OperationsAI ChatImage GenerationVideo Generation

Founders, startup operators, and early-stage teams

Startup Planning

Turn a startup idea into a one-page working plan, an evidence-and-assumption register and a small validation experiment. Use AI to structure the questions and alternatives while checking market evidence, cost arithmetic and business decisions independently.

Practical guidance

Know what you are making

A founder-review packet explaining the customer problem, proposed offer, evidence gaps, cost assumptions and next experiment. It is a planning aid, not proof of demand, an investment recommendation or a forecast of revenue, funding or profitability.

Before you start

A concrete idea and customer problem

Describe who might use the offer, what task or problem it addresses and how people currently solve it. Identify direct experience separately from assumptions. Do not invent interviews, testimonials or traction to make the business sound more established.

Dated evidence and known costs

Provide public research excerpts, authorized interview notes and actual supplier quotes where available. Include geography, time period, units and source links. Use synthetic or sanitized information; do not share bank details, confidential investor documents or customer records without authorization.

Constraints and a decision horizon

Set the time, budget and skills available for the next experiment, not an unlimited launch plan. State which decision the evidence should inform. Ask an appropriate accountant, lawyer or other qualified adviser to review specialized business, financial and legal questions.

Practical guidance

From input to reviewed output

  1. Write a hypothesis-led one-page plan

    Ask chat to organize the proposed customer, problem, offer, alternative solutions and likely delivery needs. Mark each claim as supported, inferred or unknown. Keep the first plan narrow enough that the weakest assumption is easy to see.

    Checkpoint: A convincing story is not validation; identify which important claims still lack direct evidence.

  2. Map the evidence and research gaps

    Separate broad market information from questions about the specific audience. Draft neutral interview or survey questions and a source ledger. Record how a finding was obtained so a national industry statistic is not presented as local purchase intent.

    Checkpoint: Open every cited source, check the date and definition, and exclude fabricated or unsupported references.

  3. Build a transparent cost outline

    List one-time and recurring categories using actual quotes where possible. Record estimated ranges with their assumptions and time periods. Keep revenue scenarios separate from cost observations, then verify calculations in a spreadsheet outside chat.

    Checkpoint: Unknown costs stay unknown or explicitly estimated; do not let a model fill the budget with plausible-looking prices.

  4. Design a small validation experiment

    Choose one uncertainty and a practical way to learn about it within your own budget. Define who to speak with, what to observe and what would change the plan before starting. Consider a manual test before building a full product.

    Checkpoint: Record the method and decision rule without promising a sample size or result will establish the entire market.

  5. Update the narrative with actual findings

    After the experiment, supply sanitized observations and ask for a revised plan. Explain what changed, what remains uncertain and what the next decision is. Keep assumptions and real traction visibly separate in any pitch or stakeholder update.

    Checkpoint: Do not turn projected revenue, hypothetical users or AI-written praise into claims of observed business performance.

Technical details

What the workflow does—and does not do

SBA market-research guidance

Broad market data and direct research differ

SBA distinguishes using existing sources from asking customers directly and suggests examining demand, market size, alternatives and pricing. Use that distinction to label the evidence in your plan. Broad statistics can frame a question, but they do not establish that a specific audience wants your proposed offer.

SBA startup-cost guidance

Separate one-time and recurring costs

SBA's startup-cost guidance distinguishes initial expenses from ongoing monthly expenses. Use those categories to organize owner-supplied figures and estimates. This page does not verify supplier pricing, tax treatment, funding needs or customer charges; calculations and specialized decisions need independent review.

Editorial planning boundary

Scenario arithmetic is not a prediction

State the formula, units and time period for each scenario and verify the arithmetic in your own spreadsheet. A model can help explain assumptions but may miscalculate or mix monthly and annual inputs. Do not present a sensitivity case, break-even sketch or projected growth rate as observed performance.

App integration boundary

Planning text does not operate the business

This guide does not establish bank, accounting, CRM, incorporation, investor or market-data integrations. It does not file documents, contact prospects or launch experiments. Use authorized source material and your own approved tools, with qualified advice for legal, tax, financing or investment questions.

Technical details

Choose by task, not by a universal ranking

These are editorial starting points, not measured comparisons. Check the exact model guide and current app access before using a model; linked catalogs may also contain historical entries.

Structuring the initial plan

Evaluate a chat model on a short idea brief and ask it to expose missing assumptions rather than sell the concept. Check whether proposed activities fit your actual budget, skills and time.

Tradeoff: A detailed plan may conceal unsupported demand or costs. More sections are not a substitute for a better test of the central problem.

Explore Anthropic models

A challenge to the proposed direction

Give another available model the same evidence and ask for alternative explanations, narrower experiments and questions a reviewer might ask. Keep the goal learning, not a confident verdict on whether to invest.

Tradeoff: Model agreement cannot predict business success or validate financial projections. Decisions need real evidence and appropriate professional review.

Compare OpenAI models

Source-oriented market questions

Check the exact model and session tools before using a search-oriented workflow. Ask for dated public sources addressing a defined market question, then inspect the underlying pages yourself.

Tradeoff: Search or citation support depends on the route and tools. A linked source may not support the claimed market size, geography or current price.

Read the Perplexity guide

Practical guidance

Create, refine and review

Replace every bracketed placeholder with authorized context. These are original templates, not generated-output demonstrations or generation-tested recipes. Copying does not submit anything or incur a generation charge.

Create an assumption-led startup plan

Where to use it: Paste into AI Chat with a concise idea brief and dated evidence.

Build a one-page working plan for [IDEA] serving [CUSTOMER AND PROBLEM]. Evidence: [DATED SOURCES OR AUTHORIZED NOTES]. Constraints: [TIME, BUDGET AND SKILLS]. Separate supported claims, assumptions and unknowns. Describe the offer, current alternatives, delivery needs and three important risks. Propose one narrow learning experiment. Do not invent market size, interviews, traction, supplier prices or predictions of funding and revenue.

Check: Confirm the plan identifies what must be learned rather than presenting an attractive story as validation.

Refine the cost and experiment assumptions

Where to use it: Paste into AI Chat with owner-supplied figures and an initial plan.

Refine [WORKING PLAN] using [ACTUAL QUOTES, ESTIMATES AND SOURCE DATES]. Separate one-time and recurring costs with units and time periods; leave missing figures unknown. Show proposed formulas for external spreadsheet verification, not certified results. Design a small experiment for [KEY UNCERTAINTY] within [LIMITS], including what to observe and what would change the plan. Do not promise demand, profit or investment returns.

Check: Recalculate outside chat and have the appropriate professional review specialized financial or legal assumptions.

Review the evidence before sharing a pitch

Where to use it: Paste into AI Chat with the revised plan and actual experiment notes.

Audit [PLAN OR PITCH] against [SOURCE LEDGER] and [OBSERVED EXPERIMENT RESULTS]. Identify invented traction, unsupported market claims, mixed units, hidden cost assumptions and statements that sound certain without evidence. Cite the exact passage and propose accurate wording. Separate forecasts from actuals and list unresolved questions for qualified advisers. Do not recommend an investment or declare the business validated or ready for funding.

Check: Verify source links and measurements yourself, then ensure the shared version keeps the uncertainty visible.

Illustrative example

What a useful handoff could contain

An editorial example of structure, not a generated result, customer case study or measured outcome.

Illustrative problem and offer

A fictional startup proposes a scheduling aid for local workshop organizers. The assumed problem is time spent coordinating availability. Whether organizers will change tools or pay for the offer remains an explicit hypothesis, not a claimed market finding.

Evidence and cost ledger

Rows track the claim, source, date, geography, confidence and next check. Cost categories include a setup item and recurring service expenses, with no invented amounts. Real quotes and calculation results must be added by the owner and independently verified.

Small learning experiment

The proposed next step is to discuss current scheduling work with appropriate organizers and record concrete examples. The plan names what would support, weaken or change the idea. No interviews have been conducted and no outcome is implied by this example.

Technical details

Check the result before relying on it

Quality checklist

  • Customer demand, market size and competitive claims link to sources with dates, definitions and appropriate scope.
  • Actual observations are separated from hypotheses, forecasts, synthetic examples and hoped-for traction.
  • Cost figures retain sources, units and time periods, and arithmetic is verified outside the chat response.
  • The next experiment has an owner, a bounded budget and a clear learning question, with specialized decisions referred for qualified review.

When the result misses the brief

The plan sounds certain because it is detailed

Add an evidence-status column and remove unsupported numbers. Ask which single assumption could invalidate the proposal rather than requesting more polished sections.

A large market statistic becomes a sales forecast

Check its definition, geography and period. Distinguish overall industry activity from the customers you can actually reach and from any independently tested willingness to pay.

Estimated costs are copied as real quotes

Label estimates, preserve the source date and request actual supplier information through your normal process. Verify taxes, contracts and financing questions with appropriate professionals.

Provider maximums, picker estimates and actual customer charges are different things. Read the pricing explanation.

Common questions

Before you start the workflow

Will this tell me whether the startup will succeed?

No. It can organize assumptions and help plan learning, but it does not predict demand, revenue, funding or profitability. Real evidence and business judgment remain necessary.

Can I put generated numbers in an investor pitch?

Do not present them as facts. Use verified observations or clearly labeled scenarios with transparent assumptions and independent arithmetic checks. Obtain appropriate professional review for financial and legal representations.

Does this conduct market research automatically?

It can help organize supplied research and draft questions. Current source access depends on the selected route and tools; it does not conduct interviews, contact customers or validate demand on your behalf.

Evidence and review

Sources & boundaries

The linked technical references were reviewed on . Documentation and app availability can change.

The plan, prompts and fictional example are editorial aids, not market research results, financial advice or investment recommendations. SBA references support the attributed planning distinctions. No business was evaluated, no money was invested and no customer, banking or accounting system was accessed.

Opening EZ Ai Assist does not prefill these templates or automatically run the workflow. Choose the relevant mode and review your input before submitting it.

Continue exploring

View all use cases