once for the whole org.
01 / STARTBase capacityHow many outputs can be reviewed before quality slips.
02 / APPLYRisk weightHigher exposure demands a more deliberate review.
03 / APPLYComplexity factorMore judgment reduces the safe operating span.
Questions this raises
There isn't one number. It depends on the task's risk and complexity for that specific role. A low-risk, routine task might support ten or more agents per person. A high-risk, judgment-heavy task might support two. That's why this calculator runs the formula per role instead of giving one company-wide target.
Quality drops before anyone notices. A person reviewing more agent output than their capacity allows starts approving things faster and catching less. The fix isn't always fewer agents. Sometimes it's tighter risk tiering, more oversight, or a second reviewer.
It adds a second axis to it. Headcount planning has always asked how many people you need. This asks how many agents each of those people can responsibly run. Both numbers belong in the same conversation now.
ROI answers whether the spend is worth it. This answers something that comes before ROI: whether the ratio you're running is safe. A strong ROI number built on an unsafe ratio is a mistake waiting to surface.