Disclaimer: This article provides general information and is not legal or technical advice. For official guidelines on the safe and responsible use of AI, please refer to the Australian Government’s Guidance for AI Adoption →
An entrepreneur repeatedly converts assumptions into evidence and evidence into a decision.
Choose
Name a specific customer, problem and assumption that could be wrong.
Test
Run the smallest ethical experiment that can change the decision.
Decide
Continue, change or stop—and record why before starting the next cycle.
Direct answer
The work is deciding what deserves the next dollar and day.
Founders talk to customers, build and sell—but the deeper job is allocating scarce time, money and trust under uncertainty. A useful first 90 days therefore produces evidence and decisions, not merely a logo, feature list or long business plan.
Before day one: define the constraint
Begin with the runway and responsibility you actually have. Write down the hours and money you can risk, income you still need, caring or health constraints, and the date you will review the commitment. The Australian Government starting guide treats personal readiness, business definition and planning as separate steps for good reason: enthusiasm does not remove practical obligations.
Then write one sentence that can be disproved:
We believe [specific customer] experiences [specific problem] often enough to commit [money, time, access or behaviour] to [proposed outcome].
“Everyone needs this” cannot be tested. “Independent allied-health clinics with more than five practitioners lose at least three hours a week reconciling referrals” can be investigated.
Days 1–10: investigate the problem, not your pitch
Recruit people who have recently experienced the workflow. Ask about a specific occasion: what triggered it, what they did, what it cost, what they tried and who controls a purchase. Avoid explaining your solution until you understand the current behaviour.
Useful prompt
“Tell me about the last time this happened.”
Follow the sequence, tools, workarounds, delay, people involved and consequence. Recent behaviour is more informative than a prediction about a hypothetical product.
Weak prompt
“Would you use my AI platform?”
It reveals the concept, invites politeness and leaves price, urgency and decision authority undefined.
Summarise repeated facts, contradictions and unknowns. Do not turn three similar conversations into a market statistic. The output of this phase is a narrower assumption and a list of people willing to continue—not proof of demand.
Days 11–30: test a commitment before a build
Choose the smallest test that exposes the riskiest assumption. That could be a paid diagnostic, manually delivered service, letter of intent, prototype review with the actual decision-maker, or a time-boxed pilot. Set the success signal before running it so the result cannot be rewritten afterwards.
Problem decision
Look for: Recent examples, measurable consequence and an owner actively trying to improve it.
Do not substitute: Broad survey interest or agreement that the topic sounds important.
Offer decision
Look for: A customer accepts a defined outcome, price or meaningful next commitment.
Do not substitute: Feature requests from someone with no buying or adoption role.
Delivery decision
Look for: You can produce the outcome safely at a cost and effort the model can support.
Do not substitute: A polished demo that has never touched the real workflow.
Two illustrative experiment designs
These examples show the structure; they are not MLAI case-study results.
Assumption
Test
Pre-set signal
Decision
Clinic managers will pay to reduce referral triage time.
Offer a two-week, manually operated pilot with a defined privacy boundary and baseline measurement.
At least two qualified clinics sign the same paid pilot and provide approved test data.
Continue only if the commitment and delivery constraint are both met.
Technical founders need help producing investor updates.
Run a concierge update session from approved company evidence rather than building automation first.
Three founders return the next month and independently use the output with stakeholders.
Change the audience or workflow if usage does not repeat.
MLAI founder tool
Turn assumptions into decisions
Set the success signal before running the test. Record what actually happened, then choose to continue, change or stop. Entries stay in this browser until you clear its site data.
Your ledger will save locally after this page loads. Nothing is submitted to MLAI from this worksheet.
Days 31–60: deliver the outcome manually and measure it
A manual or partly manual service can expose missing permissions, edge cases, support work and adoption friction before software hides them. Define the baseline, target, test period, owner and failure condition. Record both the claimed outcome and the independently verified outcome.
For an AI workflow, this is also when to map the data, integrations, human approval points and recovery path. If the output affects people, money, access or reputation, decide who can challenge and override it. Do not put real personal or confidential information into an unapproved public AI tool.
A proposed weekly operating rhythm
This is a planning template, not a measurement of how successful founders spend their time:
Monday: choose one decision and the evidence needed.
Tuesday–Thursday: talk, sell, deliver and observe.
Friday: reconcile cash, review evidence, record risks and make the continue/change/stop decision.
Protect time for customer work. A week filled entirely with tooling, branding and internal planning can feel productive while leaving the main assumption untouched.
Days 61–90: decide what becomes a business
By day 90, aim to answer four questions honestly:
Does a defined customer experience the problem with enough consequence and frequency?
Have customers made a meaningful commitment to the proposed outcome?
Can you deliver that outcome safely, repeatedly and within a viable cost envelope?
Do the founders still want to pursue it under the actual constraints?
“Stop” is a valid evidence-based decision. “Change” should specify which assumption changed. “Continue” should name the next risk rather than simply authorise more building.
Australian setup: use decision gates, not a generic checklist
Formal steps depend on what you are doing. The official guide to starting a business covers readiness, structure, planning, registration, finances, customers, hiring and protection. Use it as the source of truth, then get legal, accounting or other professional advice where the choice depends on your circumstances.
Before trading or signing
Check structure, personal liability, contracts, ABN, business name, tax registrations, licences and insurance. Do not choose a company simply because it sounds more like a startup.
Document roles, time, decision rights, vesting approach, departures and IP ownership with qualified advice before conflict makes the conversation harder.
Keep one evidence trail rather than rebuilding the story for each investor, adviser or teammate:
customer problem and segment changes;
experiments, pre-set signals, results and decisions;
revenue, cash, runway and material commitments;
product usage and verified customer outcomes;
risks, incidents, assumptions and help needed; and
the next milestone and accountable owner.
This is the foundation for a useful company plan and stakeholder update. For a deeper customer-acquisition sequence, continue with how to get your first customers.
Australian primary sources
Official handoffs checked for this 28 July 2026 revision.
Office of the Australian Information Commissioner • Official checklist for whether the Privacy Act covers a small business and the obligations that may follow.
Fair Work Ombudsman • Official employment obligations, best-practice checklist and help for small businesses.
Government
First-founder questions
What does an entrepreneur actually do?
An entrepreneur turns uncertainty into decisions: choose a customer problem, test assumptions, assemble people and resources, deliver an offer, learn from evidence and decide what to continue, change or stop.
What should a first-time founder do first?
Write one falsifiable customer-and-problem assumption, identify people who genuinely experience it and run the smallest ethical test that could change your decision.
Do I need to register a company before testing an idea?
Not every early conversation requires a company, but trading, contracts, tax, liability, licences, ownership and hiring can create obligations. Use official Australian registration guidance and get professional advice for your circumstances before making structural commitments.
What belongs in a 90-day founder plan?
A useful plan names the riskiest assumptions, experiments, success signals, customer commitments, delivery milestones, cash constraints, owners and decision dates. It should change when evidence changes.
How is evidence different from positive feedback?
Compliments describe an opinion. Stronger evidence shows behaviour or commitment: sharing real workflow data safely, introducing a decision-maker, signing a pilot, paying, using the product or returning repeatedly.
Is this guide legal, tax or financial advice?
No. It is a general operating guide. Business structure, tax, employment, privacy and intellectual-property decisions depend on the facts, so use the linked regulators and qualified advisers.
