Jessie McRae

Flishworks · Current operating business

Building an AI-enabled business end to end

Transforming the business, not just the Product.

Build in progressOperating Product · active learning · open questions

90-second case

A business problem, treated as one connected system.

Problem
A high-judgment service can benefit from AI, but faster generation does not by itself create a trustworthy Product or a viable business.
Mandate
As founder, I am building and operating the customer Product, service workflow, offer, payments, measurement, and acquisition direction as one connected system.
Strategic choice
Define the customer promise, truth standards, human authority, and commercial journey before maximizing automation.
What exists
A customer-facing AI-enabled Product and service experience; packaged offers; payment and checkout architecture; GA4, PostHog, ecommerce, and server-side purchase instrumentation; and explicit human-review principles.
Current state
The operating foundation exists. The active work now is Product quality, customer outcomes, acquisition, unit economics, reusable capability, and the next Product direction.
Why it matters
This is the current environment where I can make—and be accountable for—decisions across Product, Experience, AI behavior, commercialization, and operating constraints.

Three layers of the work

Build the Product. Transform the system around it. Productize only what earns reuse.

The three layers move at different speeds: operating work, system change, and reusable capability.

01

Build

Current state
Operating now
Purpose
Create a useful AI-enabled Product and experience. Define the customer job, offer, workflow, AI-assisted creation, truth standards, checkout, payments, and human approval as one usable Product system.
Current focus
Improve customer value, Product quality, acquisition, and the economics of delivery.
02

Transform

Current state
Active work
Purpose
Redesign the surrounding work and operating model. Connect human and AI responsibility, quality, measurement, acquisition, commercialization, adoption, learning, and economics instead of treating AI as an isolated feature.
Current focus
Use instrumentation and operating changes to improve quality, throughput, customer outcomes, and economics.
03

Productize

Current state
Long-term experiment
Purpose
Extract reusable capability only after repetition is proven. Use repeated operating patterns to decide whether a capability belongs in Foundry, remains business-specific, or earns a distinct anonymous Product direction.
Current focus
Learn which capabilities improve quality, speed, safety, cost, or learning across more than one Product.

01 · Operating business

A real customer job creates Product, service, and commercial constraints at once.

The current operating environment centers on a consequential career-positioning job: understand a person’s source evidence, clarify a credible market position, create coherent artifacts, and preserve human authority over what is true and strategically useful.

Current operating model

From source evidence to an approved customer artifact

AI provides leverage in analysis and creation; human judgment remains accountable for positioning, evidence, quality, and approval.

Why it matters: The Product is designed around responsibility and evidence—not generation alone.

Human + AI allocation

Automate repeatable work. Keep consequential authority explicit.

WorkAI contributionHuman authority
Source evidenceOrganize and synthesize supplied materialProvide, clarify, and validate the record
PositioningGenerate and compare candidate directionsChoose the market position and strategic emphasis
Artifact creationDraft, rewrite, tailor, and compareJudge specificity, truth, usefulness, and coherence
RevisionPrepare alternatives from feedbackResolve tradeoffs and approve consequential changes
Final authorityNo independent authorityJessie and the customer retain approval

Decision

Package the customer problem before maximizing automation.

Why it was hard
The work contains repeatable structure, but the value still depends on evidence, positioning, and judgment that cannot be reduced to output volume.
My judgment
Define clear customer jobs and deliverables first, then decide where AI can improve speed, consistency, or quality without weakening authority.
Consequence
The build now has a legible offer and operating model; customer use will determine what changes next.

Decision

Treat unsupported claims as Product failures, not editorial cleanup.

Why it was hard
Plausible language can look strong while changing ownership, inventing evidence, or overstating a customer’s record.
My judgment
Separate source fact, strategic interpretation, and generated expression; retain human responsibility for truth, positioning, quality, and approval.
Consequence
Truth and approval became Product requirements, with quality evaluation and recovery designed into the workflow.

02 · Transformation work

Change the Product and the work system together.

The agenda is deliberately broader than adding AI features. It connects the customer promise, workflow, quality, measurement, commercialization, economics, and learning—while keeping the state of each stream clear.

01

Product + workflow

What exists today
Customer problem, offer structure, AI-assisted creation, human judgment, payments, and the commercial journey are treated as one Product system.
What the work has exposed
Generation is abundant; the difficult Product decisions are evidence, positioning, quality, and authority.
What I am testing
How much of the workflow can become repeatable without making consequential judgment implicit.
Next move
Map the deployed workflow and document where people intervene, revise, approve, or recover.
02

Quality + evaluation

What exists today
Truth standards and human approval are explicit Product principles.
What the work has exposed
Fluent output is not a sufficient definition of quality for consequential career material.
What I am testing
A useful quality model spanning grounding, role relevance, specificity, consistency, and customer acceptance.
Next move
Create a versioned evaluation set, failure taxonomy, denominators, and recovery behavior.
03

Commercialization + measurement

What exists today
Offer architecture, pricing, checkout, analytics, and acquisition direction operate as one commercial system.
What the work has exposed
A funnel can be observable without yet explaining what intervention will improve it.
What I am testing
Which signal gaps and interventions deserve scarce founder attention.
Next move
QA the event dictionary, establish cohorts and measurement windows, and use an experiment to change a decision.
04

Economics + learning

What exists today
The operating questions are explicit: price, human effort, model and infrastructure cost, acquisition, revisions, support, and fees.
What the work has exposed
Automation percentage is not the business objective; successful execution and learning are.
What I am testing
Where effort and value accumulate across the Product and service workflow.
Next move
Build an operating ledger that connects cost and effort to pricing, scope, automation, and scaling decisions.

03 · Marketing + acquisition

From acquisition analytics to an adaptive growth system.

The immediate problem is concrete: understand traffic and channels, QA the events, diagnose the funnel, choose an intervention, and measure what changed. Only then does the larger question become useful: how might an AI-enabled Marketing system increasingly sense, decide, act, and learn?

GA4Page and ecommerce events documented
PostHogAutocapture, funnel behavior, and masked replay documented
StripeIdempotent server-side purchase event documented
Experiment recordNo public dated intervention/result ledger yet

Implemented and proposed states separated

From observation to a measured acquisition decision

Instrumentation is in place; the next step is to run a focused intervention and use the result to decide what changes.

Why it matters: The acquisition direction begins with signals, diagnosis, and decisions rather than content volume.

Decision

Instrument the commercial journey before attempting meaningful acquisition scale.

Why it was hard
Traffic or content volume cannot explain whether the offer, funnel, purchase flow, or customer journey is the limiting constraint.
My judgment
Connect GA4, PostHog, ecommerce events, masked replay, and server-side Stripe purchase tracking so the operating system can support focused experiments.
Consequence
The measurement architecture is in place; event QA and focused interventions come next.

In place

Measurement foundation

GA4, PostHog, ecommerce events, masked replay, and server-side Stripe purchase tracking provide the current measurement foundation.

Next to learn

Intervention performance

The next intervention will test which change improves customer progression through the acquisition journey.

Longer-term direction

Selective AI assistance

AI may eventually detect, analyze, recommend, and prepare. People retain approval over consequential external and investment decisions.

04 · Foundry / long-term experiment

What makes an AI Product factory effective?

Foundry is a reuse experiment beneath the current build: which capabilities should remain specific to one business, and which should become shared only after another real Product shows the advantage?

Keep business-specific

Customer proposition + Product behavior

Offer, pricing, workflow, domain rules, content, specific AI behavior, and the Product experience.

Observe repetition

Candidate reusable capability

Shared contracts + infrastructure

Only capabilities with demonstrated cross-Product value should move toward a reusable layer.

What exists

This portfolio’s design system has semantic tokens, content-independent primitives, explicit variants, and an export layer designed for future registration.

What is being designed

A clear separation between business-specific proposition, workflow, AI behavior, and visual identity—and capabilities that may eventually be reusable.

What comes next

Use a second real Product to learn whether reuse improves implementation time, cost, quality, or risk.

Read the Foundry reflection

05 · Anonymous Productization path

From field learning to reusable Product strategy.

The question is whether a repeated operating problem can produce a useful capability—and whether that capability earns investment as a separate Product. The direction remains anonymous because it is not a mature public Product.

  1. 01Operating problem

    Operating the current business raises a recurring question: can market signals be turned into better, more deliberate acquisition decisions without creating a content machine?

  2. 02Repeated pattern

    Look for the same operating problem across real interventions.

  3. 03Reusable capability

    Turn repeated patterns into shared capability when reuse improves the work.

  4. 04Possible Product

    If that problem repeats and a useful capability survives real interventions, it may justify a standalone Product direction.

What exists now
A distinct Product idea is taking shape inside the current work while customer exposure and reusable capability develop.
Decision already made
Keep the direction anonymous and inside the Flishworks chapter until the operating pattern—not a name or proposed architecture—earns a separate case.
What would justify investment
A real Product artifact, customer exposure, differentiated learning, and an explicit investment decision.

06 · What I’m working on next

Customer use will determine where the work goes next.

The priorities are better customer outcomes, reliable quality, sustainable economics, stronger acquisition, and learning whether repeated infrastructure should become a reusable Product.

  1. 01
    Product state + use

    Confirm deployment, customer availability, cohort, package, and completion state.

  2. 02
    Quality + trust

    Measure grounding, failure types, revision burden, approval, and recovery.

  3. 03
    Customer value

    Define a bounded outcome and distinguish Product contribution from external factors.

  4. 04
    Acquisition learning

    Publish a dated experiment ledger with denominators, windows, results, and decisions.

  5. 05
    Operating economics

    Measure active human time, model and infrastructure cost, fees, revisions, support, and acquisition.

  6. 06
    Foundry reuse

    Reuse a capability in a second real Product and compare time, cost, quality, and maintenance.

  7. 07
    Productization decision

    Put a real artifact in front of customers, capture differentiated learning, and make a continue, pivot, or stop decision.

Continuity, not origin story

Earlier work provides standing; Flishworks is the current test.

Current direction

A real business makes the theory answerable: what should AI do, where must people remain accountable, which signals matter, and what deserves to become reusable?

Customer use will determine which operating decisions create value and which capabilities deserve to become reusable.