Hundred Ideas, One Hour
In Lesson 3, you defined the problem worth solving. You interviewed CFOs, synthesised their frustrations into a ranked pain map, and arrived at a crisp "How Might We" statement: how might we give CFOs the confidence to face their auditor without AP anxiety?
That problem statement is now your creative brief. The next step — ideation — is where most innovation teams underperform. In a typical brainstorm, the same 10-15 ideas surface every time. The creative range of the session is bounded by the diversity of the room and the willingness to say things that sound strange. AI removes these constraints. Given a clear problem statement, it can generate ideas across every dimension — incremental and radical, technical and non-technical, feature-level and business model level — faster than any team.
The 100-idea sprint is not just about volume. It is about structure. Without categories, 100 ideas become 10 ideas repeated with slightly different wording. With 10 categories, you explore 10 fundamentally different lenses on the same problem — and the best ideas, almost always, are in the second half.
Why Brainstorms Underperform
The academic literature on group brainstorming is unequivocal: unstructured group brainstorming consistently produces fewer and lower-quality ideas than the same number of people working alone, then sharing. The culprits are well-understood — production blocking (you cannot say your idea while someone else is talking), evaluation apprehension (fear of judgment), and social loafing (assuming others are doing the hard thinking).
The 100-idea sprint sidesteps all three mechanisms. The AI has no social pressure and no evaluation apprehension. It does not run out of ideas when the obvious ones are exhausted. And the 10-category structure forces it into creative territory that group brainstorms never reach.
| Unstructured Brainstorm | 100-Idea Sprint |
|---|---|
| Same ideas every session | 10 categories force 10 distinct creative lenses |
| Quality peaks early | Best ideas in categories 6–10 |
| Social dynamics limit range | No evaluation apprehension |
| Stops at 20-30 ideas | Discipline to reach 100 |
| No filter for follow-on action | DVF scoring creates an actionable shortlist |
The insight embedded in the 100-idea target: your first 50 ideas are not bad ideas — they are the ideas you were already carrying. The second 50 are where original thinking begins.
The 10 Categories
The sprint structure divides 100 ideas across 10 categories, 10 per category:
| Category | The Question It Answers | Why It Matters |
|---|---|---|
| Product | What could be built? | Core feature-level thinking |
| Business Model | How could it make money differently? | Revenue model innovation is often more powerful than product |
| Distribution | How would customers find and buy it? | Channel strategy is a competitive moat |
| Analogy | What worked for similar problems elsewhere? | Cross-industry pattern transfer accelerates idea quality |
| Technology | How could new tech (AI, mobile, IoT) change the solution? | Forces specific tech application, not vague "AI-powered" |
| Crazy / 10x | What would 10x better look like? | Suspends disbelief; reveals unstated assumptions |
| Boring | What is the most obvious solution? | Grounds the sprint; often misses market; reveals baseline |
| Contrarian | What is the exact opposite of the obvious solution? | Challenges incumbent assumptions; finds blue-ocean space |
| Partnership | Who could solve this if they combined forces? | Surfaces distribution leverage and complementary assets |
| Incumbent | What would the market leader do if they noticed this? | Defines the competitive response and required differentiation |
The Crazy and Contrarian categories feel uncomfortable. They are supposed to. The point is not to generate implementable ideas — it is to expand the solution space before collapsing back to a shortlist. Ideas that survive from these categories are often the ones that become category-defining businesses.
DVF: The Three-Dimensional Filter
Once you have 100 ideas, you need a systematic filter that is faster than "does this feel right?". DVF — Desirability, Viability, Feasibility — gives you three independent lenses:
Desirability — Would customers pay for this today?
- GREEN: Clear willingness to pay; customer is already spending on an adjacent solution
- YELLOW: Pain is real but payment behaviour is not established
- RED: Nice to have; customers tolerate the problem without urgency
Viability — Could this become a significant revenue business?
- GREEN: Clear path to scale; large addressable market; strong unit economics
- YELLOW: Uncertain scale; niche market or unclear monetisation at volume
- RED: Too small a market; unit economics problematic at scale
Feasibility — Can a small team build an MVP in under three months?
- GREEN: Core technology is available; team has or can access required skills
- YELLOW: Three to six month MVP; one significant technical risk
- RED: Requires proprietary technology or hardware; more than six months to first test
The shortlist targets ideas with at least two GREEN scores and no RED scores. But DVF is a filter, not a formula. Exceptional founder-specific advantages — deep domain knowledge, existing customer relationships, proprietary data — can override a YELLOW score.
The "Why Now?" Test
Before finalising any shortlist candidate, apply one additional filter: why now?
What changed in the last two to three years that makes this possible or urgent today?
Valid "Why Now?" answers:
- New technology (LLM capabilities, mobile penetration, API availability)
- Regulatory change (new requirement creates demand)
- Behaviour change (remote work, platform adoption, pandemic-driven shift)
- Cost inflection point (cloud costs, AI inference costs dropping)
- Demographic shift (new generation in the workforce)
- Market gap (incumbent exited, raised prices, lost trust)
If you cannot answer "Why Now?" for an idea, it may be a genuinely good idea with bad timing — or an idea that someone already tried and it did not work. Either way, timing is one of the most critical variables in venture success.
For the AP automation example: the "Why Now?" is clear. WhatsApp Business API became production-ready, LLMs brought invoice OCR accuracy to commercial viability, and mid-market CFOs — newly remote — now manage AP processes through messaging apps by default rather than by exception.
Using /idea for the 100-Idea Sprint
Worked example. You have your HMW problem statement from Lesson 3:
"How might we give mid-market CFOs the confidence to face their auditor without AP anxiety — knowing that every invoice, approval, and payment is documented, searchable, and defensible?"
You open Cowork and type:
/idea
Idea generation sprint.
Problem statement: "How might we give CFOs the confidence to face
their auditor without AP anxiety — knowing that every invoice,
approval, and payment is documented, searchable, and defensible?"
Constraint: Must work with or alongside WhatsApp.
Generate 100 ideas across these categories:
- Product ideas (10): what could be built
- Business model ideas (10): how could it make money differently
- Distribution ideas (10): how would people find and buy this
- Analogy ideas (10): what has worked for similar problems elsewhere
- Technology ideas (10): how could new tech (AI, mobile, IoT) change this
- Crazy/10x ideas (10): what would 10x better look like
- Boring ideas (10): what would the most obvious solution look like
- Contrarian ideas (10): the exact opposite of the obvious solution
- Partnership ideas (10): who could solve this if they combined forces
- Incumbent ideas (10): what the market leader would do if they noticed this
Format: numbered list, 10 per category, one sentence each.
No self-censorship — include obvious, impossible, and strange ideas.
Condensed sample output (Product and Contrarian categories shown):
IDEA SPRINT — AP Automation / CFO Audit Confidence
════════════════════════════════════════════════════════════
PRODUCT IDEAS (how the product could work):
1. WhatsApp-native approval workflow that logs every CFO response
as a timestamped, immutable audit record
2. AI invoice matching that highlights discrepancies before
they reach the approval step — not after
3. Real-time AP dashboard that CFOs can share a live read-only
link with their auditor during audit week
4. Automated vendor payment remittance emails that generate
a paper trail without manual entry
5. Invoice intake via forwarded email — no portal login required
6. "AP health score" dashboard widget: percentage of invoices
approved with full documentation vs. exceptions
7. Audit mode: one-click export of every transaction, approval
chain, and supporting document for a specified date range
8. Anomaly alerting: flag invoices that arrive from new vendors
or deviate from historical patterns before approval
9. Approval delegation workflow for when the CFO is travelling
— backup approver with full audit chain maintained
10. Mobile push notification for urgent approvals with one-tap
respond from the phone without opening a dashboard
CONTRARIAN IDEAS (the exact opposite of obvious):
71. Remove the approval step entirely for invoices under $500
from approved vendors — automate payment, log for audit
72. Make the audit trail visible to the vendor too: mutual
transparency as the product's moat
73. Charge by the invoice (not subscription) — align cost
to value for low-volume customers
74. Build for the auditor, not the CFO — sell auditor productivity
tools that happen to require the CFO's data
75. Start with the vendor side: give vendors a portal to submit
invoices in structured format, removing the CFO's OCR problem
76. Zero onboarding: connect to CFO's existing email account,
parse invoices automatically, require no behaviour change
77. Offer "AP-as-a-service" — outsource the entire AP function
with AI + human oversight; no software purchase
78. Make failure visible: real-time dashboard of invoices
NOT processed on time — accountability as the product
79. Reverse the approval flow: auto-approve, with rejection
as the exception (reversal of the current model)
80. Partner with accounting firms to white-label the tool —
make auditors the salesforce
════════════════════════════════════════════════════════════
Step 2: DVF Shortlisting
/idea
From the 100 ideas, apply the DVF filter:
1. DESIRABILITY: Would mid-market CFOs pay for this today?
2. FEASIBILITY: Can a 3-person team build an MVP in 3 months?
3. VIABILITY: Could this become a $10M+ revenue business?
Shortlist the top 10 ideas that score highest across all three.
For each: DVF scores (GREEN/YELLOW/RED each), a one-sentence rationale.
Then: recommend one idea with a two-paragraph case for why.
Sample shortlist output (two examples):
SHORTLIST — Top 10 AP Automation Ideas
════════════════════════════════════════════════════════════
IDEA 1 (Product #1): WhatsApp-native approval with immutable audit log
Desirability: 🟢 HIGH — directly solves the #1 pain from discovery
Feasibility: 🟢 HIGH — WhatsApp Business API; 3-person team feasible
Viability: 🟢 HIGH — SaaS model; repeatable per customer; scales
Rationale: Solves the core emotional job (audit confidence) via the
channel CFOs already use; no behaviour change required.
IDEA 5 (Contrarian #75): Start with vendor-side structured submission
Desirability: 🟡 MEDIUM — vendors want faster payment; less certain on fees
Feasibility: 🟡 MEDIUM — requires vendor adoption; two-sided market risk
Viability: 🟢 HIGH — platform model once vendor side reaches density
Rationale: Eliminates OCR problem at source; stronger long-term moat;
but requires two-sided adoption strategy which complicates early sales.
════════════════════════════════════════════════════════════
SELECTED IDEA: WhatsApp approval workflow + audit trail SaaS
Case: [Two-paragraph rationale based on discovery data]
Pressure Testing: The Devil's Advocate
Once you have selected an idea, you do not immediately move to building it. You try to kill it.
Pressure testing is not pessimism — it is the act of finding the objections before your investor, competitor, or customer does. Every objection the devil's advocate raises is a hidden assumption inside the idea. The pressure test converts your selected idea into the foundation for the assumption map in Lesson 5.
/idea
My chosen idea: WhatsApp approval workflow SaaS for mid-market CFOs,
with AI invoice matching and full audit trail.
$500/month subscription.
Play devil's advocate: give me the 5 strongest arguments against
this idea succeeding. Then: for each argument, what would I need
to prove to overcome it?
Sample pressure test output:
PRESSURE TEST — AP Automation SaaS
════════════════════════════════════════════════════════════
OBJECTION 1: AI accuracy may not reach 90%+ on real invoice formats
(especially handwritten, non-standard, multi-language formats)
To overcome: Demonstrate >90% accuracy on 500 real invoices
from 3 pilot customers before committing to V1 build.
OBJECTION 2: CFOs may not have authority to purchase without CEO
sign-off — lengthening the sales cycle to 3-6 months
To overcome: In next 5 discovery conversations, ask directly:
"If this existed today, could you approve the purchase yourself?"
OBJECTION 3: WhatsApp Business API costs or policy changes could
eliminate the core product channel overnight
To overcome: Build on an abstraction layer; test with one backup
channel (email/SMS) so the product survives a WhatsApp exit.
OBJECTION 4: Enterprise ERP systems may require native integration
before CFOs will trust the product with live financial data
To overcome: Test whether 3 pilots will sign without ERP integration
in the first 8 weeks; if all require it, reprioritise roadmap.
OBJECTION 5: Market may be winner-takes-all — if a large incumbent
(SAP, Oracle, BILL) enters with the same feature, we cannot compete
To overcome: Design for a customer segment the incumbents ignore
(companies below $50M revenue) and move fast enough to own the
early-market relationship before incumbents notice.
════════════════════════════════════════════════════════════
Each objection in the pressure test becomes an assumption in Lesson 5. The pressure test is not the end of ideation — it is the handoff to hypothesis.
The Incumbent Ideas category (ideas 91-100) is especially relevant if you are innovating inside an existing organisation. Your own company is the incumbent. Ask: what would your organisation do with this problem if it noticed it? The answers reveal two things: which internal capabilities you could leverage (distribution, customer relationships, data) and which internal obstacles you would face (organisational inertia, competing priorities, "not invented here" resistance). The intrapreneur version of the pressure test adds a sixth objection: "Our existing IT systems cannot support this without a 12-month integration project." What is the minimum viable version that avoids that dependency?
Exercise: Idea Generation Sprint
Type: Design Thinking — Ideate Time: 40 minutes Goal: Generate 100 ideas for the problem you defined in Lesson 3; shortlist to 10; select one and pressure-test it
From Exercise 2 (Lesson 3), you have: a discovery synthesis, a pain ranking matrix, and a primary HMW problem statement. This exercise uses that HMW as the input.
Step 1 — Confirm your HMW statement (5 minutes).
Retrieve your primary HMW statement from Lesson 3. If you produced multiple variants, identify the one that:
- References a specific customer (not "users" or "people")
- Frames a desired outcome, not a feature
- Is specific enough to direct ideation but open enough not to prescribe a solution
If your original HMW was weak, revise it now using the format: "How might we [verb] [specific customer] so that [desired outcome]?"
Step 2 — 100-idea sprint using /idea (20 minutes).
Open Cowork and type:
/idea
Idea generation sprint.
Problem: [Your HMW statement]
Context: [2-3 sentences about the customer and their world]
Generate 100 ideas across:
- Product ideas (10): what could be built
- Business model ideas (10): how could it make money differently
- Distribution ideas (10): how would people find and buy this
- Analogy ideas (10): what has worked for similar problems elsewhere
- Technology ideas (10): how could new tech change this
- Crazy/10x ideas (10): what would 10x better look like
- Boring ideas (10): the most obvious solution
- Contrarian ideas (10): the exact opposite of the obvious
- Partnership ideas (10): who could solve this together
- Incumbent ideas (10): what the market leader would do
Format: numbered list, 10 per category, one sentence each.
Step 3 — Shortlist with DVF (10 minutes).
/idea
Apply the DVF filter to the 100 ideas above.
Shortlist the top 10 by:
- Desirability: would my target customer pay for this today?
- Feasibility: can a small team build an MVP in under 3 months?
- Viability: could this become a sustainable business?
For each shortlisted idea: GREEN/YELLOW/RED per dimension,
one-sentence rationale.
Then recommend one idea with a two-paragraph case.
Review the shortlist. Does the recommendation align with your customer discovery data from Lesson 3? If the AI's top-ranked idea does not match what you heard from customers, adjust — your discovery data takes precedence.
Step 4 — Pressure-test the selected idea (5 minutes).
/idea
My chosen idea: [Describe it in 2-3 sentences]
Devil's advocate: give me the 5 strongest arguments against
this idea succeeding. For each, what would I need to prove
to overcome it?
Deliverable: Your primary HMW statement, a 100-idea sprint output, a DVF shortlist of 10, one selected idea with rationale, and a pressure test with 5 objections and their required proofs.
The selected idea and its pressure test are the direct inputs to Lesson 5 (The Assumption Stack). Each objection from the pressure test corresponds to one or more assumptions you will need to map. Save this in your Cowork session before closing.
Try With AI
Use these prompts in Cowork or your preferred AI assistant.
Reproduce — Run the chapter's worked example:
/idea
Idea generation sprint.
Problem: "How might we give mid-market CFOs the confidence to face
their auditor without AP anxiety — knowing that every invoice,
approval, and payment is documented, searchable, and defensible?"
Constraint: Must work with or alongside WhatsApp.
Generate 100 ideas across:
- Product ideas (10)
- Business model ideas (10)
- Distribution ideas (10)
- Analogy ideas (10)
- Technology ideas (10)
- Crazy/10x ideas (10)
- Boring ideas (10)
- Contrarian ideas (10)
- Partnership ideas (10)
- Incumbent ideas (10)
Format: numbered list, 10 per category, one sentence each.
What you are learning: The 100-idea sprint reveals that most of the first 50 ideas are obvious. Notice which categories produce genuine surprises — Contrarian and Analogy tend to uncover the non-obvious ideas. The structure of the output (10 categories × 10 ideas) matters as much as the total volume.
Adapt — Modify for a different industry:
/idea
Idea generation sprint.
Problem: "How might we help independent restaurant owners track
food costs in real time so that they stop losing margin without
knowing why?"
Context: Restaurant owner, small team (5-15 staff), no finance
background, manages everything on WhatsApp and a paper notebook.
Generate 100 ideas across all 10 categories. Format as above.
What you are learning: Notice how the Incumbent Ideas category changes when the incumbent is a spreadsheet (Excel) rather than enterprise software. Contrarian ideas for a small business customer look very different from those for a CFO — the constraints and the objections are fundamentally different.
Apply — Generate ideas for your own venture:
/idea
Idea generation sprint.
Problem: [Paste your HMW from Lesson 3]
Context: [2-3 sentences describing your target customer and
the world they operate in]
Generate 100 ideas across all 10 categories.
Then apply DVF scoring, shortlist to 10, and recommend one
with a two-paragraph case for why.
What you are learning: Running the sprint on your own problem forces you to evaluate whether your HMW statement is specific enough. If the output ideas are generic, the problem statement is generic — revise it until the ideas are targeted.
Flashcards Study Aid
Continue to Lesson 5: The Assumption Stack →