Experiment 03 · Canada · Product-Market Discovery

We built the product. The harder question was: who should care first?

Instead of choosing an ICP in advance, FlyPig used its own process to research who in Canada had the strongest workflow fit. Then we ran the same business problem through a fresh high-capability LLM without FlyPig's process rules.

ICP shiftBROAD_ADOPTER_DISCOVERY → SERVICE_MULTIPLIER_FIRST

The strongest first-wave targets became firms that repeatedly perform prospect research for clients, plus AI/CRM implementers that can reuse the workflow across many customers.

01

The business problem

This is the same problem many new products face after launch: the product exists, but “who should we sell to first?” is still unresolved.

Possible ICP

Direct users

Founders, sales teams, export teams, consultants, lead-generation firms, or market-entry providers could all plausibly use prospect research.

Possible multiplier

AI / CRM implementers

These firms could embed the workflow into CRM, qualification, research, or sales-automation projects across multiple clients.

Possible ecosystem

Communities and amplifiers

Accelerators, training organizations, founder communities, and ecosystem channels can influence adoption even if they are not direct buyers.

02

What FlyPig found

Archetype A

Direct workflow practitioners

Market-entry firms, outsourced prospecting teams, partner-matching consultancies, territory-research providers, and international BD services already perform this work repeatedly.

Archetype B

Implementation multipliers

AI and CRM firms with explicit lead qualification, sales copilots, CRM enrichment, outreach, data validation, or guarded automation showed stronger adoption logic.

Mission v1
BROAD_ADOPTER_DISCOVERY
→
Workflow fit
Repeated leverage
→
Mission v2
SERVICE_MULTIPLIER_FIRST
03

The FlyPig shortlist

The biggest cuts came from rejecting broad “AI company” adjacency and asking whether the actual workflow matched.

27
material identities
20
fully researched
18
initially QUALIFIED
13
final SHORTLIST_READY

Five initially qualified AI-related firms were downgraded because their public evidence did not clearly show prospect research, target-account selection, lead qualification, CRM/sales workflow, or outreach/follow-up work.

04

Then the naked LLM found something FlyPig had missed

The control run was strong — and it explored a broader set of launch paths.

≈30
organizations considered
15
final control prospects
0
formal mandatory audit
0
formal dedup procedure
What the control added

Broader ecosystem thinking

It considered accelerators, training organizations, founder communities, events, and other amplification channels more aggressively than the original FlyPig run.

That was useful enough to change the product. FlyPig added a Discovery Diversity Check so an ambiguous campaign must test materially different routes before converging too early.
05

The business takeaway

When a new product does not yet have a clear ICP, research should not merely “find companies.” It should help discover which type of customer, partner, or multiplier deserves the first sales conversation.
What this result does not claim
The 13 shortlisted organizations were evidence-backed targets for a first adoption conversation. The experiment did not test actual replies, adoption, budget, or willingness to change workflow.
PUT THIS TO WORK

Built something valuable, but still unsure who your first real customers should be?

FlyPig can treat ICP discovery itself as a research problem — then build the first evidence-backed target set around the strongest adoption paths.

FlyPig AI · Page updated: