Decision SystemDecision 6 of 6

2,000-person company

Do not automate organisational unreadiness

An €8 million AI agreement, a competitor’s announcement, and no agreed account of how the company works.

An illustrative scenario. It does not describe a client or a real company.

Frame

Objective, horizon, non-negotiables and who holds authority.

A 2,000-person company is considering an €8 million, three-year agreement with an enterprise AI provider.

The provider proposes company-wide assistants, connected knowledge and automation, projecting €15 million in savings.

However

Data is fragmented and inconsistently governed.
Processes are poorly documented.
No reliable savings baseline exists.
Vendor dependence would be substantial.
Leaders have not defined AI authority.
Departmental pilots look impressive but adoption is inconsistent.
A competitor has announced a major AI transformation.

Should the company buy a company-wide intelligence layer before it has agreed how the company itself works and decides?

Objective: Create measurable value from AI in this company, without committing the organisation to a platform it is not ready to use.
Horizon: Two quarters for a bounded pilot. Three years for the agreement on the table.
Non-negotiables: Value is measured against a baseline the company has established itself.
Decision authority is granted explicitly. It is never acquired through convenience or platform design.
The company can leave with its data.
Who decides: The chief executive, with the board. AI may support defined workflows within explicit permissions. What it may decide is set by leaders, not by the platform.

Observe

What is established, what is assumed and what is missing.

Every item is typed. Facts come from the scenario. Assumptions are illustrative quantities added so the consequences can be calculated. Unknowns are left unknown.

Established 8

Fact:
The agreement is €8 million over three years, covering company-wide assistants, connected knowledge and automation.
Given in the scenario. High confidence.
Fact:
Data is fragmented and inconsistently governed.
Given in the scenario. High confidence.
Fact:
Processes are poorly documented.
Given in the scenario. High confidence.
Fact:
No reliable savings baseline exists.
Given in the scenario. High confidence.
Fact:
Vendor dependence would be substantial.
Given in the scenario. High confidence.
Fact:
Leaders have not defined AI authority.
Given in the scenario. High confidence.
Fact:
Departmental pilots look impressive but adoption is inconsistent.
Given in the scenario. High confidence.
Fact:
A competitor has announced a major AI transformation.
Given in the scenario. High confidence.

Assumed 5

Assumption:
The provider projects €15 million in savings.
Given in the scenario. Low confidence.
Assumption:
Across the departmental pilots, about 35% of intended users work with the tools each week. Pilots elsewhere range from 15% to 75%.
Illustrative modelling assumption. Low confidence.
Assumption:
About 40% of the data the assistants would need is governed. About 30% of the processes they would touch are documented.
Illustrative modelling assumption. Low confidence.
Assumption:
A bounded implementation around 2 workflows would cost about €1.2m.
Illustrative modelling assumption. Medium confidence.
Assumption:
Once workflows depend on the platform, leaving would cost about 60% of the contract value.
Illustrative modelling assumption. Low confidence.

Not yet known 2

Unknown:
What the candidate workflows cost today: baseline cost, time and quality.
Given in the scenario. Low confidence.
Unknown:
Whether measurable value appears in real work here, under real adoption.
Given in the scenario. Low confidence.

Model

The relationships that can be calculated, and their limits.

Only relationships that can honestly be calculated are modelled. Apply one bounded pressure, or move one assumption, and follow the consequence across the organisation.

Apply one pressure

As briefedThe scenario as it was first described. The board wants an answerAfter the competitor’s announcement, the board asks for a company-wide AI commitment before its next meeting.The provider discountsThe provider cuts the price to €6 million if the agreement is signed this quarter.Adoption doublesA push from department heads takes weekly adoption in the pilots to 70%.

Or change one assumption: Share of intended users who adopt

35%

Bounded to 15% to 75%.

What changed

First effects

Second effects

Third effects

Fourth effects

Thresholds

Technology and data

Net benefit on the provider’s numbers€7.0m

Finance

Adoption multiplied by readiness needed to break even53.3%

Pilot as a share of the contract15%

Realisable savings€1.6m

Realisable savings less the contract−€6.4m

Savings cannot be measuredCrossed

Present readiness does not pay for the contractCrossed

Operations

Organisational readiness30%

Distance from break-even5.1x

The organisation is below working readinessCrossed

People

Adoption multiplied by readiness10.5%

Governance

AI authority is undefinedCrossed

The board wants a response to the competitorNot crossed

The conditions for expansion are demonstratedNot crossed

Risk

Cost of leaving the platform€4.8m

No exit is securedCrossed

Apply a pressure to see what moves, in what order, and where it stops.

Contest

Functional positions, conflict and a critic.

Each function answers in the same structure: a recommendation, the evidence it rests on, the assumption underneath, its confidence, the consequence for another function and the authority it would need.

Technology and data

Argues for: sign the company-wide agreement

Sign the company-wide agreement. The provider projects €15.0m of savings against €8.0m over 3 years, and the pilots already impress.

Rests on: Assumption:The provider projects €15 million in savings.
Assumption:Across the departmental pilots, about 35% of intended users work with the tools each week. Pilots elsewhere range from 15% to 75%.
Assumes: What impressed in departmental pilots will repeat across the company.
Consequence for operations: Every department connects its knowledge and processes to the platform within the first year.

Finance

Argues against: sign the company-wide agreement

Do not sign. With no baseline, €15.0m cannot be measured, let alone banked. At 35% adoption and 30% readiness the realisable figure is €1.6m against €8.0m committed.

Rests on: Fact:No reliable savings baseline exists.
Fact:The agreement is €8 million over three years, covering company-wide assistants, connected knowledge and automation.
Assumption:Across the departmental pilots, about 35% of intended users work with the tools each week. Pilots elsewhere range from 15% to 75%.
Assumes: Savings scale with adoption and readiness, and with nothing else.
Consequence for technology and data: Leaving would cost about €4.8m once workflows depend on the platform. Portability goes into any contract.

Operations

Argues for: run a bounded pilot, expand in stages

Pilot 2 valuable, measurable workflows. Only 30% of processes are documented and 40% of data is governed. Real work will show what has to change. The workflow savings are already in the operating plan.

Rests on: Fact:Processes are poorly documented.
Fact:Data is fragmented and inconsistently governed.
Assumption:A bounded implementation around 2 workflows would cost about €1.2m.
Assumes: Two workflows can be bounded tightly enough to measure.
Consequence for finance: The pilot, at €1.2m, produces the baseline finance is missing.

People

Argues against: sign the company-wide agreement

Do not deploy company-wide. Leaders have not said what an assistant may observe, recommend, draft or decide. Without that, authority moves to whoever configures the platform.

Rests on: Fact:Leaders have not defined AI authority.
Fact:Departmental pilots look impressive but adoption is inconsistent.
Assumes: Inconsistent adoption in the pilots reflects unclear permission as much as unfamiliar tools.
Consequence for governance: Authority boundaries need an owner on the leadership team before any rollout.

Conflict

The projection against the baseline

Technology and data against Finance

Technology counts €15.0m of savings. Finance has no baseline to measure them against.

Conflict

Deployment against permission

Technology and data against People

Technology wants assistants everywhere. People points out that nobody has said what they may decide.

Calculated

Realisable savings

€1.6m

The projection, scaled by what the organisation can reach.

Assumption: Assumption carrying weight

The provider projects €15 million in savings.

Low confidence. If this is wrong, the calculations that rest on it move with it.

Unknown: Evidence to obtain first

What the candidate workflows cost today: baseline cost, time and quality.

Authority gate

AI authority is undefined

What an assistant may observe, recommend, draft or decide is set by leaders before deployment.

Human decision required

The critic

A separate pass reads the positions for conflict, unsupported certainty, benefits counted twice, omitted effects and false precision.

Contradiction between functionsTechnology and data, Finance
Technology and data and Finance take opposite sides on "Sign the company-wide agreement". Both cannot be followed.
Contradiction between functionsTechnology and data, People
Technology and data and People take opposite sides on "Sign the company-wide agreement". Both cannot be followed.
Unsupported certaintyTechnology and data
Technology and data reports high confidence, but nothing it cites is established. Every item is an assumption or an unknown.
Unsupported certaintyTechnology and data
The €15.0m is the provider’s projection. With no baseline it can be neither demonstrated nor disproved. It is a sales figure until the company measures its own work.
Benefit counted twiceTechnology and data, Operations
Technology and data and Operations are both counting the same benefit. It can only be banked once.
Omitted second-order effectRisk
Nobody has priced the exit. Once workflows depend on the platform, leaving costs about €4.8m.
Highest-value missing evidenceFinance, Operations
The most valuable unknown is whether measurable value appears in real work here. Finding out needs a baseline first, and a pilot is the cheapest way to get both.

Where that leaves the room

Agreed

Every function wants AI in the company’s work. The disagreement is about sequence and commitment.

Still in dispute

Technology and data and Finance. Whether €15.0m of projected savings is a business case or a claim.
Technology and data and People. Whether deployment can come before leaders define what the assistants may decide.

Not for analysis to decide

AI may support defined workflows within explicit permissions. It may not silently acquire decision authority through convenience or platform design.

The disagreements are kept. Someone accountable has to choose between them.

Decide

Options, trade-offs and an accountable call.

Three credible courses of action. Each one is carried by assumptions that would need to be true. Choose the one you would take. Your choice is not scored.

Sign the company-wide agreementCommit to €8 million over three years and move at the pace of the competitor. For this to be right:

The projected €15.0m is real and reachable here.
Adoption and readiness will follow deployment.
Dependence on one provider is acceptable.

Run a bounded pilot, expand in stagesOne or two valuable, measurable workflows, with the foundations built alongside. For this to be right:

One or two workflows can show value that is measurable.
Real work reveals what must change faster than a readiness programme.
A provider, this one or another, will agree a bounded start.

Wait until the organisation is readyFix data, processes and governance first. Buy later. For this to be right:

Readiness can be built in the abstract, without real work to test it.
The cost of waiting is small.
The organisation will know when it is ready.

Christianna’s judgement

Reject the full agreement. Pilot, bounded and staged.

On the provider’s numbers the agreement returns €7.0m. At 35% adoption and 30% readiness, the realisable saving is €1.6m against €8.0m committed. A pilot on 2 workflows costs €1.2m, 15% of the contract, and produces the baseline that does not exist today.

The decision as signed, at the starting position

Reject the full agreement in its current form. Negotiate a bounded implementation around one or two valuable, measurable workflows, with staged expansion.

Do not wait for organisational perfection. Build the required foundations alongside the pilot, using real work to reveal what must change.

Another company’s announcement is not evidence that this investment creates value here.

Rejected: Signing the company-wide agreement now. Waiting for organisational perfection before starting.
The assumption carrying it: That one or two workflows can be bounded tightly enough to show value that is measurable.
Authority I keep: What AI may observe, recommend, draft, approve or execute, and every expansion, correction and termination gate.
What I would delegate to AI: Support for defined workflows within explicit permissions.
What would change my mind: I expand only after reliable data, real adoption, measurable value, clear authority and acceptable vendor dependence are demonstrated.

Technology cannot compensate for organisational unreadiness; deployed at scale, it often amplifies it.

Christianna Hamilton

This is one accountable judgement for the stated objective and evidence. It is not offered as the universal answer.

Act

Bounded actions, owners and approvals.

A decision is finished when it has owners, limits and a way of being checked.

First action

Choose the one or two workflows, name their accountable owners, and baseline cost, time and quality.

Measure to watch

Real adoption and output quality in the pilot workflows, against the baseline.

Baseline cost, time and quality.
FinanceDraft
Bound the workflows and name accountable owners.
OperationsHuman decision required
Do the minimum necessary data cleanup.
Technology and dataDraft
Set explicit observe, recommend, draft, approve and execute boundaries.
GovernanceHuman decision required
Measure real adoption and output quality.
PeopleObserve
Secure vendor portability and data-export rights.
RiskPrepare for approval
Set token, tool-step and cost ceilings.
Technology and dataPrepare for approval
Define expansion, correction and termination gates.
GovernanceHuman decision required

Where calculation ends and human authority begins

Calculate and surface. Evidence, assumptions, arithmetic and consequences are made visible.
Investigate and challenge. AI can extend the search, prepare options and test for conflict or unsupported certainty.
Decide and own. An accountable person weighs the trade-offs, makes the call and remains responsible for what follows.

Learn

What would show the judgement was wrong.

A judgement should say in advance what would overturn it. Change one material condition and see whether it holds.

The pilot proves the conditionsTwo quarters on, a baseline exists, authority is defined, export rights are signed, adoption is 75%, and the data and processes the pilot touched are 85% and 80% ready.The provider refuses a bounded startThe provider will only sell the company-wide agreement.

Next decisionFunding the present without selling the futureLive-entertainment company Previous decisionContain the downside while improving the evidenceIndustrial manufacturer

A decision of your own

If a situation in your organisation has this shape, I would be glad to talk it through.

hello@christiannahamilton.com

Method and provenance

The scenario is illustrative. It does not describe a client, an engagement or a real company.

What the tools do, and what remains human

Calculate and surface. Arithmetic and thresholds are handled consistently so the same inputs produce the same result.
Investigate and challenge. Functional positions expose disagreement. The critic looks for conflict, missing evidence, double counting and unsupported certainty. A live reading, where offered, can extend that challenge but cannot alter the underlying numbers or the signed decision.
Decide and own. The signed judgement is Christianna Hamilton’s. It is written in advance and remains the responsibility of a person, not a system.

Evidence

Items marked given in the scenario come from the written brief. Items marked illustrative modelling assumption are quantities added so that consequences can be calculated. Evidence last reviewed 2026-09-21.

Calculations in this decision

Quantity	Unit	What it assumes
Net benefit on the provider’s numbers	EUR million	The projection is achieved in full.
Organisational readiness	per cent	The weaker of data and process readiness binds.
Adoption multiplied by readiness	per cent	Savings need both a ready workflow and a person using the tool.
Realisable savings	EUR million	The projection is right about the ceiling, which is itself untested.
Realisable savings less the contract	EUR million	The full contract value is committed.
Adoption multiplied by readiness needed to break even	per cent	Break-even against the contract alone, before internal cost.
Distance from break-even	multiple	None beyond its inputs.
Cost of leaving the platform	EUR million	Workflows have come to depend on the platform.
Pilot as a share of the contract	per cent	None beyond its inputs.