NYU Leslie Entrepreneurship Institute Spring 2026 AI Systems · Program Design

Startup Intelligence System: An AI Discovery Partner an Entire Cohort Adopted in Two Weeks

How I gave 18 founder teams the same AI discovery workflow the best founders build for themselves, got an entire program to run on it in two weeks, and designed it so every founder could use it whatever their skill level.

Live deployment. Spring 2026 Startup Sprint cohort. Quantitative outcomes are being consolidated now. The qualitative signal is already in.

18 teams Cohort served, Spring 2026 Startup Sprint
~900 interviews Supported across the cohort, roughly 50 per team
10 skills, 2 tracks Custom commands and access paths for every skill level
Best Use of AI Award typically reserved for participating teams

The Problem

Every Startup Sprint team runs about 50 customer interviews in two weeks. Across the cohort that is roughly 900 conversations. The volume has never been the hard part. The hard part is knowing which assumption to attack next and which question will validate or kill it. What a team needs to learn shifts every week, sometimes every day.

The founders who do discovery well keep a living map of their riskiest assumptions and design each interview around them. The most sophisticated founders we coach already use AI to close the loop between what they hear and what they ask next. Most teams do not, and the gap widens by the month.

The failure modes are predictable. Founders hear what they want to hear and discard the signal that complicates it. Insights from early interviews decay before they matter. Teams drift from the method the moment a coach leaves the room. Setting up the tooling to fix all of this is a steep onboarding cliff, and under a two-week deadline most teams fall back to sticky notes and gut feel.

So I built the setup once and made it the floor for the entire cohort, live on Day 1.

"We really need to create a centralized mind within our own team just so we can really be on the same page with our understanding of the problem and therefore our vision of the solution."

— David Cui, Agora

Before and After

Before
Manual coaching check-ins
No persistent AI support between sessions
No structured evidence tracking
Teams rebuilt context from scratch every session
Workshops captured on flipcharts and phone photos
After
Every team has a persistent AI thinking partner that reads their interviews
Assumptions tracked against a four-level evidence framework
Workshops designed and run AI-first
Coaches working inside each team's workspace
Context that compounds across the full two weeks

The Hardest Design Problem

The Spread

The hardest problem here was never the AI. It was the range of people who had to use it.

The cohort ran from founders who had been building AI products for years to founders who had never opened a terminal. A system that only works for the top of that distribution is a perk for the already capable. So I set the least technical founder as the design target and let everyone else inherit the same system.

Three decisions followed from that one constraint.

Two parallel tracks on a single methodology

Technical founders run Claude Code in the terminal, connected to Notion over MCP. Everyone else uses Claude in the browser with the Notion connector. Same workspace, same skills, same outputs, different door in.

Setup stripped to three steps

Duplicate a Notion template, drop in a config file, run one command. Nothing else stands between a founder and a working system.

A silent start protocol

Claude reads the team's context, initializes it if empty, and installs any missing skills before the founder types a word. A founder's first interaction is a working system, not an empty prompt box.

The result was near-universal adoption across the cohort with almost no technical issues. For a two-week program where every hour counts, the reliability was a design outcome, not an accident.

How It Works

Two layers, ten skills, zero happy ears

01
Two-layer infrastructure

Notion is the team's persistent workspace and memory. Claude is the reasoning layer that reads and synthesizes across it. The Notion MCP connects them with live read and write access. A team's interviews, assumptions, and business model all live in one place that Claude can actually reason over.

02
Methodology encoded as skills

The Institute's discovery framework is built directly into the system as ten custom skills. A founder runs one command instead of writing a prompt from scratch. The flagship skill maps all of a team's interview evidence against a four-level discovery ladder and returns a self-contained read, every claim cited to a specific interview.

03
Built to resist happy ears

The most common way discovery fails is founders hearing what they want. The system never validates and never celebrates. When a team reads its results optimistically, Claude surfaces the specific evidence that complicates the picture and asks the question that points at the contradiction.

Tools and Stack

What I built it with

Claude Code
Terminal access for technical founders. Full skill support and persistent project context.
Claude.ai
Browser access for everyone else. Notion connected as an integration, skills used as prompt workflows.
Notion
Persistent workspace, interview database, session memory, and the read-only methodology library.
Notion MCP
The live read and write bridge between Claude and each team's workspace.
Custom skill system
Ten installable commands that turn the discovery methodology into one-command workflows.

What I Actually Built

The system was half the work. The other half was getting people to adopt it.

I saw the opportunity, built the case, and got buy-in from a team that had no reason to assume an AI layer would help. Then I built the whole thing, tested it, and trained every user myself: all 50 founders and every coach. I did it for the Startup Sprint, the most high-pressure program the Institute runs, where there is no slack in the schedule to recover from a bad rollout.

The coaches are the signal I am proudest of. They started skeptical, the way good coaches should be about anything that touches their craft. By the end of the sprint they had redesigned their workshops to run AI-first, because the system raised the floor on what every team walked in with.

This is the second time I have taken an AI system from idea to full organizational adoption at the Institute, after the Leslie coaching copilot. The pattern I trust now: the build is the easy half, and getting an organization to reorganize around the build is the rare one.

Signal So Far

Quantitative outcomes from the cohort are being consolidated now and will be added here. The qualitative signal is already clear.

Recognized with a "Best Use of AI" award, an honor typically reserved for the participating teams rather than staff.
Invited to submit a proposal on the system to the Global Consortium of Entrepreneurship Centers (GCEC), currently under review.
Coaches redesigned their workshops to run AI-first around the system.
Founders left the program meaningfully more AI-native, many having not known a setup like this was possible.
Near-universal adoption across every skill level in the cohort, with almost no technical issues.

Design Principles

01 Design for the floor, not the ceiling. Setting the least prepared user as the target is what produced universal adoption. A system built for the power user only ever serves the power user.
02 Make every output self-contained. A founder who has never read the framework can act on the result, because the insight travels with it and references nothing external.
03 Reflect evidence, stay out of the way. The system surfaces contradictions and cites its sources. It never tells a founder what to do next unless asked. Under time pressure, unsolicited advice reads as noise.
04 Build the method and the tool in the same loop. The methodology was written into the system while the system was being built and used. Ship, learn, revise, ship again, the same loop we ask founders to run.