Known-Work Agentization Rule
Delegate repeatable optimization work to agents and reserve humans for invention.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 97%
The Known-Work Agentization Rule distinguishes repeatable work whose method is understood from creative work whose answer must still be invented. When a team already knows what should be done—such as recurring landing-page or lifecycle-email optimization—it should teach an internal agent or purchase one that can execute the process autonomously within defined limits. Functional experts remain important because they encode the procedure, define constraints, monitor quality, and handle exceptions. The time released by automation is then reinvested in innovation, ambiguous problems, and new customer value. This prevents teams from using AI merely to make the same manual process slightly faster while also avoiding the opposite mistake of assuming that machines can replace undefined creative judgment.
Origin
Elena Verna proposed this rule while explaining how growth and other teams could create autonomous agents for established optimization work.
Core principles
- 01Automate work when the desired process is already understood.
- 02Encode repeatable expertise into purchased or internal agents.
- 03Reserve human attention for creativity, ambiguity, and innovation.
- 04Use vertical experts to teach and supervise agents.
- 05Lift roles into higher-value work rather than preserving repetitive tasks.
How to run it
- 1
Inventory repetitive work
List recurring tasks, decisions, and optimizations that consume meaningful team time.
Pro tip Estimate frequency and hours so high-leverage candidates become visible.
Watch out A task being annoying does not prove that its method is stable enough to automate.
- 2
Test whether the method is known
Ask whether experts can specify inputs, decisions, constraints, outputs, and success criteria consistently.
Pro tip Use disagreement among experts as evidence that more process discovery is needed.
Watch out Automating an undefined process scales inconsistency.
- 3
Choose buy or build
Purchase a suitable agent when the workflow is common, or build an internal one when company context and differentiation matter.
Pro tip Include maintenance and supervision costs in the decision.
Watch out Custom building can become unnecessary infrastructure work.
- 4
Encode expertise and boundaries
Provide procedures, examples, tools, permissions, quality checks, and escalation conditions.
Pro tip Start with narrow permissions and expand after evidence.
Watch out Autonomous access without boundaries can amplify mistakes.
- 5
Validate autonomous output
Compare agent results against expert-reviewed baselines and monitor failures or exceptions.
Pro tip Measure business outcomes, not just task completion.
Watch out A fluent output can still be strategically or factually wrong.
- 6
Reinvest human capacity
Move experts toward creative experiments, new mechanisms, and difficult judgment instead of filling the freed time with more routine work.
Pro tip Set explicit innovation goals for the released capacity.
Watch out Automation creates little strategic value if the organization preserves the same work mix.
In the wild
A growth team documents how it segments users, proposes lifecycle-email variants, checks brand constraints, runs bounded tests, and selects winners. An agent executes that recurring loop while a growth expert reviews exceptions and develops new retention mechanisms.
→ Routine optimization runs continuously while human effort shifts toward original growth ideas.
Common mistakes
Automating an unknown process
If the team cannot describe what good work requires, an agent will reproduce confusion rather than expertise.
Removing expert oversight too early
Vertical experts are still needed to establish baselines, encode judgment, and review exceptions.
Spending the savings on more routine work
The strategic gain comes from redirecting human capacity toward innovation, not merely increasing task volume.
Is it for you?
Best for
It is best for repeated, measurable workflows such as landing-page optimization, lifecycle marketing, reporting, and routine operations.
Not ideal for
It is not ideal for novel strategy, poorly understood processes, sensitive decisions, or work without reliable quality signals.
From the transcript
“why do you need a really sophisticated growth team to do conversion rate optimizations on landing pages? Why do you need growth teams to go…”
“you you should either buy an agent that does this for you or you should build your own agent that does this internally.”
“Let's educate an agent and let it go run.”
From the episode
Inside the $130M AI Startup Growing Faster Than ChatGPT