Hypotheses, shipped.

Tokyo Doesn't Have Parking

Tokyo and New York break routing assumptions that work everywhere else. There is nowhere to park, the last hundred feet are on foot or by bike, and a single bad stop sequence costs the driver twenty minutes. The hypothesis was that density is not a penalty to absorb. It is a modelling problem with a specific answer.

We rebuilt route construction for those geographies around parking reality, alternative transport modes, and walking sequence. Route volume rose double digits in exactly the markets where a lift is hardest to get.

Delivering Where the Map Doesn't Exist

Most of the world does not have reliable address data. The patented approach, with Alex as a named inventor, grades the driver experience to the fidelity of the GIS underneath it: where the map is good, turn-by-turn; where it is thin, the interface degrades gracefully instead of confidently lying.

Every completed delivery then writes back into a historical map, so the next route through that block is better than the last. Compounding accuracy, from the fleet itself.

Building the Tool Clinicians Actually Wanted

A mobile clinical workforce was being managed with tools built for schedulers, not clinicians. MySignify is a bespoke workforce application built around the provider's day, with efficiency and personal safety treated as first-order product requirements rather than compliance checkboxes.

Provider sentiment improved markedly and attrition fell substantially. It also carried the operational weight of transitioning the workforce to W-2, which is the kind of change that usually costs you your best people.

The Margin Is Between the Visits

In-home care economics are drive time. Clinicians were spending close to half their day between visits, and the usual fix, hiring more clinicians, makes the margin worse.

Route Logic attacks both sides. On the demand side, ML shapes the schedule itself: patients on the same block are called for complementary windows rather than whenever they ask. On the fulfilment side, best-in-class routing sequences what remains. Average drive time fell by more than half.

Three Care Models, One AI Playbook

MinuteClinic, Oak Street Health and Signify Health are three distinct business models with three different risk profiles. Rather than pick one flagship AI feature, we walked the entire clinician journey and asked where machine assistance actually removes work: ambient listening in the visit, patient scheduling, inbound fax handling and routing, care coordination, decision support.

Each of those is unglamorous on its own. Together they cut enough documentation and coordination time that clinicians could finish their day on schedule. Millions of patients sit downstream of that work.

The Underwriter's Copilot

The tempting version of this project replaces the underwriter. The version that works splits the population. We built an adjudication engine that risk-tiers mortgage applications with ML, automates the approvals that are genuinely routine, and hands the remainder to a human.

The large majority cleared automatically. For the remainder, the engine produces guided troubleshooting steps rather than a rejection, so the adjudication team spends its judgment where judgment is the scarce input. Productivity rose sharply and end-to-end application time fell by more than half.

A technology-first kitchen system

A dedicated product and engineering team, a roadmap targeting a double-digit reduction in labor cost, and a genuinely unglamorous starting point: the back-of-house system nobody outside the restaurant ever sees. We rebuilt it around the crew's actual shift, not the org's process diagram.

It generated tens of millions in operational savings and measurably improved employee experience. The same team launched the Cantinas restaurant concept and demoed two working prototypes at the annual franchise meeting.

A Digital Bank Built Inside a Brick-and-Mortar One

An established brick-and-mortar bank wanted a digital offering, which usually means a worse version of the branch on a phone. We built a native digital bank instead. A separate product with its own account opening, its own service model, and none of the legacy assumptions.

Deposit volume grew double digits. The customer reviews did a lot of the marketing.

Pricing, Rebuilt Location by Location

Pricing was set centrally and applied uniformly, which means every location was either leaving money on the table or pricing itself out of its own market. We built an ML pricing model that produces location-specific guidance across thousands of items.

Revenue rose double digits. Most of the remaining work went into fitting the guidance into the commercial team's existing workflow.

Consumer personalization at hundreds of millions of users

Advisory to the founding CEO from early product through hypergrowth: core product strategy, consumer growth, and the personalization capabilities that drove viral adoption across the major messaging platforms.

We supported the positioning and deal process through the acquisition by Snap in March 2016. Bitmoji went on to become one of Snapchat's most-used features.

Alex
Blackstock.

linkedin ↗github ↗

Alex Blackstock built AB Venture Labs around a narrow question: where does AI actually move the P&L, and where is it theater? Twenty years of shipping products at scale have made him good at telling the difference, and unsentimental about the answer.

Most recently Chief Product Officer of CVS Healthcare Delivery, owning digital product across Signify Health, Oak Street Health and MinuteClinic. Before CVS, he led product for machine learning and geospatial systems at Amazon, including the routing intelligence behind last-mile delivery worldwide. He started at McKinsey, and for over a decade advised the founding team behind Bitmoji through its acquisition by Snap.

One partner,
not a pyramid.

The next name on this page will have shipped something at scale and be able to say what it cost. Until then the seat stays visibly empty — we would rather show the gap than pad the roster.

Every engagement is run by someone who has built the thing before, not someone learning on your budget.