Alan Seideman · Head of AI · VP of Product

I turn AI confusion into aligned execution.

I run product organizations, and I ship production AI with my own hands. Twenty-five years shipping at scale, six years building AI systems myself. I find the bet that moves the business — when the org is chasing the impressive one — and get everyone behind it.

I need help with AI → I’m hiring a leader ↓

The problem

Executive pressure to do AI, with no agreement on what.

Pilots that impress and never ship.

A team that can't tell which bets matter.

Selected work

Four engagements, in the order I'd want you to read them.

CPO
2016–2019
10MM MAU

OCC — Mexico's largest jobs marketplace

Twenty years of dominance eroding. The founder gone, the acquisition fresh, no product strategy, 45-day release cycles. Our in-house matching was precise and thin on coverage — and when the group stood up a central AI recommendations service, the data-science team resisted becoming its consumer.

I championed adopting it anyway, because coverage was what recruiters were paying for. Then I replaced the 45-day cave-build cycle with roughly fifteen autonomous squads, each on a deploy stack it controlled.

The harder problem sat at the top of engineering. Unlocking it required a leadership change; SEEK asked me repeatedly to assess the situation, and I gave it months before reaching a conclusion. I then ran product and tech on an interim basis and rebuilt trust from the inside — weekly 1-1s with every tech lead, moving them from waiting to be told what to build to owning their domains.

The culture was the deeper constraint: a rigid top-down hierarchy where openly critiquing a director was unthinkable, being merged across a border post-acquisition. My instrument was a weekly newsletter to both OCC and SEEK leadership that reported failed experiments as plainly as wins. It became the group's benchmark for product-team transparency — cited by senior executives and by the CEO as the standard he wanted other markets to match.

+30%

YoY applications

8×

release velocity

1→20

product org

ADVISOR
ONGOING
TAX PRACTICE

A national financial-advisory firm

Executive pressure to adopt AI inside the Office of the CFO of a 1,300-person, 18-office advisory firm, and an entirely non-technical tax practice being offered tooling nobody there could maintain.

I diagnosed that what they needed was sustainable no-code automation, built credibility with live demos instead of decks, and sequenced a three-phase roadmap: workbook roll-forward, return-to-provision reconciliation, trial-balance categorization.

100s

hours projected saved,
firm-wide

One AI-assisted review surfaced a material error in a live client workbook. The debate about whether this was real ended that afternoon; the engagement has since expanded from one sponsor to a second workstream.

ADVISOR
HEALTH, FITNESS
& WELLNESS

Three health & wellness organizations

The same pattern in three sizes: everyone wanted to talk about the AI product, and in each case the money was somewhere else.

NASM

$100MM ARR
CERTIFICATION

Advised the head of product at a fitness-certification firm on AI-assisted product strategy — which consumer AI opportunities to fund first, and how to frame the roadmap so the executive team would actually back it.

Herbal Medics Academy

4 WEB PROPERTIES

The founder was ready to fund a speculative AI build. I ran AI-driven analysis across four web properties instead, surfaced conversion and top-of-funnel gaps manual review had missed for years, and traded the build for a roadmap ordered by return. The unglamorous work paid better.

Healthgena

100+ STARTUPS
VISION 2030

Resident AI and product advisor for Saudi Arabia's Vision 2030 HealthTech accelerator — coaching founders on where AI belonged in the product, user need, value proposition, business model, and three-year feasibility.

SENIOR PM
2012–2016
25MM MAU

Funny or Die

A household-name comedy brand losing ground to BuzzFeed. Expanding into snack-size media and video recommendations drove a record 25MM MAUs — and then the metrics stopped meaning what everyone assumed.

Engagement was climbing while revenue stayed flat, because ad CPMs were collapsing industry-wide. A business monetized on ads couldn't win on time-on-site alone. I reframed the strategy from keeping users on our site longer to going where the audience already was, and built the mobile-apps team from zero to do it.

Among what shipped: Dips, a full-screen short-video app for comedy that anticipated the format TikTok would take mainstream, and the Weather app my leadership was lukewarm on until our lead Sequoia board investor singled it out as the standout concept.

25MM

record monthly actives

+15%

MoM time on site

+20%

YoY, from the
mixed-media shift

Why I still build

By 2021 this was clearly the largest shift in technology since the mobile revolution, and moving far too fast to understand from the outside. Building with AI changes the product development paradigm, and you can't set strategy against a paradigm you've only read about.

Prototyping is nearly free now, so the expensive parts moved: evaluation, context design, and the maintenance bill on anything the frontier will do better next year. Six years of shipping production systems myself is how I know which of those a given bet will run into — so I can tell a team where to invest, where to buy instead of build, and what to refuse.

Two of my own products, as the evidence.

Meldoria

A multi-step LLM pipeline: assessment webhook, psychometric scoring, multi-domain completions, dynamic PDF report generation. Then a living journal — a week of entries returned as a chapter of your own story, written against your psychometric profile. Alongside it, 222 published articles from an automated content factory, drawing 150K+ search impressions a quarter.

Big Invoice: a dark invoices dashboard showing $67,536.72 received of $71,372 invoiced, 95% paid, above a list of client invoices

Big Invoice: an estimates screen tracking draft, pending and accepted totals above a filterable list of client estimates

Big Works / Big Invoice

iOS invoicing for the trades, built solo in Flutter. A million tradespeople sold bloated software that all looks like accounting — in the part of the economy least exposed to AI displacement. Big Apps. Tiny Price. Black and yellow against a category of blue.

Live on the App Store →

Before

All nine retired

Agent 01

38 steps

Agent 02

38 steps

Agent 03

38 steps

Agent 04

38 steps

Agent 05

38 steps

Agent 06

38 steps

Agent 07

38 steps

Agent 08

38 steps

Agent 09

38 steps

Nine separate agents, 38 steps each. Every new persona meant cloning and maintaining another full copy.

After

Live in production

17 thin webhooks
3 steps each

Shared core engine

92 steps

One engine, every persona — coach, companion, educator.

Adding a persona went from a 38-step clone to a 3-step webhook. Nine personas became seventeen while the system got simpler to maintain.

The lesson

The expensive part of an AI system is rarely building the first one — it's the maintenance bill on the tenth. Duplicating a working agent nine times is the fastest way to build and the most expensive way to own. Collapsing them into a shared core with thin per-persona routers is the decision that made the system survivable.

The estate

49 workflows in production · ~22 services · one person

Assessment pipeline
Content factory
Agent layer
Memory tiers
Social video pipeline
Email & notifications
Payments

This is the part that changed how I think about org design. The question stopped being how many people a function needs, and became what the smallest team plus the right automation looks like.

Dispatcher Video Generator Publisher

Most recent build, shipped this month: a social video pipeline with separated dispatch, generation, and publishing stages.

Writing

21 SECTIONS
ILLUSTRATED

Primer

Working with AI

What these tools actually are, why they slow down, and how to get noticeably more out of them. Written for a team who'd been handed AI tools and left to get on with it — no jargon, and it applies to Claude and Copilot alike.

Read it →

91 TRENDS
2022–2031

Research

AI tools & human behavior

Ninety-one trends since ChatGPT launched, each scored for how hyped it got against how long it actually lasted — then what that pattern implies for 2027 to 2031. Built to separate the trends worth planning around from the ones worth ignoring.

Read it →

Services

AI Opportunity Audit

I audit the org, the strategy and the products, and come back with where AI creates value here, where it doesn't, and one working demo.

3 WEEKS
FIXED FEE

AI Transformation Roadmap

The sequenced plan, a working proof-of-concept, and the internal narrative that gets leadership aligned behind it.

4–6 WEEKS
FIXED FEE

Fractional Head of AI

I run the AI program for you — the roadmap, the build-or-buy calls, the vendor and data decisions, the team’s capability — as your Head of AI on a fractional basis.

1–2 DAYS/WEEK
RETAINER

Fractional Product Leadership

I run product for you — strategy, roadmap, the team's operating rhythm — until you're ready to hire the permanent seat.

1–2 DAYS/WEEK
INTERIM OR ONGOING

Advisory Retainer

A standing call and async access, for a product leader who wants a second read on the AI decisions.

MONTHLY
RETAINER

About

I've spent twenty-five years learning to tell a real bet from an impressive one.

Sometimes that meant making the unpopular call and being right: adopting an AI service my own team wanted to build, and lifting applications 30%. Sometimes it meant killing my own work — I decided not to build a coaching chatbot because frontier models were going to outrun anything I could maintain.

I've also been on the other side of it. I shipped a follow-a-topic network as Twitter launched and the location check-in two years before Foursquare. Being early taught me what being right actually requires, and I've been selling that distinction ever since.

I also own a direct-to-consumer products company I founded in 2014 — profitable every year on organic demand — which keeps me honest about margins, inventory, and what marketing actually converts.

I work in Spanish as readily as in English — I ran a 400-person Mexican company's product organization in it — so I can lead or advise LATAM teams directly, with no layer of translation between the strategy and the people executing it.

Remote-first. MBA, University of Denver. B.S., Haverford.

Book a call → Download resume