Four Innovation Agents
Every tool in the Innovation OS so far has been on-demand: you invoke /discovery when you need synthesis; you invoke /pitch when you need a narrative; you invoke /sprint when you need a sprint plan. This is reactive intelligence — useful when you know what question to ask.
The four innovation agents work differently. They run continuously, monitor your venture's key indicators, and surface the decisions that need attention before you ask. Each Monday, your Idea Generator delivers unsolicited provocations. Your Customer Intelligence agent synthesises last week's feedback without being prompted. Your Business Model Architect flags when validated data contradicts the canvas. Your Fundraising Readiness agent tracks the investor pipeline and prepares you for every meeting 48 hours in advance.
This is the shift from reactive to proactive intelligence — from answering your questions to asking questions you have not thought to ask.
This lesson orchestrates everything. The Idea Generator builds on /idea. The Customer Intelligence agent builds on /discovery. The Business Model Architect builds on /canvas and /financials. The Fundraising Readiness agent builds on /pitch. You have used all 10 commands. Now meet the four agents that run them continuously.
Skills vs. Agents
Before exploring the agents, understand how they differ from the skills you have been using:
| Dimension | Skills (Commands) | Agents (Persistent) |
|---|---|---|
| Activation | You invoke them when you need them | They activate on schedule or trigger |
| Frequency | On-demand; once per invocation | Weekly automated + on-demand |
| Context awareness | Reads innov.local.md if you provide it | Reads innov.local.md automatically every invocation |
| Output | Single-task response | Continuous intelligence over time |
| Primary use case | Specific synthesis, analysis, or creation | Pattern monitoring, proactive surfacing, readiness |
| Example | /discovery synthesises 10 interviews now | Customer Intelligence agent synthesises all week's feedback on Monday |
Skills are tools. Agents are team members with standing briefs.
Agent 1: The Idea Generator
Purpose: Maintain a steady flow of fresh ideas into the venture's innovation pipeline. Prevent ideation from becoming a bottleneck — where the team only thinks about new directions when forced to by a crisis.
Weekly task — Monday Innovation Brief:
Every Monday, the Idea Generator loads your innov.local.md, identifies the current stage and open assumptions, searches for relevant market signals from the past week, and delivers three unsolicited ideas — small provocations to keep thinking fresh without requiring a formal brainstorm.
MONDAY INNOVATION BRIEF — Week of [Date]
════════════════════════════════════════════════════════════
THIS WEEK'S 3 UNSOLICITED IDEAS:
(Stage: MVP | Assumptions being tested: A-004, A-005)
IDEA 1: WhatsApp-to-ERP auto-sync
What: Automatically sync approved invoices from our WhatsApp workflow
directly to the customer's accounting software (Xero, QuickBooks, Sage)
Why now: Two pilot customers mentioned manual re-entry as a friction point
this week. This is the integration step that could unlock enterprise.
Effort to test: 1 week spike to assess API availability for top 3 ERPs
Assumption it would test: A-007 (ERP integration feasibility)
IDEA 2: CFO WhatsApp nudge for at-risk vendors
What: Proactively flag when a vendor payment is overdue based on PO date —
notify CFO via WhatsApp before vendor calls.
Why now: Pilot 1 CFO mentioned a vendor dispute this week; early warning
could prevent that category of friction entirely.
Effort to test: 1 sprint; builds on existing WhatsApp notification pipeline
Assumption it would test: A-009 (proactive notification adoption)
IDEA 3: Dual approval threshold
What: Auto-approve invoices below $1,000 (after first human approval per vendor);
only route exceptions above threshold to CFO.
Why now: CFOs have mentioned approval fatigue for small invoices.
Effort to test: 3 days; configuration change in approval logic
Assumption it would test: A-006 (CFO approval fatigue for small invoices)
THIS WEEK'S MARKET SIGNAL:
A regional bank has announced an SME finance digitisation programme
that will require AP audit trails from their SME lending clients.
This is a potential distribution partnership — financial institutions
as a sales channel. Worth tracking.
UNCOMFORTABLE QUESTION:
We have 3 paying pilots but 6 months in, Pilot 2 still only uses the system
for 60% of their invoices. We have never asked them directly: what is
preventing the other 40%? Is it a workflow problem, a trust problem,
or a feature gap we are not seeing?
════════════════════════════════════════════════════════════
Monthly task: Review the idea backlog in innov.local.md. Flag ideas that have become more relevant due to new learnings. Flag ideas that have become irrelevant. Identify any idea that should move to /hypothesis for formal assumption mapping.
On-demand: 100-idea sprint facilitation, pivot ideation after a failed assumption, analogy library generation ("What would Stripe do with this problem?").
Agent 2: Customer Intelligence
Purpose: Maintain a continuous feedback loop between customers and the team. Ensure no customer signal goes unsynthesised. Update customer profiles and assumption statuses in innov.local.md as new data arrives.
Weekly task — Customer Signal Digest:
Every Monday, the Customer Intelligence agent synthesises all customer signals from the past week — NPS responses, support tickets, direct messages, usage data, interview notes, sales call notes, and churn events.
CUSTOMER SIGNAL DIGEST — Week of [Date]
════════════════════════════════════════════════════════════
CUSTOMER HEALTH SNAPSHOT:
Active customers: 3 | Healthy (>70% usage): 2 | At risk (<50% usage): 1
NPS this week: 8.5 (2 responses) — stable vs. last week
Churn this week: 0 customers ($0 MRR impact)
TOP 3 THEMES FROM CUSTOMER FEEDBACK:
THEME 1: Approval fatigue for small invoices — 2 mentions this week
Evidence: Pilot 2 CFO: "I get 40 WhatsApp notifications a day. Most are
under $500. I just want to auto-approve those."
Implication: A-006 (approval fatigue) is likely VALIDATED — needs a sprint.
THEME 2: ERP re-entry still manual — 2 mentions this week
Evidence: Both Pilots 1 and 3 manually re-enter approved invoices into
Xero. One mentioned "this takes 20 minutes a day."
Implication: ERP integration may be a retention risk, not just a feature request.
THEME 3: New finance staff onboarding friction — 1 mention
Evidence: Pilot 3 new hire took 3 days to become proficient independently.
Implication: Onboarding assumption A-008 may need revision upward.
ASSUMPTION UPDATES THIS WEEK:
A-006 (approval fatigue): Evidence from 2 customers this week supports
VALIDATING this assumption. Recommend moving to TESTING by designing
auto-approve threshold feature in next sprint.
CUSTOMER AT RISK:
Pilot 3 (Regional Manufacturer): Usage dropped from 67% to 52% this week.
Likely related to the 3 new staff hires and onboarding friction noted above.
Recommended action: Founder calls the CFO today — not to sell, to listen.
NEW CUSTOMER INSIGHT THIS WEEK:
Finance teams treat WhatsApp as an approval tool and email as the
record of invoice receipt. They want both channels — WhatsApp for
speed; email for the audit trail they can forward to their accountant.
Our current system only covers one side of this.
════════════════════════════════════════════════════════════
On-demand: Interview synthesis (applies /discovery structure), NPS driver analysis (segments promoters/passives/detractors), churn analysis (stated + inferred reason + assumption impact), persona maintenance.
Agent 3: Business Model Architect
Purpose: Maintain the Business Model Canvas and financial model as living documents that reflect current validated state — not the original vision. After every significant validation event, propose specific canvas updates. Provide regular health assessments.
Trigger-based: The Business Model Architect activates when new validation data arrives. It does not wait for you to ask. When the Customer Intelligence agent identifies a theme that affects a canvas block, the Business Model Architect proposes a canvas version update.
VERSION 1.4 UPDATE PROPOSAL — [Date]
Block: Revenue Streams
Current: $500/month flat SaaS subscription (all company sizes)
Proposed: Tiered pricing — $500/month (mid-market, >$5M revenue);
$350/month (SME tier, $1M–$5M revenue)
Evidence: Pilot 1 CFO (SME) mentioned price sensitivity at $500/month.
Two LOI conversations stalled on price point. Tiered pricing
could accelerate SME pipeline without cannibalising mid-market.
Status: PROPOSED — approve to update canvas to v1.4
Monthly task — Financial Health Review (first Monday of the month):
MONTHLY FINANCIAL HEALTH REVIEW — [Month/Year]
════════════════════════════════════════════════════════════
HEADLINE: 🟢 HEALTHY — Unit economics strong; runway comfortable
UNIT ECONOMICS:
CAC (measured): $275 vs. target $300 🟢
LTV (projected): $36K vs. target $30K 🟢
LTV:CAC: 131:1 vs. target 10:1 🟢
Payback period: 1 month vs. target 3 mo 🟢
Monthly churn: 0% (3 months; too early to measure) 🟡 WATCH
CURRENT STATE:
MRR: $1,500 (3 customers × $500)
ARR: $18,000
Paying customers: 3
Monthly burn: $4,400
Cash on hand: $83,200 (month 2 of bootstrap)
Runway: 19 months
RUNWAY ALERT: 🟢 19 months > 6 months threshold
CANVAS HEALTH:
Most validated blocks: Value Proposition (pilot evidence), Revenue Streams (3 paying customers)
Least validated blocks: Channels (only founder-led sales), Key Partnerships (no ERP integrations)
Stale assumptions: A-007 (ERP integration, 45 days untested), A-009 (proactive notifications, 60 days untested)
RECOMMENDED ACTIONS:
Priority 1: Run a quick test on ERP integration feasibility (A-007 is 45 days stale)
Priority 2: Design auto-approve threshold to test A-006 (validation from customer feedback)
════════════════════════════════════════════════════════════
On-demand: Alternative business model exploration, runway recalculation after burn rate changes, sensitivity analysis ("what happens to runway if churn is 25% instead of 15%?").
Agent 4: Fundraising Readiness
Purpose: Ensure the founding team is always fundraising-ready — not scrambling when a warm introduction arrives, but prepared with a current pitch narrative, data room, and investor pipeline.
Continuous monitoring — Readiness Score:
The Fundraising Readiness Agent evaluates readiness against a 16-item checklist whenever fundraising.current_round is populated in innov.local.md:
FUNDRAISING READINESS SCORE — [Date]
DATA ROOM READINESS:
✓ Pitch deck (current; <4 weeks old) COMPLETE
✓ Financial model (current; <2 weeks old) COMPLETE
✓ Cap table COMPLETE
✓ Customer LOIs / contracts (3+ preferred) COMPLETE (3 pilots + 6 LOIs)
✗ Team bios (1 page each; achievement-focused) NOT STARTED
✓ Product demo or screenshots COMPLETE
✓ Reference customer list COMPLETE
NARRATIVE READINESS:
✓ 30-second elevator pitch (memorised; natural)
✓ 3-minute pitch (practised ≥5 times)
✗ 15 hard Q&A answers prepared and practised NOT STARTED
✓ Financial model walked through out loud
INVESTOR PIPELINE:
✗ 20+ target investors identified and researched NOT STARTED (only 4 identified)
✓ 5+ warm introductions in progress
✓ 0 investors in pipeline >14 days without update
READINESS SCORE: 12/16 items complete
Recommendation: Minimum threshold (12/16) reached. Can begin active
outreach but address team bios and Q&A prep before first investor meeting.
Weekly task (during active fundraising) — Investor Pipeline Update:
INVESTOR PIPELINE — Week of [Date]
════════════════════════════════════════════════════════════
HEADLINE: 4 investors in active process | 1 at diligence | 2 at intro meeting
PIPELINE STATUS:
Vertex Capital DILIGENCE [Date] Send data room [Date+3]
Angel X INTRO MEETING [Date] Follow-up email [Date+2]
Regional VC INTRO MEETING [Date] Schedule follow-up [Date+5]
Seed Fund Y OUTREACH PENDING — Send intro [Date+7]
AT RISK: None >14 days without contact
THIS WEEK'S PRIORITY:
1. Send data room to Vertex Capital (diligence started — time-sensitive)
2. Schedule follow-up call with Angel X (strong signal from intro meeting)
════════════════════════════════════════════════════════════
On-demand tasks:
Pre-pitch ritual (48 hours before any investor meeting): The agent generates a personalised investor brief — their portfolio, known investment thesis, past questions from public sources, and the 5 most likely hard questions for this specific investor. This prevents the avoidable mistake of pitching to an investor whose portfolio includes your direct competitor without knowing it.
Pitch practice feedback: When you share a transcript of your verbal pitch, the agent identifies what was strongest, what was weakest, what investors will not believe without more evidence, and the one thing to fix before the next practice.
Monthly investor update draft: For investors or advisors in the pipeline, the agent drafts a 400-word factual, honest update. The format: one-sentence headline, MRR/customers/runway, two wins, one real challenge with a plan, two specific asks.
Configuring All Four Agents for the AP Venture
Use this prompt to generate a Week 1 output from all four agents simultaneously:
I want to see what Week 1 looks like with all four innovation agents active.
My venture context is loaded in innov.local.md (AP automation SaaS;
MVP stage; 3 paying pilots; $500/month each; 6 LOIs; raising $500K seed).
Generate:
1. Monday Innovation Brief (Idea Generator) — 3 unsolicited ideas,
1 market signal, 1 uncomfortable question
2. Customer Signal Digest (Customer Intelligence) — using the pilot
feedback from our exercises as the week's customer signals
3. Financial Health Review (Business Model Architect) — using unit
economics from Lesson 9
4. Fundraising Readiness Score (Fundraising Readiness) — using the
pitch deck and Q&A from Lesson 12
The four agents translate directly to the intrapreneurship context with one key adaptation — the Fundraising Readiness Agent becomes the Business Case Readiness Agent:
| Fundraising Agent Element | Business Case Agent Equivalent |
|---|---|
| Investor pipeline | Stakeholder alignment tracker (CHAMPION / NEUTRAL / RESISTANT) |
| Pitch deck | Business case document |
| Data room | Evidence pack: pilot results, financial justification, risk assessment |
| 15 hard Q&A answers | Objections from CFO, Legal, IT, and HR |
| Readiness score | Gates: business case approved → pilot approved → scale approved |
| Pre-meeting investor brief | Pre-meeting stakeholder brief: what they care about, likely objections, the ask |
The Monday Innovation Brief and Customer Signal Digest work identically — internal users are customers; their feedback goes into the same digest structure.
Try With AI
Use these prompts in Cowork or your preferred AI assistant.
Reproduce — Generate the Monday Innovation Brief for the AP venture:
Generate a Monday Innovation Brief for the AP automation venture.
Context: MVP stage; 3 paying pilots ($500/month each); currently testing
A-004 (WhatsApp integration feasibility) and A-005 (adoption improvement).
Recent signals: Pilot 2 CFO mentioned approval fatigue for small invoices;
Pilot 3 experiencing onboarding friction with new staff hires.
Output:
- 3 unsolicited ideas (each with: what, why now, effort to test,
assumption it would test)
- 1 market signal relevant to the venture this week
- 1 uncomfortable question the team should sit with
What you are learning: The Monday Innovation Brief is valuable because it arrives whether or not you would have asked. The uncomfortable question is the most powerful element — it surfaces the thing the team is avoiding. Notice whether the ideas generated are specific to the AP venture context or generic.
Adapt — Generate a Customer Signal Digest from hypothetical feedback data:
Generate a Customer Signal Digest based on this week's customer feedback
for a B2B SaaS venture at MVP stage (3 pilot customers).
Feedback received this week:
- Pilot 1 NPS: 9 (comment: "saves me 2 hours a week")
- Pilot 2 usage data: down from 75% to 60% adoption this week
- Pilot 3 support ticket: "I can't figure out how to add a new vendor"
- Sales call with LOI prospect: "Your pricing is too high for our stage"
- Churn event: a trial user (not yet paying) said they're going with a competitor
Synthesise into: health snapshot, top 3 themes, assumption updates,
at-risk customers, and one new customer insight.
What you are learning: The digest forces you to interpret multiple signals simultaneously. Notice how the usage drop in Pilot 2 and the pricing objection from the LOI prospect might be related — or might not be. The agent's job is to surface the pattern; your job is to decide whether it is signal or noise.
Apply — Use your venture's innov.local.md for a Financial Health Review:
Based on my innov.local.md, generate a Monthly Financial Health Review.
My current data:
- Unit economics: [From your L09 exercise]
- Current MRR: [Your actual or estimated MRR]
- Monthly burn: [Your estimated burn rate]
- Cash on hand: [Your starting capital]
- Runway: [Calculated]
- Canvas validation status: [Most and least validated blocks from your L08 exercise]
Include: unit economics with RAG status, current state, runway alert,
canvas health, and 2–3 recommended actions.
What you are learning: The monthly review reveals which assumptions in your financial model are measured vs. assumed. Every metric marked ASSUMED is a number you are betting on without evidence. The review makes those bets visible — before they become expensive surprises.
Flashcards Study Aid
Continue to Lesson 15: Capstone — Build Your Innovation OS →