Most organisations do not have a data problem, an analytics problem or an AI problem. They have a sequencing problem wearing three different costumes, and no agreed way to decide what comes first.

The 7D method. Seven gates from a company or business unit goal through to booked value. Dividend feeds the next Direction, which is the arrow that turns a piece of work into a capability.

The 7D method is how a company or business unit goal becomes a set of things people can act on. It works across the whole chain, because the chain is where the value leaks: an estate nobody trusts, reporting nobody opens, and AI sequenced onto both.

The seven gates

Direction. Which company or business unit goal does this serve, and by how much? Fails when a programme is named after a technology rather than a goal.
Decision. Which decision changes, how often is it made, and what does a bad one cost? Fails when a capability arrives first and nobody ever counts the decision, so the business case rests on a guess.
Data. Does trustworthy ground truth exist, and does a named human own it? Fails in month four when three systems disagree.
Design. Buy, configure or build, with the third option honestly priced? Fails as a bespoke build of something a licence already does.
Deployment. Does it land in the workflow people already use, and are they still using it in month six? Fails twice over: either it never reached the workflow, or it did and people drifted back to the spreadsheet once the attention moved on.
Discipline. Who governs it, and can you evidence that? Fails in week ten when legal blocks go live.
Dividend. Is the value booked against a named budget holder with a baseline? Fails as savings no finance director can locate.

Nothing gets funded until it clears gates one, two and three. That rule costs nothing to adopt and is usually the difference between a productive year and an expensive one.

What it covers

This is not an AI readiness review. Gates three and four look hard at the data estate and the analytics layer, because those are where most of the value is lost long before anyone reaches a model.

Data. Estate, architecture, quality baseline, ownership, migration sequence, what to decommission.
Analytics and BI. What reporting exists, who opens it, which decisions it supports, and the governed metric layer underneath.
AI and agentic. Portfolio, build or buy, where agentic workflows genuinely fit and where they are a fashion.
Governance. Policy, risk, classification, assurance and audit trail.
Adoption. Whether the organisation can absorb any of it, which decides everything else.

Why it is a loop, not a line

Dividend feeds the next Direction. That single arrow is the whole argument, and it is the reason this is a method rather than a checklist.

Most data and AI work is funded as a project. A project has a start, an end, a budget and a closure document. It is a sensible way to build a bridge and a poor way to build a capability, because the thing you are building does not stop needing decisions the day the funding stops.

The organisations that get value out of this treat it as a product. Funded continuously rather than in one lump. Owned by a named person who is still there in year two. Measured on whether anybody uses it, not on whether it shipped. That shift is usually worth more than any individual thing in the portfolio, and almost nobody makes it deliberately.

	Funded as a project	Run as a product
Funding	One approval, one budget, one year	A standing allocation, reviewed against value delivered
The team	Assembled, then dispersed at closure	Stable, and it keeps the context
It ends when	The scope is delivered	The decision it serves stops being made
Success is	On time and on budget	Still in use in month twelve, with the value booked
Ownership after	Handed to whoever will take it	Named before the first release, and unchanged
The knowledge	Leaves with the contractors	Stays, and compounds into the next thing
What you get	An asset	A capability

You do not have to reorganise the company to start. Gate seven is where it begins: a named owner, a measured baseline and a review at ninety days. Do that on one thing and you have run your first product cycle, whatever your funding model officially says.

A project ends. A decision does not. Funding the first and expecting the second is the most expensive mistake in this whole chain.

Why not use an existing framework

A fair question, and I looked before building this. The established ones are good at what they measure. The problem is what sits between them.

What exists	What it measures well	Where it stops
Data management maturity models	Discipline across data governance, quality and architecture	The data layer. They say nothing about whether anything built on top of it delivers value
Capability maturity models	Process repeatability and organisational maturity	Capability, not outcome. An organisation can score well and still deliver nothing
Cloud adoption frameworks	Migration readiness and platform design	The platform. They are vendor shaped by design, and the vendor does not sell the hard part
AI maturity models	AI ambition, skills and capability	They assume the data and analytics beneath are sound, which is exactly where most value is lost
Regulatory baselines	Compliance exposure and classification	Compliance is a floor. Passing means you are allowed to proceed, not that proceeding is worth it

Each of those covers one link. None of them scores the whole chain, from the company goal at one end to booked value at the other, in a single number you can put in front of a finance director.

Every framework I could find measures how good you are at a discipline. None of them measures how much value survives the journey from objective to accounts.

Two things this measures that I have not found elsewhere

Compounding loss across the chain. Maturity models score each discipline separately, which hides the multiplication. Two thirds trusted data, times two thirds useful analytics, times two thirds adopted output, is not a third lost. It is seventy per cent. Scoring the links independently makes an organisation look considerably healthier than it is.

Agentic authority. Almost every framework in common use was written for systems that advise. An agent that takes action needs three things a recommendation engine does not: a defined authority boundary, a rollback path, and a named human accountable for what it does unsupervised. Gates five and six ask for all three, because the frameworks written before 2025 largely do not.

I am not claiming nothing like this exists anywhere. I am saying I could not find one, and I needed it for my own work, so I built it and now use it with clients.

Over three weeks

Week one. Stakeholder interviews across leadership, finance, operations and technology. Typically eight to twelve conversations.
Week two. Estate and data assessment, BI inventory with usage data, first pass at the portfolio.
Week three. Scoring, sequencing, value cases for the top candidates, and the leadership session.

What you are left with

A decision register with volumes, cycle times and cost to serve.
A data modernisation roadmap, sequenced and costed, with an owner against each step.
An analytics and BI blueprint: one version of the numbers, and who owns them.
A prioritised portfolio, with the things to stop marked as clearly as the things to start.
A governance framework: policy, risk register, assurance, audit trail.
A benefits realisation plan with baselines, owners and review dates.

Every one of them owned by a named person before I leave. That is the point of the exercise, and it is why the last gate is Dividend rather than Delivery.

You can run a short version of this yourself right now. The free gate check asks seven questions and tells you which gate you are stuck at, with no form in the way.

The 7D Method