Every company runs on more than one version of itself.Execution drag lives in the gaps.
Ask four functions who committed the go-live date
and you will get four confident answers.
- SalesDelivery“Delivery owns it.”
- DeliverySales“Sales committed it.”
- The SOWEngineeringDates need Engineering acceptance.
- The CRMNot recordedFour dates and no acceptance record.acceptance
Nobody lied. The accounts conflict, so ownership has to be validated.
Conflicting answers do not prove the date has no owner. They show that nobody can say for sure yet. The method keeps each account separate, compares it with the trail, and allows for an owner who is unclear, disputed, or legitimately shared, without forcing one confident answer.
The five evidence states
Three are evidence we collect. One is inferred from that evidence. One is the future state.
Role descriptions, RACI charts, SOPs, onboarding guides, the decision rights nobody has opened since they were written. Collected from the company's own documents. It is the company's intent, frozen at the moment someone wrote it down.
The founder's account, each leader's account, each function's account of who decides what. Collected through structured intake. Honest and incomplete, because everyone describes the company from their own seat. When two people confirm the same account it is stronger, and it is still reported evidence.
Message threads, ticket history, CRM records, meeting notes, the approvals that went through and the ones that stalled. Collected from the systems where decisions leave a trail. Usually the least flattering of the three.
Not collected directly. It is inferred by reading the first three against each other: the workarounds, who people really go to when something is stuck, which meetings decide and which only report.
How each decision should run: who is accountable, who approves on which paths, where it escalates. The Sprint points toward it as a direction. The Install designs it in detail and takes it through sign-off.
Compared, never averaged. Never weighted by seniority.
The founder's accountdoes not outrank the CRM.
The SOPdoes not outrank the Slack thread.
Each account is collected on its own terms and laid next to the others.
- agreeagreedisagrees
When reported and documented agree and observed disagrees with both, the company has a process it believes in and does not run.
- agreesays something elseagree
When reported and observed agree and documented says something else, the company has outgrown its own manual.
- three owners
When three functions name three different owners for the same decision, the accounts conflict, and ownership has to be validated before anything is redesigned.
Each of those is a specific structural gap with a specific fix. None of them is visible from inside any single account, which is why leadership offsites keep discovering the same problems and keep not fixing them.
From recurring decisions to future state
Identify
The work starts with decisions, not functions.A recurring decision is a repeatable organizational choice that arises on a regular schedule or when defined conditions occur. It may sit within one function or cross several. We prioritize decisions where unclear authority, dependencies, exceptions, or delayed resolution create meaningful execution friction. Each decision is validated against the evidence before examining its ownership, inputs, approval paths, handoffs, escalation rules, and measurement.
Nothing is redesigned just because it exists. The work follows the decisions that need it.
Collect
Documented and reported evidence comes from the company's own documents and structured intake.Observed evidence comes from read access to the systems where decisions leave a trail. De-facto patterns are inferred by reading the three against each other.
Nothing said, decided, or documented is allowed to get lost between conversations. Every finding is traceable to where it came from.
Measure
Defensible baselines where the evidence allows, and an agreed measurement plan where it does not.Founder escalations, decision cycle time, and rework are defined with you, and each number is labeled: observed, validated with you, pending, or insufficient evidence.
A later number only counts as improvement when it measures the same decisions the same way. Where no defensible starting point exists, we do not claim one.
Diagnose
We read the gaps.Diagnosis is a judgment call built on evidence, made by an operator who has run programs across BI, engineering, product, and regional operations, and it is the part of the method that does not automate.
Every finding traces back to the evidence it came from. Point at one.
- Which recurring decision has conflicting accounts of who owns it.Traced from documented, reported, observed.
- Where a seam sits: a gap or conflict in authority that work keeps falling into.Traced from reported, observed, de-facto.
- Which handoff has no definition of done.Traced from documented, observed.
- Which escalation has no threshold, so everything escalates or nothing does.Traced from observed, de-facto.
Design
The designed state:who is accountable for each decision, who approves on which paths, and where an unresolved decision escalates. The Sprint points toward it. The Install designs it in detail and takes it through sign-off.
A handoff moves work between people whose authority is clear. A seam is where that authority is unclear. Correct escalation is often the system working as designed.
When the trail has no single owner
- An AI agentrecommends.
- An automationtriggers.
- One systemrecords the result.
- A humanacts on it somewhere else.
In that chain, no single person holds the full trail of how a decision got made, and the reported version of events gets less reliable every quarter.
More decisions, more exceptions, more handoffs.Companies that added AI without redesigning the operating layer around it, which Deloitte's 2026 data puts at 48 percent of respondents, now generate more decisions, more exceptions, and more handoffs than their informal system can carry.
The method is built for exactly that condition.It reads what the systems recorded, not what the team remembers. In the designed state, who holds a decision is set separately from what AI contributes to it. AI assistance does not confer authority, and not every decision should be automated.
Three places you will meet it
Questions about the method
Why not just interview the leadership team?
Because interviews capture reported evidence only, and reported evidence is where every company looks healthiest. The go-live date example above would have produced four confident, contradictory interviews and no diagnosis. The gaps only appear when the trail is read alongside the accounts.
How much of my team's time does the evidence review take?
Less than an interview-led approach. Most of the evidence already exists in your systems and documents. The team's time goes into a short intake, one review, and the handoff.
Is this an audit?
It is a diagnostic. An audit checks compliance against a standard. This method checks the company's accounts of itself against each other and against the trail.
Do you use software to do this?
The evidence review is tool-assisted where that makes it faster and more reliable, and the diagnosis is made by a person. What gets published is the finding and where it came from, never a score from a black box.
What does the output look like?
Three parts: the validated current state, a T0 baseline with the confidence of each number stated, and a directional future state with a possible 90-day sequence. The full list is on the Sprint page.