How I work

Implementation

Implementation

50% faster value realization

The problem

The implementation process worked. Customers got live, teams hit their milestones. But customers were waiting 30 days to realize value from a platform they'd already bought because the process was built around our internal needs, not their time-to-value.

What we changed

We audited every step and asked a simple question: "Does this belong to the customer or to us?" The complex, ambiguous, expertise-heavy decisions. We took those back. Built the internal capability to absorb them. Stopped leading with questions and started leading with recommendations. Customers brought the context. We brought the expertise, knowing what questions to ask, what was missing, and when to push for more.

The outcome

Time to first value dropped from 30 days to 15. Not only because we moved faster, but because we put our effort in the right places and stopped asking customers to do work that was never really theirs to do. That shift became the model on which everything else was built. I talked through this redesign at Propel26.

AI Systems

Reconciliation before kickoff

Every new customer arrives described by four different sources: the request that starts implementation, two CRM records, and the sales calls that got them here. Each one is accurate. None of them are fully aligned.

Nobody read all four side by side, because that takes an hour nobody has before a Thursday kickoff. So the disagreements surfaced live, in front of the customer. Sample size was recorded one way in the contract and promised another way on a call. A launch date that had moved four times with no record of why.

I built a tool that reconciles the sources instead of summarising them. The CRM already holds the facts. The value is in what doesn't add up. It marks every claim as stated, implied, or unknown, and the unknowns are what the team asks about before kickoff rather than during it.

The first version was worse. It tried to serve two readers at once and the output was inconsistent, so I rebuilt it. I also cut my own handover condition and moved the team onto it earlier than planned.

The outcome

It runs on every new account now. Two of the patterns it caught have recurred, so they're permanent checks.

AI Systems

My digital bench

Most people use AI as a very capable stranger. Every session starts from nothing; you re-explain the same context, and you get advice that ignores decisions you already made. The output is good, and the accumulation is zero.

I built an alternative. A set of specialized agent colleagues instead of one general assistant, sharing a context I own.

There's an orchestrator that handles routing, prioritization, and synthesis. A strategic advisor for pressure-testing decisions and operating model design, which is the one that tells me when my reasoning is thin. A knowledge architect that owns documentation structure and decides where things belong. An analyst for implementation data and reporting. Each has a defined scope, and the routing between them is explicit rather than me picking whichever chat window is open.

They read from a shared context layer that holds what's true right now: current priorities, the decisions I've made and the reasoning behind them, how the team operates, what each part of the system is responsible for.

Three design rules do most of the work.

Nothing writes without approval. Updates get drafted, shown to me, and written only after I say yes. A system that remembers what you tell it is useful. A system that quietly rewrites its own understanding of your work is a liability, and you find out which one you have at the worst possible time.

Contexts stay isolated by default. My career development work runs in a separate system that delivery work cannot read. Information can move one direction, deliberately, and not the other. Without that boundary the two contaminate each other and neither stays honest.

Every material decision gets logged with its reasoning. Not what I chose, why I chose it and what it commits me to. It stops me relitigating settled questions, and it catches the times I'm about to contradict something I decided three weeks ago for reasons I no longer remember.

The outcome

This is how I work, not a side project. The parts I'd expect to matter most were the capabilities. What actually mattered was governance: deciding what the system is allowed to change on its own, which turned out to be almost nothing.

AI Systems

Phase-mapped AI integration: implementation operating model

8 of 11 stages live in production

The problem

Implementation runs on repeatable phases, but the highest-friction moments (kickoff prep, status updates, audience structuring, handover documentation) were still being handled manually, every time, at every account.

What I designed

Mapped AI assistance to the specific friction points across an 11-stage implementation methodology. Built and deployed seven tools, each targeting a distinct phase, and established a pipeline structure for future additions as new friction points are identified.

The outcome

Eight of eleven implementation stages now have active tooling in production. Recurring admin at each touchpoint is handled by the system. New capabilities are added against a defined methodology rather than as one-off fixes.

AI Systems

AI agent setup automation

The problem

Launching AI-powered Marketing Agents required manually reviewing and categorizing large volumes of client files, a 20-hour process that created a bottleneck before any customer value was delivered.

What I built

Designed and built an AI workflow that ingests client files at scale and creates agent-friendly content summaries, including making visual assets usable as structured inputs for faster, more efficient processing.

The outcome

Setup time dropped from ~20 hours to 8–10. That compression reduced internal delivery cost and cut customer lead time — accelerating time-to-value on a flagship AI product for every new deployment.

Operational Design

Pricing & scope defense through measurement

0→1 visibility · pricing leverage gained

The problem

We had no idea where the time was actually going. Which meant every pricing conversation was a guess, scope had no floor, and the cost accumulation had nowhere to land.

What I built

Conducted an internal audit to define what needed to be measured and why. Designed a tracking framework from scratch, calibrated for accuracy, and built the measurement system that captured effort at enough granularity to surface real trends.

The outcome

Identified hidden effort sinks growing month-over-month. Provided the data foundation that defended scope in client conversations and justified price increases that previously lacked any supporting evidence.