How to Build a New Business: The Complete Guide | A 9-Step, AI-Powered Framework for Your Business Plan

Hello, this is Ryuta Hamamoto from TIMEWELL.

Something I hear all the time from people who want to start a new business is, "I have the idea, but I don't know where to begin." I understand the feeling completely. I have launched several businesses in-house myself, including TRAFEED, our export-control AI agent, and the first wall I always hit is the "order" wall. When a good idea strikes, people naturally want to start building features and screens. Yet the businesses that start in that order are exactly the ones that end up building something nobody wants.

This article is a map for assembling a new business in the "right order." Who, what problem, value, solution, business model, competition, advantage, market size, and pitch. Follow these nine steps and a passing thought turns into the backbone of a business plan. The nine steps are simply well-known frameworks that have been used around the world, such as the Lean Canvas, Jobs-to-be-Done, and TAM, SAM, and SOM, rearranged to follow the flow of a launch. There is nothing novel invented here. What matters is doing the work in this particular order.

For each step, I have also prepared a prompt you can paste straight into an AI to think out loud with it, and a sample PowerPoint template you can fill in as you think. As you read, please try applying your own business idea. And if you want to check your own AI fluency before making AI your partner in building a business, there is a free AI Literacy Check that takes just three minutes.

First, let me boil this article down to three points.

• Build a new business starting from the customer side: who and what problem. Starting from the solution or the idea is the single biggest pattern of failure.
• The deliverables from the nine steps become the skeleton of your business plan and pitch deck as they are. Working on a single Lean Canvas keeps the whole picture from drifting.
• AI is powerful as both a first-drafter and a critic for each step. But verify your numbers with primary sources and your problems with customer interviews.

The big picture of the nine steps, and how to use this article

First, take in the whole map. The nine steps fall into three big blocks. The first half (1 to 3) is the phase for sharpening your understanding of the customer and the problem, the middle (4 to 5) is the phase for shaping the solution and the business, and the second half (6 to 9) is the phase for nailing down why you can win and how big it is, then turning that into a message. Thinking in these chunks keeps you from losing track of where you are.

There is one tip for how to proceed. Keep a "Lean Canvas," a single sheet of paper divided into nine boxes, at hand, and write the result of each step into its box as you go. The Lean Canvas is a one-pager devised by entrepreneur Ash Maurya in 2010, used to visualize the hypotheses of a highly uncertain new business quickly. Thinking only in your head scatters the issues, but putting them on one sheet makes the "boxes still empty" obvious at a glance.

The table below lists all nine steps. The detailed how-to for each step is written up in its own article. Take in the whole first, then read into whichever step interests you.

Step	What to do	Main tools	Read more
1. Who	Narrow the target and early adopters down to one person	Customer segment, persona	Article on setting your target
2. What problem	Dig into the job the customer truly wants done	Jobs-to-be-Done, empathy map	Article on digging into the problem
3. Value	Say your answer to the problem in one sentence	Value Proposition Canvas	Article on value
4. Solution	Cut down to the top three features and validate with an MVP	MVP, Lean Startup	Article on solutions and ideas
5. Business model	Who pays, how much, and how it is delivered	Business Model Canvas	Article on business models
6. Competition	Look at competitors including existing substitutes	3C, competitive map	Article on competitive analysis
7. Advantage	Put into words the assets that are hard to copy	Positioning map	Article on advantage
8. Market size	Estimate the size with TAM, SAM, and SOM	Fermi estimation, official statistics	Article on market sizing
9. Pitch	Convey who, what, and why you in 30 seconds	Elevator pitch	Article on building a pitch

As a base for getting hands-on, I have also handed out a slide deck you can fill in with these nine steps directly. Download the free sample PowerPoint template, write the content of each step into it, and by the time you finish reading, the skeleton of your business plan will be built. Rather than just skimming, I recommend reading while you fill it in.

First half: sharpen your understanding of the customer and the problem (steps 1 to 3)

The place where a new business is most likely to fail is, in fact, this first half. Skip it and start building from the solution, and no matter how exquisitely you craft it, you will finish something nobody wants. This is the most common gateway to failure, known as the product-out trap. That is precisely why you thrash out the customer and the problem first.

Step 1, "Who," is the work of narrowing your target as tightly as you can. The tip here is to think of the customer in three roles: the decision-maker, the person who pays, and the person who actually uses it. Businesses where these three diverge are surprisingly common, and if you miss the gap you will get the sales approach wrong. And the first group to aim for is the one for whom the problem is most severe, who are already spending money or effort trying to deal with it somehow. In other words, your early adopters. A setting as broad as "business professionals from their twenties to their sixties" resonates with no one. Try to picture that first single person in enough detail that you could give them a name.

Step 2, "What problem," is where you dig into the job the customer truly wants done. What helps here is the way of thinking called Jobs-to-be-Done. Proposed by the late Clayton Christensen, it takes the view that "customers don't want the product's features themselves; they hire the product to make progress on a job they want done." The milkshake anecdote is famous. So before you talk about features, you pin down "what is this person trying to accomplish, and what are they making do with right now?" The existing substitutes, such as Excel, phone calls, or simply putting up with it, become your first competitors as they are. Then narrow the problems down to the top three and evaluate them on three axes: severity, frequency, and willingness to pay. The most important thing here is not to decide by assumption. Always verify with five to ten customer interviews.

Step 3, "Value," is the work of saying your answer to the problem you unearthed in a single sentence. The Value Proposition Canvas is effective here. It is a tool devised by Alex Osterwalder, known for the Business Model Canvas: on the right you line up the customer's jobs, the gains they want, and the pains they want to avoid, and on the left your product along with what produces those gains and relieves those pains, then check that the two mesh. What many people tend to do here is describe value in terms of features. Explanations like "automated with AI" or "feature-rich" are the maker's words, not the customer's. Write value in terms of "how the customer changes." Once you can state "what existing thing you replace, and for whom" in one line, you pass the first half.

These first three steps become the foundation for everything in the second half. Calculating market size while the customer is still vague is a castle built on sand. Sometimes the long way round is the fastest, so spend your time here.

Middle: shape the solution and the business (steps 4 to 5)

Only once the customer and the problem are set does the solution get its turn. This is where keeping the order pays off.

Step 4, "Solution," is where you narrow the means of delivering value down to the top three features. You will be tempted to pack in this and that, but aiming for a finished product from the start is forbidden. In the Lean Startup thinking systematized by Eric Ries, the emphasis is on validating before you build it out. You first prepare the minimum product needed to test the hypothesis, the so-called MVP. An MVP does not need to be a polished app. A single landing page, a concierge-style operation where people run things by hand behind the scenes, or a paper prototype is fine. What you should decide here is to narrow down to one single most-critical hypothesis: "what has to be true for this business to hold up?" Then write a plan to check it at the lowest cost and in the shortest time. Build, measure, learn. How many times you can go around this loop determines the speed of a new business.

Step 5, "Business model," is the work of turning the business into a mechanism where money flows. Using the Business Model Canvas, you design the revenue streams, the cost structure, and the channels that reach the customer. Who pays, how much per unit, and is the pricing a subscription, usage-based, or a commission. What is surprisingly often overlooked here is the case where the user and the payer are different people. For a service where, say, a parent uses it and a child pays, unless you build this "gap" into the design from the start, you will struggle later with how to convey it. And one more thing: work out your unit economics, even roughly. Look at the gross profit per customer and the relationship between customer lifetime value and acquisition cost, and check that you are not "losing more the more you sell." Skip this and you end up in the un-funny situation of running out of money even as you grow.

By the middle, the outline of the business gets quite clear. It raises accuracy to pause here once and have someone stress-test it from an outside perspective. In WARP, the AI consulting service we provide, we routinely do this kind of sounding-board work to turn new-business ideas into the shape of a business. I feel that not trying to finish it all alone, and having someone who will criticize you without holding back early on, is the single best shortcut for avoiding detours.

Second half: why you can win, how big it is, and how to convey it (steps 6 to 9)

Once the shape of the business is done, you nail down "why our company can win" and "how big the market is," then finally turn it into a message.

Step 6, "Competition," is where you draw a map that includes not just direct competitors but existing substitutes. What you must not do here is conclude "there are no competitors." Even if there is no direct rival, the customer is always substituting with something. They manage it in Excel, they check by phone, or they give up and put up with it. All of those are your competitors. Organize them with what is called the 3C (customer, competitor, and company) and find the dissatisfaction that existing substitutes fail to resolve. That hole of dissatisfaction is exactly your opening.

Step 7, "Advantage," is the work of evaluating why your company can keep winning, judged by how hard it is to copy. Ash Maurya defines a true advantage as "something that cannot easily be copied or bought." A common mistake here is to list "first-mover advantage," "financial muscle," or "having more features" as advantages. Unfortunately, these get caught up with quickly. What to aim for are assets that cannot be bought and are hard to imitate: data that accumulates uniquely over time, a team of experts, a channel that is hard to copy, a community, or network effects. It is fine if you don't have it now. It is enough if you can put "how you will build it from here" into your plan.

Step 8, "Market size," is where you estimate the size in three layers: TAM, SAM, and SOM. TAM is the maximum achievable market, SAM is the range you can realistically reach, and SOM is the figure you can actually capture from your first year over the next several years. A common failure here is looking at TAM alone and being satisfied with "the market is a trillion yen." Big alone is not proof of demand. Estimate it two ways and reconcile them: top-down, by apportioning from official statistics, and bottom-up, by stacking up assumed number of customers times unit price times frequency. Don't forget to always note the numbers you assumed and their sources. A big number with no basis is seen through in an instant by investors and bosses.

Step 9, "Pitch," is the work of rearranging the deliverables so far into an order that lands. Problem, value, solution, business model, market size, advantage, and the results you gained from validation. Line them up in this order and first build a 30-second elevator pitch. Here too, showing off features is forbidden. What the listener wants to know is one thing: "whose pain, and why can you solve it?" If you have filled in the nine steps carefully, you already have all the material.

Thinking out loud with AI at each step (a prompt collection)

Here is another highlight of this article. The nine steps go around astonishingly fast when you make generative AI your partner. There are three ways to use AI: have it draft first versions, have it play devil's advocate, and have it think up the questions you should ask customers. But as a fundamental premise, AI output is raw material for hypotheses, not fact. Always verify market-size numbers, statistics, and proper nouns against primary sources, and confirm customer problems through real interviews. Get speed from AI, and reliability from people. Just don't let that line blur.

You can copy the prompts below and paste them straight into an AI. Replace the parts in brackets with your own business.

First, a prompt to have it draft first versions of all nine steps at once.

You are a business-development consultant who has supported the launch of many new businesses. Please create a "first draft" that organizes my business idea into the following nine steps.

How to proceed:
(1) Read the input first; where a premise is missing, you may fill it in, but label it clearly as an "assumption."
(2) Write each step in concrete terms—specific personas, situations, numbers, and proper nouns—avoiding abstract words such as "convenient," "high-quality," or "innovative."
(3) Keep each step to three bullet points or fewer, and present each as a "hypothesis" rather than a definitive statement.
(4) At the end of each step, add one "question I should ask a customer to validate this."

Steps = (1) Who / target and early adopters (2) What problem (3) Value (UVP) (4) Solution (5) Business model (who pays, unit price, channel) (6) Competition and existing substitutes (7) Overwhelming advantage (8) How to think about market size (9) 30-second pitch

Output format:
- Separate each step with a "## Step N: name" heading.
- Under each heading, place "Hypothesis" (three bullets or fewer) and "Validation question" (one).
- Whenever you use a number, label it "assumption" and add one line stating the premise behind it.
- End with "## Top 3 weakest hypotheses that most need validation," listing three hypotheses that would sink the business if they fail, each with the reason.

Constraints:
- Do not fabricate facts or data. Write "unverified" for anything you do not know.

Business idea: [describe in 1-3 lines here]
Hypothesis about the target customer: [ ]
Hypothesis about the problem to solve: [ ]

When you want to sharpen your understanding of the problem, have it dig in from a Jobs-to-be-Done perspective. It will even create customer interview questions, so you can use them in the field as is.

You are a customer researcher specializing in Jobs-to-be-Done. Please dig into the "job the following customer truly wants done."

Steps:
(1) Hypothesize the job, split into functional, emotional, and social dimensions.
(2) Identify the existing substitutes (workarounds) the customer uses today, and estimate the points of dissatisfaction with each.
(3) Weigh the jobs and dissatisfactions by severity, frequency, and willingness to pay, and identify the most pressing one.
(4) Create 10 customer interview questions to confirm these hypotheses (non-leading, phrased to ask about specific past behavior and experience—use "When did you last do X, and what did you do?" rather than "Do you think X?").

Output format:
- Summarize the "job hypotheses" in a table (columns: dimension / the job / why you think so), with rows for functional, emotional, and social.
- Summarize "existing substitutes and dissatisfactions" in a table (columns: substitute / when it is used / point of dissatisfaction).
- Present the "10 interview questions" as a numbered list.
- Label any number or proper noun as an "assumption."
- End by listing the "top 3 hypotheses most prone to bias," each with the reason.

Constraints:
- Avoid abstract words; write in concrete scenes and behaviors. Do not fabricate facts.

Customer: [who they are and what situation they are in]
Assumed scene: [ ]

For a Lean Canvas draft, accuracy rises when you have the AI create it and then critique itself. The trick is to make it switch to an investor's point of view.

You are a new-business mentor experienced with the Lean Canvas. From the information below, create a draft of the nine Lean Canvas items, then critique it as an investor.

Steps:
(1) Fill in the nine Lean Canvas items (customer segments / problem / unique value proposition / solution / channels / revenue streams / cost structure / key metrics / overwhelming advantage), plus existing substitutes and early adopters.
(2) Write each item in concrete terms—specific people, numbers, and proper nouns—avoiding abstract words. Label any number you cannot ground as an "assumption" and add one line stating the premise.
(3) After creating it, switch into the role of an "experienced investor" and review it critically.

Output format:
- Present the Lean Canvas as a table (columns: item / entry).
- Below it, under "Investor review," list the top three weakest items, and for each write "why it is weak," "the reasoning," and "what should be validated further."
- Do not fabricate facts or data. Write "unverified" for anything unknown.

Business overview: [ ]
Assumed customer: [ ]
Solution: [ ]
Revenue model: [ ]

For market size, have it produce the formulas and candidate sources for verification, which makes the later primary-source check easier.

You are an analyst specializing in market-size estimation for new businesses. For the following business, roughly estimate TAM, SAM, and SOM in two ways: top-down and bottom-up.

Steps:
(1) Present the formulas for TAM, SAM, and SOM both top-down (apportioning from official statistics) and bottom-up (stacking assumed number of customers x unit price x frequency).
(2) Assign the numbers used in each formula, and label every one of them as an "assumption."
(3) For each number, always add candidate sources for "which official statistics or data should be used to verify it."
(4) Reconcile the two estimates, and where they diverge, state which is more likely to be an over- or under-estimate.

Output format:
- Summarize the "assumed numbers" in a table (columns: variable / assumed value / candidate source for verification).
- Summarize "TAM, SAM, SOM" in a table (columns: layer / formula / amount (assumption) / top-down or bottom-up).
- End with the "top 3 assumptions that tend to be overestimated," and for each, how to revise it conservatively.

Constraints:
- Make clear that every number is an assumption. Do not fabricate statistics or figures that do not exist; where you are unsure, stop at proposing candidate sources.

Business: [ ]
Assumed unit price: [ ]
Target region and customer segment: [ ]

Before a pitch, have it play a tough investor and poke holes. The point is not to ask for answers, only questions.

You are a tough venture capitalist. Read my pitch outline and, before making an investment decision, poke at the points that must be closed out, in the form of sharp questions.

Steps:
(1) From the following 10 angles, create one question each that is most likely to leave me stuck for an answer: whether the problem is real / the validity of the target / evidence that the solution addresses the problem / why now / competition and substitutes / overwhelming advantage / the realism of the market size / profitability (unit economics) / the execution team / the next validation.
(2) For each question, add one line on "why this question lands."
(3) Do not ask for answers; give questions only.

Output format:
- A numbered list of 10 questions. Format each entry as "Question" then, on a new line, "Why it lands: ..."
- End by restating the "top 3 questions I would most struggle to answer," each with the reason.

Constraints:
- Ask questions that probe specific numbers, premises, and situations, not vague ones (e.g., "Will it really sell?"). Do not fabricate facts.

Pitch outline: [paste the key points of the nine steps]

To use prompts like these well, you need a feel for what and how much to hand over to AI, and where to keep your own hands on the wheel. If you want to align this feel across your team, or want a partner to accompany you all the way through building the business, please reach out from an individual consultation. We support making AI a partner in your business from the standpoint of a company that has actually launched businesses of its own.

Walking through the nine steps with a fictional business, "Nominokoshi"

Abstraction alone leaves no tactile sense. So let's walk a fictional new business all the way through the nine steps. Every number is an assumption. When you build it for real, always verify with primary sources and interviews.

The subject is "Nominokoshi," a service for watching over the medication of a parent who lives far away.

1. Who. We put the child's generation, working professionals in their forties to fifties in urban areas, as the decision-maker and the payer, and the parent living alone in the provinces, aged 70 or older, as the user. The early adopters are the meticulous members of the child's generation who worry about their parent's medication and check by phone almost every day.

2. What problem. The job is "I want to feel at ease that a parent who lives far away is taking their medicine properly, without having to check by phone every single time." They fear health deterioration from missed or double doses. The existing substitutes are a handwritten medication calendar, a daily safety-check phone call, and a hundred-yen-shop pill case. None of them conveys a "record of what was taken" to the family.

3. Value. In one line: "peace of mind that you can quietly watch over a parent's medication without calling." The high-level concept is "a watch-over camera for medication."

4. Solution. The top three features are: the pill case lights up to notify at dose time; an open-close sensor detects the dose and automatically records it in the family app; and if a missed dose continues for two hours, it notifies the family. The MVP runs in 10 households using an off-the-shelf pill case, a commercial open-close sensor, and manual notifications via a messaging app, to confirm whether families genuinely feel at ease.

5. Business model. A monthly subscription: the device is loaned free, at 1,480 yen a month. We make the "gap"—the child's generation pays, the parent uses—explicit from the start. Channels are referrals via care managers and community comprehensive support centers, plus social media ads. We validate whether customer lifetime value exceeds acquisition cost, centered on the referral channel.

6. Competition. Major watch-over sensors and cameras, carriers' watch-over apps, and existing medication-management apps. However, options that integrate "medication," "family sharing," and a "dedicated device" are scarce, and many families substitute with phone calls and calendars. That is the opening.

7. Advantage. Real data on medication adherence—whether the medicine is actually being taken properly—and trusted channels with care managers and pharmacies. App features can be copied, but trust and referral routes via the local network of professionals are assets that are hard to imitate in the short term.

8. Market size. For TAM, we assume roughly 7 million elderly people living alone in Japan and 18,000 yen a year, for about 126 billion yen (the number of single-elderly households needs verification against official statistics). For SAM, assuming 30% of the child's generation pay by smartphone, about 38 billion yen. For SOM, we bring it down to several thousand households via care managers in a specific area in the first year, on the order of tens of millions of yen up to about 100 million yen.

9. Pitch (30 seconds). "Do you check your parent living alone on their medication by phone every day? Nominokoshi lets you watch over their medication without calling, with a pill case that lights up to notify and a family app. Missed doses are automatically sent to the family. Right now we're testing with 10 households via care managers in a specific region, with a retention rate of X percent."

Walking through it this way, you can see that the deliverables from the nine steps are, as they stand, the building blocks of the pitch. All we did was rearrange the material accumulated in steps 1 through 8 into an order that lands. That is the reward for doing the work in order.

Summary

Finally, let me organize what I wanted to convey in this article.

• Build a new business starting from the customer and the problem. Start from the solution or the idea, and you finish something nobody wants.
• Keep one Lean Canvas at hand and write the results of the nine steps into it as you go, so the whole picture doesn't drift.
• The first half is the customer and the problem, the middle is the solution and the business model, and the second half is why you can win, how big it is, and how to convey it. Grasp it in these three blocks.
• AI is powerful as a first-drafter and a critic, but verify your numbers with primary sources and your problems with customer interviews.
• The deliverables from the nine steps become the skeleton of your business plan and pitch deck as they are.

A new business is not a one-shot gamble on talent or inspiration. Form a hypothesis in the right order, check it with customers, and update it. The precision of this unglamorous repetition is what separates success from failure. Start today: in the "Who" box of step 1 in the sample PowerPoint template, write your idea in just one sentence. The moment that first single person comes into focus, the business starts to move. When you want to go deeper into each step, follow the links in the table to the detailed article for each.

---

References

Related articles

• Step 1: Setting your new-business target and narrowing down early adopters
• Step 2: Digging into the customer's problem (Jobs-to-be-Done and the empathy map)
• Step 3: The Value Proposition Canvas for stating your value
• Step 4: Generating solutions and ideas, and validating with an MVP
• Step 5: Shaping the business with the Business Model Canvas
• Step 6: Competitive analysis and how to find existing substitutes
• Step 7: The positioning map and an advantage that is hard to copy
• Step 8: Estimating market size (TAM, SAM, and SOM)
• Step 9: How to put together a pitch that lands