I build the systems that win, keep and grow customers, and the data infrastructure underneath them.

AI agents, CRM and lifecycle automation, and analytics platforms built from zero. Eight years, from 100-million-row data stacks to production revenue agents. I learn the business before I build, so every system attacks a number the owner actually cares about.

Work with me See the work

120+AI systems shipped at 40 Analytics

100M+rows: raw files to daily automated reporting, built solo

Months → dailyreporting cycle after rebuild; ad-hoc in 48h

2×BI platforms built from zero as sole admin

Selected work

Systems that are still running.

Client results from my work at 40 Analytics and in-house data leadership roles. Numbers are real and I can walk through any build end to end.

5×

more qualified conversations for the sales team

AI AgentsRevenue Automation

Revenue AI Agents: lead generation to booked meeting, no manual touch

What it does

Trigger-based agent pipeline that finds leads, enriches and scores them against criteria the sales team agreed in writing, drafts personalised outreach for one-click human approval, and writes everything back to the CRM. Runs continuously off behavioural signals.

Stack: n8n orchestration · Claude reasoning steps · CRM writeback · enrichment APIs

System

+38%

lead-to-won conversion

CRMSales Automation

Self-updating CRM with automated lead routing

What it does

Rebuilt a client's CRM layer so records, stages and ownership update themselves from live activity, with routing rules deciding who works each lead and when. The pipeline report regenerates itself; nobody retypes anything on a Monday.

Stack: CRM automation · n8n / Make workflows · reporting layer

System

30s

to auto-assign each order

Custom SoftwareOperations

Logistics dispatch platform

What it does

Custom-built dispatch software for a logistics operation: orders come in, the system assigns them to the right driver against live capacity, no human dispatcher in the loop.

Stack: custom web platform · automated assignment logic · ops dashboard

System

100M+

rows rebuilt into a running data function: months to daily reporting, 48h ad-hoc turnaround

Data InfrastructureCRM & Lifecycle

US manufacturer: 100M+ rows to a running data and CRM function

The problem

Joined a Los Angeles company (remotely, from Accra) with no data infrastructure and 100+ million rows of history trapped in raw files.

What I built

Built the entire stack: ingestion pipelines, analytics database, transformation layer, Power BI on top. Then took ownership of the CRM and lifecycle stack: behavioural segmentation, churn-risk scoring, onboarding / win-back / upsell flows in HubSpot, Klaviyo and Mailchimp, A/B tested with control groups, measured on cohort retention and LTV.

Stack: pipelines · analytics DB · Power BI · HubSpot · Klaviyo · n8n / Make / Zapier sync

System

From zero

to company-wide self-serve analytics, twice

Business Intelligence

Two BI platforms from absolute zero

The problem

At two companies with no BI whatsoever, every question had to go through an analyst.

What I built

Stood up Looker end to end as sole administrator: LookML semantic models, explores, governance standards, and the CloudSQL-to-BigQuery pipelines underneath. Every department went from asking an analyst to self-serving answers.

Stack: Looker & LookML · BigQuery · CloudSQL · Power BI · data governance

System

The through line

Three jobs. One system.

Every build above is one of three jobs. Most businesses buy three disconnected tools for them and wire them together with people. I build them as one system, so each one feeds the next.

Win

Agents that find the right buyers, qualify them against criteria your sales team actually agreed, and open the conversation, writing every touch back to the CRM.

Keep

Lifecycle and churn-risk scoring that flags the customer about to leave while there is still time to act, then runs the win-back itself instead of filing a report.

Grow

The data layer underneath both: pipelines, models and reporting that make the next decision an observation instead of an argument.

How I work

The pattern behind every build.

Three steps, in this order, every time.

01Learn the business first

Before any workflow gets built, I find the number the owner actually cares about: revenue per cohort, hours burned, cycle time. If automating it isn't worth more than building it costs, I say so.

02Build the shortest path to running

n8n and Make when orchestration is the job, Claude Code and custom software when logic gets genuinely custom. The skill is knowing which problem is which. Working system in weeks, not quarters.

03Operate and prove it

Systems I ship assume failure: retries, fallbacks, human-in-the-loop checkpoints. Then the reporting layer proves the result, because I build that too.

Figuring it out is the actual skill.

A domain I have never worked in, a stack I have never touched, a problem nobody in the building has solved yet: that is the work I want. Tools change every year. Walking in, finding the real problem and building the thing that fixes it does not. I have never needed the map to already exist.

Beyond the build

Teaching and speaking.

The systems only matter if the people around them understand what they are looking at. So I teach, and I talk about this work in public.

2,000+

people taught in data and AI

I teach data and AI to people who have to use it at work, not to an audience of engineers: what the tools genuinely do, where they fail, and how to put them to work inside a real business.

Speaking

I speak at events on data and AI, on what actually survives contact with a business: where agents earn their keep, why most automation projects stall, and what a data function needs before any of it works.

Recent: MTN CTIO Roundtable, AI and the Future of Business · April 2026

Writing

Notes on data and AI.

Things I have written up, usually after a build taught me something I did not expect. All of it lives on LinkedIn.

AI at work

AI in the Workplace: Imposter Syndrome and the Power of Grace

July 2026

AI in Africa

AI in Africa: A Missed Opportunity for Real Innovation

April 2026

AI economics

Anthropic Just Proved the Price of AI Tools Is Not the Real Price

April 2026

AI economics

Sam Altman Just Confirmed AI Will Break the Economy

April 2026

Data teams

Effective Data Analysts: Curiosity, Obsession and Context

March 2026

Experimentation

Companies Abandon Tests Because of a Patience Problem, Not a Data Problem

March 2026

Skills

Upskilling vs Building Capacity in the Age of AI

March 2026

AI economics

AI's Dark Side: Concentrating Wealth and Economic Collapse

March 2026

AI economics

The Price You Are Paying for AI Tools Is Not the Real Price

March 2026

Data quality

Data Accuracy Matters: Every Source Is Guilty Until Proven Innocent

March 2026

Stack

Tools change. The pattern doesn't.

n8nMakeClaude Code OpenAI CodexZapierReplit LLM APIsGoogle BigQueryLooker & LookML Power BISQLPython GCPCloudSQL HubSpotKlaviyoMailchimp

Signal → decision → action → measurement.

Everything on the left is just how that gets expressed this year.

Let's build something that runs

Currently leading AI and data at 40 Analytics, and open to conversations about roles and projects with North American teams. I am open to relocation, I already work North American hours, and my preferred interview is a take-home build, because the work is the argument.

adamsmawulisylvester@gmail.com LinkedIn