Project Everstate

How it started

The client was about to change how the company works. Before you change how a company works, you must know how it actually works – and nobody could say, with evidence, how it actually worked.

The premise we walked in with: in the age of AI every company is starting over, whatever its age – nobody has done this before, so an established business is as much a beginner as one founded last week. Which makes "you can’t know everything before you start" a self-imposed limitation, not a law of nature. Everything an organization knows about its own operations can be written down, structured, accessible and up to date. Projects then become predictable in scope, budget and timeline, and nobody discovers in month four what could have been known in week one.

What we found

The usual pattern – decisions made on partial information obtained from a handful of interviews and workshops, and assumptions.

• The wiki stored what someone had bothered to write down, and had been decaying from day one.
• Previous consulting decks were a snapshot of a sample, filtered through office politics and outdated before they were presented.
• Requirements written against a partial picture, which is how scope creep is guaranteed.
• AI tools handed out per person. Individually faster people don’t make a structurally faster company; the bottlenecks are in how work flows, not in how productive each person is.

Workshops were never the ideal method – they were a sampling technique, because you couldn’t physically interview 200 people. That constraint no longer exists. AI adoption is not AI transformation: adoption gives people tools to be more productive, transformation redesigns the way work itself happens. Most transformation projects and AI initiatives fail because nobody establishes how the company actually works before deciding what to change.

What we did

• Interviews – AI agents ran 30–45 minute interviews in parallel across the entire organization. Adaptive conversations, catered to each employee, running deeper than human-led discovery: workflows, handoffs, dependencies, workarounds, informal habits and manual tasks that never made it into documentation.
• Structure – the results were processed into a structured map of the operation: workflows between people and systems, dependencies, bottlenecks. First a human-friendly version, no longer trapped in people’s heads, scattered Word files and old whiteboard photos. Reviewed by humans, then transformed into the overarching knowledge layer, accessible by machine and human alike.
• Knowledge layer – established and continuously updated to match business reality, not updated when someone remembers a Confluence page exists.

How it went

Coverage over sampling. Transparency over assumptions. Structured for AI and human. Always up to date.

Where a traditional engagement team spends weeks assembling something that aims to be representative, we delivered full coverage and a structured overview of the operation in a fraction of that time – organization-wide process understanding in the time it takes to schedule a discovery workshop. The guesswork was removed and replaced with an evidence-based picture of how the company operates.

What we delivered

• Operating map – a complete, navigable overview of how the company operationally works, with every workflow, dependency and bottleneck.
• Prioritized initiatives – where AI creates value, ranked by impact and feasibility. Specific to their operations, not a list of generic use cases.
• Implementation blueprints – for each initiative: what changes, who is affected, what tools are involved, what the expected outcome is.
• Ongoing access – the agents that built the map remained available afterwards, and can be asked any question.
• Continuous updates – unlike a slide deck, the assessment doesn’t expire. The operating map is kept current to reflect the organization’s situation.

Project requirements, impact analyses, onboarding and other documents now reference it.

The result

When 1890s factories replaced steam engines with electric motors, productivity barely moved. The gains only came generations later, when factories were redesigned around what electricity made possible. Handing employees an AI tool is an engine swap – the way of working still revolves around outdated, human-imposed constraints. Nobody is further along here than anyone else; every company is at the beginning of this one, however long it has been trading.

This project removed the constraint. Their AI of choice went from an intern to a person who spent the last twenty years working for the company, and project decisions now reference reality rather than recollection. The map covers the whole organization and stays alive after the consultants would have left – because keeping it current is the business.