What constitutes a startup? Examine the business, not the label
New, small and startup describe different things
A useful working distinction is whether a venture is designed for rapid growth through a repeatable offering. That ambition is not proof it can achieve it.
State your meaning
A startup label can describe an ambition, not a verified growth record.
Examine the constraint
Ask what must increase to serve another customer: founder hours, staff, infrastructure or a repeatable product.
Keep uncertainty visible
An idea, a trial, revenue and repeat purchases are different kinds of evidence.

If you are exploring Australia's startup community, it helps to know what someone means when they call a venture a startup. “Recently opened”, “uses AI” and “has investors” answer different questions. None tells you on its own how the business reaches customers or fulfils its promise.
A working definition, not a universal certification
In his September 2012 essay Startup = Growth, Paul Graham uses a growth-oriented definition: the company is designed to grow quickly. He distinguishes that intention from merely being new, and does not make technology or venture funding necessary conditions. This is an influential practitioner's framing, not a legal classification, universal age limit or guarantee of success.
Steve Blank's What’s A Startup? First Principles., published 25 January 2010, instead foregrounds the search for a business model that can be repeated and scaled. It helps distinguish testing a model from executing one already understood. These are related practitioner perspectives, not identical definitions or a certification process.
For this discussion, ask whether the venture is trying to develop an offering it can sell and deliver repeatedly to a large enough market for its ambitions. “Trying” matters: it may still be discovering who wants the product. An early venture can have startup intent without yet having evidence of repeatable demand.
Four fictional examples show the distinction
All four cases, the proposed hours and the contradictory request are fictional teaching material. No customer interviews, delivery timings, trials or payments were collected. These are not MLAI companies, investment assessments or reported outcomes. A case's interpretation applies to its stated plan, not permanently to that industry.
- F1: A new local repair shop
Scenario: The owner wants one dependable neighbourhood location. More repairs require more technician hours and bench space.
Interpretation: A new local business under the stated plan, not a rapid-growth startup merely because it just opened.
Next question or action: Decide whether one location meets the owner’s goals. A later expansion plan would need separate evidence.
- F2: An AI contractor proposing a shared quote-drafting product
Scenario: The current offer is bespoke workflow setup. A common template is proposed, but no shared product or customer trial exists.
Interpretation: A service business exploring a startup hypothesis. AI tooling and a reuse estimate do not prove repeatable delivery or demand.
Next question or action: Compare the actual setup differences before promising a product. The completed record below keeps that uncertainty visible.
- F3: A software team testing one product for several businesses
Scenario: The team intends to serve many similar businesses. In this story two businesses agree to try a demonstration; neither has paid or returned for another use.
Interpretation: Startup intent and a possible repeatable offering, not demonstrated retention or a validated business model.
Next question or action: Record what the trial actually involves and whether the same offering solves a recurring need. Two agreements are not two retained customers.
- F4: An established company trying an internal product
Scenario: A separate team is testing a new offer while the parent keeps running its established business.
Interpretation: The experiment may use startup-style discovery. That does not reclassify the whole parent as a newly founded company.
Next question or action: Name the unit being discussed. Check the actual applicant entity if considering a programme.
Models can change. A consultancy might develop a product from recurring problems, while a product team might discover that customers need substantial services. Describe the current evidence instead of forcing a permanent label.
If you want to explore these distinctions with other people, take the question “What is still custom for each customer?” to a relevant MLAI learning event. You do not need to be raising money or seeking contract work to join a startup discussion.
Inspect the setup assumptions, not just the smaller total
For F2, compare the same proposed task: setting up a quote-drafting workflow for each assumed client. Neither option includes automatically sending quotes. Illustrative arithmetic: five bespoke setups at 20 hours each require 100 delivery hours in the plan. A common setup of 40 hours plus 4 per client gives 60 hours. Neither figure is a measured result.
The table exposes two ways that proposed advantage can disappear. On a small screen, scroll it sideways; keyboard users can focus the table region and use arrow keys. The editable download contains the same calculations in plain text.
| Assumption | Clients | Bespoke per client | Common setup | Shared per client | Bespoke total | Shared total | Difference |
|---|---|---|---|---|---|---|---|
| Five assumed setups | 5 | 20 | 40 | 4 | 100 | 60 | 40 |
| Only one setup | 1 | 20 | 40 | 4 | 20 | 44 | -24 |
| More custom work | 5 | 20 | 80 | 8 | 100 | 120 | -20 |
With only one setup, the shared route takes 44 rather than 20 hours. With 80 common hours and 8 per client, five shared setups take 120 rather than 100 hours. Those are alternative assumptions, not follow-up test results. At the original rates, 40 + 4n becomes less than 20n only above 2.5 setups: three whole setups. This is an arithmetic threshold, not evidence that three customers exist or that a product is profitable.
The missing work matters: sales, support, maintenance, security and infrastructure remain unestimated, not zero. A different accounting integration or approval requirement could make the proposed tasks incomparable. Lower estimated setup effort alone does not prove demand, safe delivery or a scalable business. The example is not a margin forecast, cash-savings claim or recommendation to change business models.
Keep an evidence record, not a startup score
Download the one-page printable decision sheet for handwriting, or download the editable record and completed example. Both are real files with no signup. The PDF is a blank print sheet, not a fillable form; the text file includes all four cases, three calculations and the completed eight-field F2 record.
1. Whose problem does the venture address? 2. What is the offering today, not the future vision? 3. What repeats between customers? 4. What still requires custom work or founder time? 5. What shows demand: conversations, trials, payments or repeat use? 6. What evidence contradicts the current assumption? 7. What growth is intended, and what constrains it? 8. What remains unknown and what test comes next?
Completed eight-field record: F2's product hypothesis
- Whose problem does the venture address?
- F2: owners of small field-service businesses preparing draft quotes. This is an imagined audience, not a researched customer segment.
- What is the offering today, not the future vision?
- Bespoke workflow setup is the current fictional offer. A shared quote-drafting product is proposed, not built, sold or deployed.
- What repeats between customers?
- The proposed draft structure repeats. The 40 common setup hours and 4 hours per client are planning inputs, not time logs or evidence that the same integration will work.
- What still requires custom work or founder time?
- Each accounting system, quote approval rule and data permission may differ. Maintenance, support, security, acquisition and infrastructure are unestimated, not zero.
- What shows demand: conversations, trials, payments or repeat use?
- None collected. Five clients in the calculation means five assumed setups, not a pipeline, five customers or five payments.
- What evidence contradicts the current assumption?
- Supplied fictional request: “Our accounting app is different and two people must approve every quote.” That would add unscoped integration and approval work. It challenges the shared-setup estimate; it is not an actual interview quotation.
- What growth is intended, and what constrains it?
- The imagined contractor hopes to serve more firms with one product. Current custom delivery is the constraint. Keep describing this as a service offer exploring a product hypothesis, not a proven scalable business.
- What remains unknown and what test comes next?
- Pause the product claim. Next, compare two synthetic quote examples with different approval rules, logging common and separate work. Do not use client records or connect accounts. Then seek permission for discovery conversations. Neither test has run; no build, outreach or purchase is authorised by this worksheet.
The contradictory request changes the next step: investigate differences before repeating a scalability claim. It does not prove the product idea has failed. Replace fiction with permission-cleared notes when you have them, keep unknowns explicit, and revisit the interpretation when the offering changes.
A programme decides eligibility under its own rules
Do not add the eight answers into an invented eligibility score. As a dated example, the Australian Government's Industry Growth Program page, checked 11 September 2026, says the programme is paused to new applications. Its published Advisory Service rules include entity requirements, turnover conditions and an eligible commercialisation or growth project in specified priority areas. Calling F2 a startup would not establish any of those facts.
The same page distinguishes an Advisory Service report from eligibility for a grant and from a successful merit assessment. This is an example of why the actual programme and applicant matter, not a recommendation to apply or a full eligibility checklist. Check the current official rules and application status before acting; an accelerator, grant and tax scheme are different decisions. Seek appropriate professional advice for legal or tax questions.
Turn the definition into a useful conversation
At an event, “What have you learned about who uses this?” usually invites more substance than “Are you a real startup?” Share your own uncertainty and ask before giving advice. A person who wants a steady small business does not need to adopt a venture-growth ambition to belong in a discussion about AI.
Bring one question from the record to an MLAI event whose topic and experience level fit. Attendance does not promise investors, cofounders or customers. If you are already testing an idea, the first-customer conversation guide helps distinguish interest from stronger evidence.
Carry your preference to MLAI events
Choose a preferred MLAI event format. This does not upload your business record, assess eligibility, book advice or reserve a pitch slot.
Applied to the event CTA: All formats.
Select All formats and apply to clear it. Your selection is kept in this page's URL, not saved as a private shortlist.
Explore the business behind the idea
Have a question about what repeats and what still needs custom work? Bring it to a relevant MLAI talk or discussion, online or in person. Check the listing before registering; attendance does not promise business advice, customers or funding.
Find a startup learning eventSources and limits of this exercise
The linked Graham, Blank and Industry Growth Program pages were checked on 11 September 2026. The first two are historical practitioner definitions; the programme page supplies a specific administrative counterexample, not a universal meaning of startup. The cases, comparisons and worksheet are MLAI editorial teaching material, not research findings or third-party certification.
This revision adds a completed example and working downloads while preserving the original publication date. Independent founder and source review remain outstanding. No customer validation, legal classification, programme admission or event outcome has been established by this exercise.
Frequently Asked Questions
Does using AI make a business a startup?
Must a startup raise venture capital?
Is a small business less worthwhile?
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 →
Explore the business behind the idea
Have a question about what repeats and what still needs custom work? Bring it to a relevant MLAI talk or discussion, online or in person. Check the listing before registering; attendance does not promise business advice, customers or funding.
Find a startup learning event