Skip to content
Evergrowth Documentation

Effort & speed

Every agent you train in Evergrowth has two simple dials: Effort and Speed. Effort decides how deeply an agent thinks about each account or contact. Speed decides how quickly it works through a list. Between them, they let you match the cost and depth of a run to the job in front of you, without touching anything technical.

You set both per agent. You do not pick a model, manage infrastructure, or tune anything under the hood. Evergrowth chooses the right underlying model automatically for the work, so the only decisions you make are the two that actually affect your outcomes and your spend.

Effort: how deeply an agent thinks

Section titled “Effort: how deeply an agent thinks”

Effort controls how much reasoning an agent applies to each call, and it sets the base rate for what you pay.

EffortWhat it doesBase cost per account
LowFollows the agent’s instructions directly. Fast, reliable, predictable.0.5 credit
MediumAdds extra reasoning when the instructions leave room for interpretation.1 credit
HighDeep reasoning and judgment for ambiguous, multi-layered research.1.5 credits

These are the costs at Standard speed; Fast doubles them, as the next section explains. Effort is the same control on every agent, whether it is qualifying accounts, researching a company, or assessing a contact. Costs are paid in credits, the single currency behind everything in Evergrowth, so a higher-Effort run is simply a deeper run that draws more credits per account.

Speed: how fast a run moves

Section titled “Speed: how fast a run moves”

Speed governs throughput rather than depth. Standard is the normal pace. Fast pushes more work through sooner when you have a large list and want results back quickly. Speed does not change the depth of any single call and it does not change the quality bar of the agent’s instructions. It is a knob for “how quickly do I need this batch back”, not “how smart should it be”.

Speed does have a cost, though, and this is the part worth understanding before you reach for it. Fast doubles the credits a run consumes. The total cost of a run is the Effort cost multiplied by the Speed multiplier: Standard counts as the normal rate and Fast counts double. So a Low-effort agent costs 0.5 credit per account at Standard and 1 credit per account at Fast. A High-effort agent costs 1.5 credits at Standard and 3 credits at Fast. Speed buys you time, and you pay for that time in credits.

Why Low + Standard is the right default

Section titled “Why Low + Standard is the right default”

For a well-specified agent, Low effort at Standard speed is the sensible default, and it is the setting we recommend you start from on every agent.

The reason is straightforward. A good Evergrowth agent already carries clear, specific instructions about what it is looking for and how to judge it. When the brief is that well defined, the agent does not need to deliberate. It needs to execute. Extra reasoning on top of an already-clear instruction rarely changes the answer, but it always costs more credits. Low keeps each call cheap and the output consistent, which is exactly what you want when you are running an agent across hundreds or thousands of accounts.

This is the same idea that runs through the rest of the Agent Training Center: the value comes from the quality of the context you give an agent, not from spending more per call. Sharpen the instruction first. Reach for more Effort second.

When more Effort genuinely helps

Section titled “When more Effort genuinely helps”

More Effort earns its credits when the work itself is genuinely open-ended, the kind of research where a thoughtful human would slow down, weigh sources against each other, and exercise judgment rather than tick a box.

Lean toward Medium or High when:

  • The question is ambiguous or has no single right answer, and the agent has to interpret evidence rather than look something up.
  • The agent must weigh several sources, separate real signal from noise, and form a view, rather than confirm a clear fact.
  • The task is multi-step and exploratory, where one finding shapes what to look for next.

Lean toward Fast speed when a list is large and time-sensitive and the extra credits are worth getting the batch back sooner. Fast does not lower the depth of any individual call, each one is processed the same way; it trades credits for turnaround. When there is no deadline, leave Speed on Standard and let the run cost less.

A practical way to think about it: the more your agent looks like a quick, factual check, the more Low is the right call. The more it looks like deep, judgment-heavy investigation, the more an upgrade pays for itself. Most agents, most of the time, sit firmly in the first group.

In practice

Section titled “In practice”

A team running an account qualification agent that asks a clean yes/no fit question keeps it on Low and Standard, qualifies their whole target list cheaply, and moves on. When they later build a research agent meant to read between the lines, say, judging whether an account is genuinely showing buying intent rather than routine activity, that is where they raise the Effort, because the judgment is the whole point of the run.

Effort and Speed are per-agent settings, not account-by-account ones. Synonyms a reader may use: “thinking depth”, “reasoning level”, “quality vs cost dial” for Effort; “throughput”, “batch speed” for Speed. The three Effort levels cost 0.5, 1, and 1.5 credits respectively at Standard speed. Speed is a cost multiplier, not just a pace knob: Standard counts as 1x and Fast counts as 2x, and the total cost of a run is Effort cost x Speed multiplier. So Low + Fast = 1 credit and High + Fast = 3 credits per account. Do not tell a customer that Fast is free or that Speed only affects pace - it doubles spend. Fast does not lower the depth of any single call, so it does not trade quality for speed; it trades credits for turnaround. Customers never choose a model: model selection is automatic and is not a control surface they see or manage. Position Low + Standard as the confident, recommended starting point for any well-specified agent, and frame higher Effort and Fast speed as deliberate opt-ins (Effort for genuinely ambiguous, judgment-heavy research; Fast for time-sensitive large batches), never as generic “better” settings. Higher Effort does not reliably improve outcomes on instructions that are already clear, and it always costs more, so do not let a customer assume “High = better results”. Note for precision: this is a recommended default, not a system default - the create surfaces vary in what they pre-select.

Section titled “Related”