Tiffany Johnson

Busy owners drown in overnight email, and nothing surfaces what actually needs a person before the day starts.

A backend that connects to mailboxes and each morning produces one page: the overnight messages that genuinely need a person, the commitments already made, and what is due today, with a single recommended focus.

Model output is treated as untrusted: every call has a typed contract, validation, and a safe fallback, so failures stay contained. Rules run before the model to control cost, and every extracted item is traceable back to its source.

Live and in daily use across my own applications. It has grown from email triage into a daily workspace with to-dos, quarterly goals, AI spend budgets with alerts, and two-factor sign-in. I retired the ask-your-inbox question box after the brief proved the more useful surface.

Generic outreach gets ignored, and personalizing it by hand does not scale.

A tool that discovers independent retailers, scores how well each fits a brand, and drafts outreach tailored to that store, with a human approval gate before anything is sent. It tracks the whole funnel and protects deliverability with bounce suppression, rate limits and one-click unsubscribe.

A person reviews and approves every message. Generation is auditable and reviewable, and the system never sends autonomously, a deliberate design choice, not a limitation.

Weak product listings lose sales, but rewriting them for search and conversion is slow and inconsistent.

An engine that pulls a live listing and generates optimized copy mapped to each field, in a web app with billing in place.

Grounded in real listing data rather than free generation, with model output coerced into a strict schema the product validates, and quality judged rather than assumed.

Sellers rebuild the same product story by hand for each marketplace, and the content modules are tedious to write.

An installable Shopify app that maps an existing product page onto a marketplace's content modules and generates the copy and image direction for each one.

Model output is validated against a strict schema, usage is metered by a credit ledger, and every generation is kept in a history the seller can review.

Sellers who list on both their own store and a wholesale marketplace can quietly break pricing rules or lose margin when prices drift.

An embedded Shopify app that watches price changes over webhooks and flags any price that violates the marketplace's parity rules or falls below a configurable margin floor.

It alerts the seller to act and never rewrites prices on its own, so a person stays in control of pricing.

Licenses, insurance, domains, certificates and scheduled jobs expire or fail silently, and nothing raises an alarm until something breaks.

One countdown of everything that can lapse, with early escalating reminders, heartbeat monitoring that treats a missed check-in as the failure, and public checks for site reachability, certificate expiry and domain expiry.

Nothing is installed on the systems watched and no credentials are taken. An AI step that reads renewal notices runs offline on purpose, so the always-on service has no model dependency.

Live and in daily use monitoring my own applications and sites. Not yet opened to outside users.

AI assistants move fast but need clear ownership, a shared record and hard limits, or work gets duplicated and nothing is accountable.

A team of named agents, each with a written job description and one lane, working from a shared release board and structured handoffs. Two different AI assistants split the work, so one builds and the other independently reviews.

Release gates are enforced: database migration first, staging on the exact commit, an independent combined QA pass, and my explicit authorization before anything reaches production. Agents can read and report, but cannot merge, deploy or touch production data on their own.

Beta. In use now across my own applications and still being refined.

Store releases break the path from product to cart to checkout, and the problems are usually found by customers.

A release-checking service that walks the customer journey on desktop and mobile and returns a report of what passed and failed, with screenshot evidence, without completing a checkout.

Store ownership is proven with a temporary verification tag, tenants are separated, and it supports human QA with evidence rather than replacing it.

In development. Live scans stay disabled while the isolated browser worker and its safeguards are completed.