Systems Adoption · Enterprise AI · Cross-Organizational Delivery

Edward J. Laurent, PhD

Edward J. Laurent, PhD

Detect. Diagnose. Design. Adopt.

I identify opportunities and turn complex problems into working systems that can run without me.

I find what is not working and what could work better, figure out why, design the change, and hand it off with an owner attached. I usually work across teams and organizations I do not manage, where adoption has to be earned with evidence: stated assumptions, tested claims, and a record others can check.

With AI, I ask where it adds real value rather than speeding up workarounds, and how to keep key decisions with people.

Review selected work
20+Years turning complex systems into repeatable work
7Industries: software and information, food manufacturing, multi-unit food service, public administration, higher education, scientific research, nonprofitSame operating model

Approach

Same four steps, whether it is a restaurant POS or an AI product.

The problem people report is usually a symptom. The cause sits upstream, in a system nobody has looked at in years or a workflow so tedious that everyone stopped questioning it. The four steps run in order because each one produces what the next one needs, and the last one produces the only thing that matters.

01

Detect

Listen, notice, and question. Find the friction and the opportunity, especially where automation or AI would change what someone decides. The output is a problem worth solving, stated plainly enough to argue with.

02

Diagnose

Describe the systems and the assumptions holding them up. Find the actual levers. Identify the decision owners and work with them directly rather than around them. The output is a shared account of why the problem exists, and a named owner for each part of it.

03

Design

Prototype the product, the workflows and the roles together. Test the assumptions that would break it. Forecast cost, benefit and the measures that will judge it. The output is something small enough to be wrong cheaply.

04

Adopt

Hand it off. Write it down, name an owner, and confirm it runs without me. The output is the only one that counts: a system still working a year later.

Skip diagnosis and you design for the wrong problem. Skip adoption and you built something nobody owns.

Every step leaves a record: the assumptions, the tradeoffs, who owns what. Scientific research and regulated manufacturing taught me what that record has to look like when an outside party is going to check it. AI governance is the same problem in a different domain.

Systems and structure

Designing what the system should be before anyone builds it.

Structure is cheap to get right before anything depends on it, and expensive afterward. That window is short and it is usually skipped.

Define the required fields, control the vocabulary, and make every record carry its evidence. Schema and data architecture, decision structures, assumption exposure, and measurement design.

2008 to 2013 Metaweb Technologies Schemas and knowledge graphs

Freebase schema design

Under paid contract with Metaweb Technologies, I designed graph database type and property models for Freebase, the open structured knowledge base. Google acquired Metaweb in 2010. Google used Freebase to seed its Knowledge Graph. The contract work modeled the motorcycle domain. On my own initiative, I tested the platform and built types and properties for bird species-habitat and trait relationships, then presented at scientific conferences on what that structure made possible for ecological data.

Why it matters: Machines reason over typed relationships, not prose. Freebase taught me to build the structure first and let the applications come to it.

Structure before applications
2023 to present GoTo Foods · Jamba Product master · Data design

Canonical product master for a multi-platform retail stack

Product identity lived in fragments. Names, sizes, PLUs and platform-specific IDs sat in separate systems, and teams reconciled them by hand every time they configured, searched or troubleshot something. The errors were not the problem. The absence of a shared product model was. I built a canonical product master that maps each Jamba product across systems on standardized required fields, then designed purpose-built views over it for the workflows that actually recurred: Punchh configuration formatting, size-based search, all-sizes lookup by product name or flavor, and cross-system ID reconciliation.

The distinction: not a static reference table. An internal data product. The master holds the canonical record. The derived views make it immediately usable in the task someone is actually doing.

Still in use: I moved off Jamba administration. The file and the workflows stayed.

Still in use
2009 to 2013 US-NABCI Measurement standards

Geographic description standard for sampling designs

Project lead on national recommendations for a geographic description standard for sampling designs, for federal, state, and NGO monitoring programs that did not share a chain of command. US-NABCI, the US committee of the trinational North American Bird Conservation Initiative, endorsed the recommendations in 2013.

Evidence it held: the standard outlived the working group that produced it and is still in use in a multistate monitoring program.

Still in use
2000 to 2005 Michigan State University Research software · Modeling

HABIClass, wildlife distribution modeling

Standard models predicted where a species lived by first passing through a hand-drawn land cover map. Whatever was wrong in the map became wrong in the prediction. HABIClass skipped the map and compared raw satellite readings directly against known occurrences. Doctoral work across a 400,000 hectare forested region of Michigan's Upper Peninsula, published in Remote Sensing of Environment in 2005.

The AI parallel: AI systems make the same mistake when they answer from a tidy summary of the source material instead of the source material itself. Grounding quality depends on what the system is actually allowed to inspect.

Published method
2005 to 2008 North Carolina State University Database and schema design

LitCentral

A structured database of bird habitat associations, built for the National Gap Analysis Program. It converted published prose, scientific and popular, into structured records that could be mapped. Data entry forms held the structure. Controlled vocabularies held the language. Every record carried its source citation.

The constraint: habitat gets described in whatever words the author chose. Unstandardized, the literature will not aggregate. Uncited, the map will not survive review. The schema turned unstructured scientific literature into usable structured data that solved both at once.

Delivered
1 / 5

Enablement

The same method, applied to AI.

Getting a team to use it, not to pilot it. Curriculum, prompt standards, use case selection, decision hygiene, and business case.

2023 to present GoTo Foods AI enablement

AI curriculum for a 20-person retail IT department

The Senior Director of Product asked me to establish an AI knowledge baseline for the 20-person retail IT department. I authored the internal curriculum covering prompt standards, verification, responsible use and workflow application. I work with subject matter experts to surface automation opportunities, then build in Claude and Claude Code or route the request to an automation specialist. I refine the requirements once people use what was built, and carry the cost and benefit case to leadership.

In production
2025 to present WalelAI LLC AI product · Independent venture

WorkOutput

A generative AI decision-support product that helps professionals capture, evaluate and improve consequential decisions over time. I own it end to end: problem definition, product strategy, roadmap and requirements, the initial GPT implementation, and its evolution into a stateful multi-agent SaaS product with persistent decision history, governance, analytics and outcome tracking.

Problem it solves: You can get an answer out of AI in seconds. Getting to a decision can take much longer, and most of that work never gets saved. You remember what you chose. You may not remember why you chose it, what you assumed, or what you ruled out. WorkOutput structures the decision and keeps the record, so the next one can build on the last.

Private beta
2025 to present WalelAI LLC AI education · Independent venture

AI with Love, judgment-first AI education

An applied AI education program for women, nontechnical professionals carrying real responsibility. Across a four-month experience, twelve 90-minute live labs cover prompting, workflows, verification, privacy and boundaries, and human decision ownership. A structured curriculum with assessment and a standardized before-and-after comparison, ending in a capstone project. The program is fully designed; public launch is expected in 2027.

Co-created with Jessica Valor.

System contribution: curriculum architecture, applied exercises, responsible-use boundaries and translation of technical concepts into plain language for nontechnical learners.

Launching 2027
1 / 3

Delivery and adoption

Getting systems across the line when ownership, dependencies, and authority are distributed.

Rollout design, dependency management, stakeholder alignment, and durable ownership after handoff.

2017 to 2023 Arden's Garden Compliance systems

Rebuilt the food safety system behind two production facilities

Compliance documentation had gone stale. I rebuilt it to SQF standard and was primary audit contact for two consecutive audits conducted simultaneously by the USDA and the Georgia Department of Agriculture. Both passed.

What made it pass: I designed the documentation so the records an auditor needs are produced by the way the plant already runs.

Why it mattered: that documentation and HACCP certified processes supported the wholesale supplier approvals behind Kroger and Publix distribution.

After I left: both facilities were SQF certified on the system I designed and built.

Delivered
2023 to present GoTo Foods Enterprise systems

Retail technology administration in a seven-brand franchise system

Own point of sale and online ordering configuration for Schlotzsky's on Qu Beyond, Aloha, and Olo, and support McAlister's on Aloha. Previously primary administrator for Jamba on Toast, Olo, and Qu Beyond across 700+ franchise locations. Punchh and Decision Logic are coordinated downstream. Seven national brands and a franchise system of 7,300+ locations, with no central database behind them: only configurations and integrations, where one error propagates outward. Menu architecture, pricing and data integrity, and third-party integrations, coordinated with marketing, brand, app and loyalty teams. Team lead for Schlotzsky's.

Operational evidence: 50+ store builds, 100+ menu builds, 1,000+ escalations resolved, and 20+ knowledge base articles.

Current role
2012 to 2017 Connecting Conservation Enterprise platform

Griffin Groups

A B2B collaboration and knowledge platform delivered from concept to production for 2,300+ users across federal agencies, NGOs and universities. Requirements, roadmap and releases tracked across 800+ Jira tickets, 30+ sprints and 30+ plugins, with analytics driving prioritization and stakeholder alignment holding a cross-organizational user base together without shared authority.

How priority was set: I monitored the platform analytics daily, analyzed and graphed them, and used those results to judge which tools and group structures were actually working.

2,300 users
2017 to 2023 Arden's Garden Point of sale rollout

Point of sale conversion: two pilot stores, then 14 in one day

Two pilot stores ran Shopify first, then Toast, each scoped to surface implementation failures rather than to prove the system worked. Live use showed Shopify grouped identical products, so individual items could not be discounted. I took that finding to the COO, and the company moved to Toast. The remaining 14 stores converted in one day with no downtime, followed by Microsoft Dynamics ERP configuration across thousands of SKUs.

Why the pilot was built that way: I scoped the pilots to find failures, not to prove success. What the two stores exposed made a single-day cutover safe for the other fourteen.

Delivered
1 / 4

About

One career. Different operating environments.

I began as a conservation scientist, where the work required finding patterns in complex systems, testing explanations and making decisions from incomplete evidence. That record includes a PhD, 25+ scholarly and technical publications, reports, and handbooks and a NASA fellowship.

I carried the same operating model into enterprise software and information systems, food manufacturing, product and platform delivery, retail technology and applied AI, and into my independent venture, WalelAI LLC. The environments changed. The core work did not.

I detect what is happening, diagnose why, design the system that should exist and drive adoption so the result survives the person who built it.

Systems adoption · Enterprise AI · Technical programs · Product and platform delivery · Operational design · Evidence and governance

25+ publications, reports, and handbooks · 788 citations · h-index 10 (Google Scholar, 6 September 2026)

The papers that still shape how I work.

My complete record of papers, reports, and standards sits on Google Scholar and ResearchGate. The work below was selected for what it still contributes, not for how often it has been cited.

2011 · Springer

The role of assumptions in predictions of habitat availability and quality

Laurent, E.J., C.A. Drew, W.E. Thogmartin. In Predictive Species and Habitat Modeling in Landscape Ecology, pp. 71 to 90.

A framework for stating what a model assumes before anyone acts on its output. The same discipline every recommendation on this page runs on.

Publisher · Paywalled
2013 · US-NABCI

Standard Sampling Grids for Avian Monitoring Programs

US-NABCI Monitoring Subcommittee. Project lead: E.J. Laurent. Recommendations to US-NABCI.

Recommendations endorsed by US-NABCI, the US committee of the trinational North American Bird Conservation Initiative. Still in use in a multistate monitoring program.

Free PDF

Contact

ed@edlaurent.comCopied LinkedIn