Jay West
I build the AI agents and data systems that run marketing operations.
By day I lead marketing engineering inside a multi-location specialty retailer. On my own time I build in the open: two curricula about thinking, and this site, a working lab where AI agents do the volume in supervised loops and code holds the quality bar.
Running a multi-location business and need a system built? See consulting.
Two projects, one question
How to think, and how to share a thought
Thinking alone
How to Think: a curriculum for thinking alongside AI
How one person notices what they believe, works it into something usable, and decides what to keep for themselves when a machine can answer for them.
howtothink.aiWhere they meet
The overlap is the interesting part: a thought you can hold clearly on your own is one you can hand to someone else, and building on what someone else hands you sharpens your own. That is also what working well with AI asks of a person.
Thinking together
Physics of Connection: how to share a thought
How a thought travels between people: listening, reading a room, building on what someone offers instead of waiting to talk.
physicsofconnection.comSelected work
Case studies, with the numbers behind them
How to Think: a curriculum for thinking alongside AI
Thinking taught as a practice rather than a reading list: every lesson is one idea with an exercise, a named failure mode, and a step that carries it into the reader’s week.
74%
average view duration on the first lesson video (3:12 of 4:18)
Physics of Connection: how to share a thought
Improv, taught as the mechanics of communication: typed concepts, lessons as paths through them, and a first video that held viewers for 74% of its length.
13 → 2
checks that passed against an empty tree
Evals for AI-generated output, and testing the tests
Every defect class found in generated pages became a check that fails the build. Then an audit asked which checks would pass if there were nothing to check.
1,103 → 45
median page-title length, characters
Why a large content site drew zero search traffic
Four mechanical causes, each measured, three fixed in code, and one that changed the plan.
1
missing setting found that was silently dropping every lead
A health endpoint that told the truth about the funnel
Every form rendered and every page converted, and nothing was being kept. Nobody could see that until the system reported on itself.
0
near-duplicate pages, measured across every pair on the site
Running a content operation on supervised AI agent loops
Claude agents doing the volume in narrow loops, a person approving every change, and quality enforced by code rather than by reading.
Context engineering: a doctrine every writing agent loads
Five competing schools of copywriting, graded on their evidence and resolved into one set of rules an agent can follow.
2 of 32
offers that paired a dollar outcome with a meter you can bill on
Choosing what to sell before spending on it
Every offer scored on demand, delivery effort and whether its outcome can be billed. One wedge chosen, then paid search ruled out before a dollar went into it.
Skills
What each case study proves
AI systems
Data and technical stack
Working with people
The full skill map links every skill to its proof.
How I work
Four rules the work holds to
Measure before fixing
Rule causes out with data before choosing one. The search diagnosis eliminated duplication and link starvation before touching anything.
See it in practiceGuard what you fix
Every defect class found becomes a check that fails the build, and the checks are tested to prove they can fail.
See it in practiceAgents propose, people approve
Agents do the volume. Every change they make is a small pull request a person reviews before it ships.
See it in practiceSay what the data does not show
Write-ups state plainly when a result has not arrived yet, and retract earlier conclusions in the open.
See it in practice
Now building
What I am building next
Each of these becomes a case study when it has a measured result.
An agent other teams rely on every day
A recurring reporting narrative agent for a marketing team, with a weekly usage and accuracy log.
Python for data and agents
A lead research and routing agent written in Python with pydantic models and a pytest suite.
SQL, warehouse modelling, and dbt
A marketing data model joining paid media, web analytics, and point-of-sale data by location.
Lead enrichment, scoring, and routing
The same Python agent as above, writing to a CRM with an eval set of labelled leads.
Experiment design and A/B infrastructure
A local-marketing test across matched store groups, analysed as a difference-in-differences.
Adoption and change management
Weekly active use and time saved, tracked for every internal tool shipped this quarter.
Measured results for paying clients
Missed-call recovery for home-service businesses, priced on the bookings it produces.
Building AI into marketing or revenue operations?
Tell me the role. The reply comes from me.
Prefer LinkedIn? Message me there.