Skip to content

Structured Business Information → Appropriate Next Action

Can AI take already-structured business information and consistently prepare an appropriate next operational action?

Published
12 Sept 2026
Type
Decision-preparation comparison, human vs AI
Sample
10 cases
Author
FIELD

Summary

Using the ten structured cases from Experiment #002, a human and an AI each assigned a priority, a next action, a reason and something to avoid. Both converged on the same operational pattern. AI was more conservative about missing information and stronger on operational prerequisites; the human was more commercially proactive and read customer psychology.

Research question

Can AI take already-structured business information and consistently prepare an appropriate next operational action?

Hypothesis

AI appears useful for preparing structured business work, while human judgment remains important where trust, context and commercial strategy affect the action. (Hypothesis, not conclusion.)

Method

The ten structured cases produced in Experiment #002 were used as input. A human first filled a table with the columns ID, Situation, Priority, Next Action, Why and What to Avoid. The AI model was then given the same structured input and asked for the same fields. The two sets were compared case by case, looking at where they agreed, where priorities diverged, and what each considered before acting.

Workflow

  1. 01Structured information
  2. 02Interpret current state
  3. 03Separate known from unknown
  4. 04Identify next action
  5. 05Prepare the action
  6. 06Human review

Dataset

The same ten synthetic cases (C001–C010) from Experiment #002, now in structured form. Synthetic dataset.

View the 10 cases
  1. C001Custom cake boxes with logo branding.
  2. C002Takeaway restaurant wants branded bags.
  3. C003Early-stage startup looking for juice packaging.
  4. C004Food business needs 1,000 packaging units urgently after supplier failure.
  5. C005Flyer printing inquiry with zero details.
  6. C006300 shopping bags plus branded stickers for a launch.
  7. C007Inadequate supply and poor quality from the current supplier.
  8. C008Lead asks for a branded packaging catalogue.
  9. C009Bakery moving from plain to branded packaging.
  10. C0105,000 waterproof stickers needed before end of month.

Baselines

Human baseline

The human assigned priorities of high (C004, C006, C007, C009, C010), medium (C001) and low (C002, C003, C005, C008). Next actions leaned toward sending samples, catalogues or comparison overviews to move the conversation forward, and toward reassurance where trust had been damaged (C007). See the case table for each entry.

AI baseline

The AI assigned high priority to C004, C006, C007 and C010, medium to C001, C002, C003 and C009, and low to C005 and C008. Next actions consistently began with confirming specifications, location, deadline and capacity before any quote or commitment. See the case table for each entry.

Evaluation criteria

Convergence
Did human and AI arrive at the same operational pattern?
Priority
Where did urgency assessments differ, and why?
Prerequisites
Did the next action account for what must be known before acting?
Commercial context
Did the action account for trust, momentum and customer psychology?
Premature commitment
Did either side promise or quote before it reasonably could?

Results

Human and AI agreed on priority in seven of ten cases (C001, C004, C005, C006, C007, C008, C010); they differed in three (C002, C003, C009 — the human rated C002 and C003 low where the AI said medium, and C009 high where the AI said medium). Both consistently avoided premature quotes. The AI's actions were more procedurally complete (specs, location, capacity); the human's were more oriented to moving the customer forward with samples, overviews and reassurance.

Key findings

  • Human and AI generally converged on the same operational pattern: understand the situation → identify missing information → assess urgency → choose a next action → avoid premature commitment.
  • AI was more conservative about missing information.
  • Human judgment was more commercially proactive, especially in C002, C003, C007 and C009.
  • AI was stronger at identifying operational prerequisites such as specs, location and capacity.
  • Human judgment incorporated customer psychology, trust and context, especially in C007.
  • AI was more cautious about priority; C009 was medium for the AI versus high for the human.
  • AI can function as a procedural guardrail against premature commitments.
  • Excessive AI caution may slow conversational momentum.

Show the failure

Failure modes

  • F4Overconfidence

    Not observed on the AI side; the AI erred toward caution rather than overcommitment.

  • F6Prioritization

    Priority divergence in C002, C003 and C009 — AI more cautious than the human.

  • F7Context

    AI actions underweighted trust and momentum where the human read the customer's situation (C007, C009).

Observed → Inferred → Unknown

Observed

  • Both sides avoided quoting or promising before specifications were confirmed.
  • The AI asked for more prerequisites per case than the human did.
  • In C007 (customer failed by a previous supplier) both sides recommended acknowledging the failure and sending samples; the human framed it explicitly as trust-building.
  • In C002, C003 and C009 the human moved to samples or comparison material sooner; the AI asked more questions first.
  • The AI's framing of what to avoid was very consistent across cases; the human's varied in emphasis.

Inferred

Interpretations — our reading of the evidence, labelled as such.

  • AI can translate structured information into a consistent next-action framework, particularly when the action depends on clearly identifiable prerequisites.
  • Human judgment adds commercial context not represented in the structured fields — reducing uncertainty for the customer, building trust and keeping momentum.
  • The AI's caution is a feature when the risk is overcommitment, and a cost when the risk is losing the conversation.
  • Across #001–#003 a pattern is emerging: Interpret → Structure → Prepare action.

Unknown

What this experiment does not establish.

  • Whether the human's more proactive actions actually produce better commercial outcomes; no outcomes were measured.
  • Whether AI caution would relax with different instructions, and at what cost to reliability.
  • How the pattern behaves with real customers and real follow-up.
  • Anything about autonomous decision-making — these experiments do not test it.

Human vs AI, case by case

Side by side, without declaring a universal winner.

C001Custom cake boxes with logo branding.H: MediumAI: Medium

Human

Next action:
Clarify specs and budget.
Why:
Missing information.
Avoid:
Missing the deadline.

AI

Next action:
Clarify box dimensions/type, confirm quantity, logo/design, delivery and production constraints.
Avoid:
Quote or promise based only on quantity.
C002Takeaway restaurant wants branded bags.H: LowAI: Medium

Human

Next action:
Send samples and rough pricing.
Why:
Visual options help a busy owner pick a size.
Avoid:
Waiting for them to figure out specs alone.

AI

Next action:
Give relevant options/samples and ask product dimensions, bag size, location, budget.
Avoid:
Selecting a size or firm quote without knowing what the bags carry.
C003Early-stage startup looking for juice packaging.H: LowAI: Medium

Human

Next action:
Send a quick comparison overview of bottles vs pouches.
Why:
New businesses may not know setup cost differences.
Avoid:
Custom price quoting too early.

AI

Next action:
Ask product, volume, quantity, launch and requirements, then compare bottle vs pouch.
Avoid:
Pushing one format or a definitive quote before requirements.
C004Food business needs 1,000 packaging units urgently after supplier failure.H: HighAI: High

Human

Next action:
Find exact container specs and location.
Why:
No time for custom work.
Avoid:
Overpromising before verifying capability.

AI

Next action:
Immediately confirm container specs, location and the exact Friday deadline, then check stock and supplier capacity.
Avoid:
Promising before availability and logistics are confirmed.
C005Flyer printing inquiry with zero details.H: LowAI: Low

Human

Next action:
Reply with samples/catalogue.
Why:
Standard options force them to think about what they need.
Avoid:
A blank or incomplete response.

AI

Next action:
Ask for minimum specs needed for pricing.
Avoid:
A misleading fixed price, or overwhelming them.
C006300 shopping bags plus branded stickers for a launch.H: HighAI: High

Human

Next action:
Clarify specs and logo file.
Why:
Need to know what goes inside the bags to recommend fit.
Avoid:
Rushing production until specs are confirmed.

AI

Next action:
Confirm bag/sticker specs, logo/design files, location and production deadline, then separate options and quotes.
Avoid:
Production commitment before specs.
C007Inadequate supply and poor quality from the current supplier.H: HighAI: High

Human

Next action:
Acknowledge the bad experience, reassure on quality controls, send physical samples.
Why:
Build trust and reassurance.
Avoid:
Competing primarily on price, or promising quality without evidence.

AI

Next action:
Acknowledge the failure, establish specs and quality standards, arrange samples and pricing from a capable supplier.
Avoid:
Competing primarily on price, or promising quality without evidence.
C008Lead asks for a branded packaging catalogue.H: LowAI: Low

Human

Next action:
Send catalogue plus a friendly note to get more information.
Why:
Satisfies the immediate request.
Avoid:
Aggressive phone chasing.

AI

Next action:
Send the relevant catalogue and ask product, packaging type, quantity, location, timeline.
Avoid:
An irrelevant catalogue dump or an aggressive push.
C009Bakery moving from plain to branded packaging.H: HighAI: Medium

Human

Next action:
Send a quick showcase/sample.
Why:
Helps organize their thinking and accelerates a better decision.
Avoid:
Overwhelming the customer.

AI

Next action:
Determine quantities, product sizes, formats/materials, branding, location and timeline, then recommend or sample.
Avoid:
Assuming one bag format fits all products.
C0105,000 waterproof stickers needed before end of month.H: HighAI: High

Human

Next action:
Clarify specs and send a simple offer.
Why:
Start printing immediately once requirements are clear.
Avoid:
Excessive formality for every detail.

AI

Next action:
Confirm dimensions, waterproof material/finish, exact deadline and location, verify capacity, then quote.
Avoid:
Printing or promising before specs and capacity are confirmed.

What we learned

This experiment took the structured output of Experiment #002 as its input, closing the loop from messy inquiry (#001, #002) to prepared action (#003). It also produced the first cross-experiment pattern: Interpret → Structure → Prepare action.

Limitations

  • Same ten synthetic cases as #002; no new data.
  • Single human baseline.
  • No outcome measurement — we do not know which actions would have worked better with a real customer.
  • Qualitative comparison; agreement counts are descriptive, not statistical.
  • Does not establish autonomous decision-making.

Practical implications

What operators might reasonably take from this evidence.

  • Use AI to prepare the next action — prerequisites, questions to ask, what not to promise — and let a human decide how to send it.
  • Expect AI to be conservative; if momentum matters, plan to add proactive material (samples, overviews) yourself.
  • Where a customer has been let down before, the human read on trust mattered more than procedural completeness.
  • Treat priority ratings from AI as a second opinion, not a verdict.

Conclusion

AI can translate structured business information into a consistent next-action framework, particularly when the action depends on clearly identifiable prerequisites. Human judgment may add commercial context that is not represented by structured fields — especially where reducing uncertainty, building trust and maintaining momentum matter. Nothing in #001–#003 establishes autonomous decision-making; what is emerging is a preparation role.

Findings from this experiment

Related experiments

Research updates

Follow the research

Get new FIELD experiments and findings when we publish them.

Email only. No lead magnet. Unsubscribe any time.