Segment and job
Which microsegment has the strongest pain, urgency, budget, reachability, and outsourcing fit? Which recent event makes that buyer act now?
First move · Six-week sprint
Validate behavior, not enthusiasm.
Published research says the problem category is real. It does not say which buyer will trust a new specialist firm, approve the proposed scope, or pay the proposed price. This sprint compares two narrow segments and ends with paid decisions—not a pile of agreeable interviews.
20–24 qualified interviews
5 offer reviews
2 paid purchase paths
8–12 founder hours / week
The decisions
Eight questions must be answered
Wedge and price
Will the buyer fund a read-only-first Reliability Baseline and then a $25k–$45k managed project? Do both follow real budget and approval paths?
Trust and coverage
What proof and access controls are required? Is co-managed business-hours coverage enough, or does the target buyer require primary 24/7 response?
Delivery, recurrence, and channel
Can projects clear a 55% gross-margin floor? Which post-project work is temporary versus recurring? Does the founder sell directly or through a credible partner?
A completed interview is not a success metric. A recent event, prior spend, budget-owner access, proposal review, and paid commitment form an evidence ladder.
— Commercial research ruleExecution runbook
One controlled loop from public signal to paid evidence
AI compresses preparation and analysis. The founder controls every transition where bad data, trust, money, or a commitment could change the result.
Source candidates
Normalize public signals, referrals, companies, roles, and possible triggers.
Verify fit
Confirm the trigger, cohort, budget context, relationship path, and exclusions.
Reach out
AI drafts from verified facts; the founder edits, sends, and handles every reply.
Interview
Reconstruct a real event, consequence, spend, buying path, and next action.
Code evidence
AI structures notes and proposes scores; the founder verifies every strong claim.
Act
Every five interviews: continue, change one test, make a paid ask, or stop.
Week 1: build the machine before increasing volume
Set the boundaries
Review employment constraints, exclusions, consent, storage, and deletion. AI can turn the approved rules into checklists; it cannot interpret the obligations.
Done: one signed-off safeguard checklist.
Build and rehearse the kit
Create the tracker, interview record, concept card, consent text, and scheduling flow. Run a 25-minute mock interview before contacting buyers.
Done: version 1 of every research artifact.
Seed the sample
Founder supplies 25–30 warm names and relationship paths. AI normalizes and deduplicates records, leaving unknown fields blank.
Done: balanced seed list with no hidden conflicts.
Find and verify triggers
AI proposes public signals with source URLs. The founder spends no more than three minutes verifying fit, trigger, cohort, and conflict status per record.
Done: 50 outreach-ready records from 80 candidates.
Release wave 1
Select 20–25 balanced prospects. AI drafts from approved facts; the founder edits and sends. Follow-ups are queued for days 4–5 and 10–12.
Done: every active record has a dated next action.
Stop bad scaling
Automate a handoff only after it works correctly five times. Review source quality, cohort balance, replies, and uncoded interviews every Friday.
Done: no interview remains uncoded overnight.
Sampling design
Two cohorts, analyzed separately
Broad vertical lists create false averages. Each cohort needs at least eight qualified interviews before results are pooled or compared.
B2B software and software-enabled services
- 75–300 employees and 10–60 engineers
- Customer-facing production product
- Azure material; AWS acceptable
- No mature platform team
- CTO, VP/Head of Engineering, or infrastructure leader with budget context
Regulated services with an in-house software estate
- Healthcare, insurance, or fintech services
- Material internally operated application or data platform
- Comparable company and team size where possible
- Recent audit, customer, availability, or recovery obligation
- Buyer who understands both the consequence and approval process
Up to four additional interviews may cover leaders who bought, rejected, or replaced a similar service in the last 24 months. Those interviews are labeled separately and do not inflate pain frequency.
A trigger is required before outreach
Platform role open 45+ days
Recent incident
Audit or enterprise-customer deadline
Cloud-budget pressure
Migration or Kubernetes adoption
AI workload entering production
One overloaded infrastructure owner
Evidence hierarchy
Buyer behavior carries more weight than buyer opinion
| Level | Evidence | What it means |
|---|---|---|
| 0 | Opinion about the category | Useful language; no demand evidence |
| 1 | General complaint | Pain may exist, but may never earn priority |
| 2 | Specific event in the last 12 months | The problem is concrete and recent |
| 3 | Quantified consequence, deadline, or executive escalation | The problem competes for attention |
| 4 | Prior spend, active budget, or failed attempt | The organization has acted before |
| 5 | Budget-owner introduction or offer review | The buying system is opening |
| 6 | Procurement starts or a paid baseline is approved | Commercial demand |
Interview method
Past events first. The offer comes last.
Thirty minutes are booked and twenty-five planned. The proposed offer is shown only in the final four minutes, after the buyer’s real situation and buying process are understood.
1 · Qualify the operating system (3 min)
What runs in production? Who owns cloud, deployments, observability, incidents, and recovery? Which of those areas can this participant approve spend for?
2 · Reconstruct the trigger (8 min)
Walk through the last incident, failed release, audit request, or budget surprise. Capture who lost time, revenue, confidence, or sleep—and what happened next.
3 · Examine alternatives (5 min)
What did the team try: hiring, MSP, consultant, tooling, or internal re-prioritization? What did it cost? What remained unresolved?
4 · Map buying and trust (5 min)
Who owns budget and veto? What proof is required for read-only access and production change? Is primary 24/7 response mandatory?
5 · Review the baseline and managed project (4 min)
Show one page with the $12k–$18k baseline and $25k–$45k managed-project follow-on. Ask which budgets and buyers apply, what would be unsafe, and what should be removed. Then ask which activities would be needed for only 90 days and which would be needed indefinitely. Ask for the next real action—not “would you buy this?”
Captured within 15 minutes
Demand
Trigger, date, consequence, deadline, attempted alternatives, prior spend.
Purchase
Budget owner, category, approval path, trust requirements, price-path evidence.
Next action
Offer feedback, stakeholder introduction, security review, proposal, or paid start.
Automation boundary
Let AI move information. Keep human judgment at the gates.
The fastest safe workflow is a relay: AI prepares a structured draft, the founder verifies it against the source, and only then does the record move forward.
AI draft or calculation
Human judgment
Quality control
Owner
Prospect
Interview
Evidence
Commercial
AI
Normalize, deduplicate, source trigger candidates, draft outreach
Prepare brief, transcribe with consent, flag missed topics
Extract fields, propose codes, calculate rates, surface contradictions
Draft recap, proposal outline, charts, and findings brief
Founder
Choose sources, verify trigger and fit, approve contact, handle replies
Build trust, probe the story, test price, request the next action
Approve quotes and scores, interpret causality, choose the next test
Set scope and price, negotiate, make the paid ask, decide green/yellow/red
Gate
Source URL + date + one-sentence verification
Explicit recording consent; unsupported facts removed
Numerator, denominator, interview IDs, and negative cases
Only buyer behavior counts as a commitment
AI can say “the transcript suggests level 4 evidence.” Only the founder can verify the quote, approve the score, and decide whether the buyer actually revealed a budget path.
— Human-in-the-loop evidence ruleFive automations worth building
Draft the candidate record
Extract only source-backed fields, attach the URL and date, suggest a cohort and trigger, and send the row to a founder approval queue.
Human gate: fit, conflict, triggerPrepare the interview
Create the interview record and a T−24 brief with verified facts, two hypotheses, three unknowns, and at most two contextual follow-ups.
Human gate: factual reviewStructure the evidence
Extract fields, mark quote versus inference, propose H1–H8 codes, expose unknowns, and draft the thank-you note.
Human gate: approve levels 3–6Run cohort synthesis
Recalculate rates, cluster language, list support and counterevidence, flag missing data, and draft the dashboard with denominators.
Human gate: audit two recordsOpen the commercial path
Draft a tailored recap and stakeholder-review request from approved evidence. The founder models effort, sets the fee, and personally asks for the paid baseline.
Human gate: scope, price, promise, sendAnalytics
Compare demand and deliverability
Opportunity score
- Frequency20%
- Severity20%
- Urgency / forcing event15%
- Prior spend or budget15%
- Outsourcing fit15%
- Founder credibility and reachable channel10%
- Repeatable delivery potential5%
Commercial dashboard
- Recent-trigger and quantified-consequence rates
- Prior-spend or active-budget rate
- Interview → offer review → committed next step
- 24/7 requirement and trust-blocker rates
- Expected sales cycle and acquisition channel
- Modeled baseline, managed-project, and stabilization hours and gross margin
- Temporary versus recurring post-project work
Every percentage includes a cohort denominator: “6 of 10 qualified Cohort A interviews,” not an unqualified “60%.”
Schedule
Six weeks, with commercial pressure by week three
| Week | Founder focus | AI / automation | Evidence target |
|---|---|---|---|
| 1 | Approve safeguards and kit; verify triggers; send 20–25 messages | Build draft rows, deduplicate, source public signals, queue follow-ups | 80 candidates; 50 ready; balanced cohorts |
| 2 | Run 4–5 interviews; review records; send wave 2 | Prepare briefs, structure notes, propose codes, maintain funnel | First trigger and trust patterns |
| 3 | Run 5–6 interviews; hold checkpoint; invite offer reviews | Calculate cohort metrics; surface contradictions; summarize offer feedback | 10+ cumulative; 1–2 offer reviews booked |
| 4 | Run 5–6 interviews; test buyer, budget, price, and partner paths | Compare cohorts; draft tailored recaps; flag sampling gaps | 16+ cumulative; three offer reviews |
| 5 | Close sample gaps; model delivery; make paid asks | Draft economics scenarios, proposal outlines, and evidence charts | 20–24 interviews; five offer reviews |
| 6 | Meet stakeholders; verify findings; sign the decision | Draft anonymized brief and green/yellow/red memo | Two paid paths or an explicit pivot |
The repeatable week
MondayApprove the funnel, cohort gaps, interview queue, and outreach wave.
Tue–ThuCluster interviews. Protect 15 minutes immediately after each for coding.
WednesdayVerify new trigger candidates and release the next outreach batch.
FridayReview evidence, contradictions, open actions, and denominators.
Every 5thRun a 60–90 minute synthesis and audit two source records.
Suggested 8–12 hour split: 4–5 hours interviews and coding, 2–3 hours prospect verification and outreach, 1–2 hours synthesis, and 1–2 hours commercial follow-up.
After the first projects
Productize inside the delivery team
Discovery can approve the first baseline and project. It cannot justify a separate subscription team. The people doing the work must remain responsible for learning what repeats.
Reserve 10–15% for productization
Build templates, automation, acceptance checks, onboarding, change control, handoff, and effort classification from real delivery—not an imagined service catalog.
Use stabilization to observe recurrence
Measure which tasks repeat, how often they occur, the skill and coverage required, incident load, and margin. End the retainer when the need is temporary.
Four to six similar clients
Do not split the team until 70–80% of work is standardized and recurring gross profit funds two engineers plus a 20% coverage buffer.
Retention, handoff, margin, and runway
Require two renewal cohorts, at least 55% gross margin, proven escalation and backup coverage, six months of payroll runway, and no client above 25% of recurring revenue.
Decision gates
The sprint ends with a decision
Green · Sell baseline + project
One cohort has eight qualified interviews; 60% report a recent event; 40% show spend, budget, or deadline evidence; five offer reviews yield two paid paths; co-management works; and modeled baseline/project gross margin is at least 55%.
Yellow · Extend one test
Pain is strong but scope, price, buying path, or channel is unclear. Run 8–10 more interviews in one cohort and change only one major variable.
Red · Reshape
Fewer than two paid paths; body-only demand dominates; primary 24/7 is mandatory; trust is unattainable; or realistic project economics fall below the 55% floor. This rejects investment in a recurring team—not necessarily a project consultancy.
What a pivot could preserve
Project-only consultancy
Keep the diagnostic and bounded remediation as the business without forcing a recurring model.
Channel-only practice
Deliver reliability work through an established MSP with existing trust and coverage.
Narrower Azure offer
Focus on recovery readiness, cost governance, or another repeatedly funded job.
The standard is paid evidence.
The sprint is successful if it identifies a segment and earns two real purchase paths—or rejects the current model early enough to preserve time and capital. A separate recurring team is a later operating decision, not a discovery-stage commitment.
Return to the company concept →