Beyond the
Magic Quadrant

Most management teams do not suffer from a shortage of lists. They have longlists of vendors, AI projects, process improvements, capability investments and strategic bets. The problem begins when every item is discussed in a different language. Finance talks about return. Operations talks about feasibility. Technology talks about architecture. Risk teams talk about exposure. Sponsors talk about urgency.

A single ranking compresses those different judgements into one number and often gives the appearance of agreement. A two-axis model does something more useful: it keeps two important questions separate long enough for the organisation to see where the real disagreement lies.

Why the Magic Quadrant idea travels

Gartner's Magic Quadrant™ is familiar because it turns a complicated market into a visible comparison. Gartner positions technology and service providers against two broad dimensions - Ability to Execute and Completeness of Vision - using up to 15 weighted criteria across those categories. Gartner also describes the output as a starting point rather than a stand-alone vendor selection tool, and explicitly notes that a challenger or niche provider may fit an organisation's needs better than a market leader. [1, 2]

That is the useful lesson to borrow. It is not the quadrant names, the vendor placements or Gartner's proprietary research process. It is the decision logic underneath the picture: two different forms of strength are assessed separately, then displayed in a way that makes comparison, challenge and trade-offs easier.

The quadrant is not the analysis. It is the visible surface of an argument about what matters.

An internal two-axis model should therefore not be presented as a homemade substitute for external market research. Its purpose is different. It helps an organisation structure a decision that depends on its own goals, evidence, constraints and appetite for risk. It can be used to compare AI opportunities, vendors, process candidates, product ideas, capability investments, training priorities or governance risks.

Breadth and fit solve different problems

The Magic Quadrant gains strength from broad external research, consistent headline dimensions and comparison across a defined market. It helps an organisation understand the supplier landscape from the outside.

An internal model can start from the organisation's actual decision and choose dimensions that expose local fit: compatibility with existing systems, the stability of the workflow, implementation capability, governance maturity, consequences of failure, acceptable dependency, service-delivery constraints or strategic importance.

Assuming the evidence is collected and analysed well, that customisation can make the model more decision-relevant - and therefore more useful for the choice at hand - even though it rests on a much smaller evidence universe. It is not more comprehensive than Gartner's research. It is better fitted to a narrower question.

On a narrow screen, each row is shown as a separate card.

The distinction is complementary rather than adversarial: external intelligence helps an organisation understand the market; internal modelling helps it decide what that market means for its own circumstances.

The trade-off is that customisation creates room for post-hoc engineering. A team can choose axes that favour the incumbent, the cheapest option or the project with the strongest sponsor. Customisation is a performance advantage only when the design choices are made before scoring and are recorded well enough to audit.

Done well, the model becomes a compact decision surface tailored to the organisation. Done badly, it becomes four boxes arranged around a conclusion someone had already chosen.

The axes are the decision

Managers often begin by asking what should go in each quadrant. That is too late. The first question is what the two axes should measure. Axis selection determines which forms of value become visible, which trade-offs can be discussed and which facts disappear from the picture. The axes should therefore be defined and approved before anyone sees where a favoured option will land.

Labels such as 'value', 'readiness', 'risk' or 'strategic importance' sound clear until several people try to score the same option. One person may interpret value as revenue. Another may mean service quality, risk reduction, time saved or organisational learning. Unless the terms are defined, the matrix merely relocates ambiguity into a more attractive shape. For each axis, record why it was selected, what it excludes and what evidence will be accepted.

Good decision-analysis guidance makes the same demand. Criteria should be measurable in qualitative or quantitative terms, should distinguish options in ways that matter to stakeholders, and should be sufficiently independent to avoid hidden causal overlap or double counting. [3, 4]

An illustrative AI portfolio model

Consider an illustrative professional-services organisation with six proposed AI initiatives: knowledge search, proposal drafting, contract review, meeting summarisation, client-report generation and compliance monitoring. Each sponsor believes their initiative is important. Technology sees different levels of integration difficulty. Risk teams see different consequences of failure. A single priority score encourages everyone to argue about the number.

The management team instead chooses two axes: Organisational Value and Delivery Readiness. These axes answer different questions.

Horizontal axis: Organisational Value

This measures the value available if the initiative works within acceptable limits. It might include:

the frequency and scale of the problem being addressed;

the likely effect on cost, revenue, service quality, risk or capacity;

alignment with the organisation's strategic priorities;

the extent to which the capability can be reused across teams or services;

the learning value created for future AI work.

Vertical axis: Delivery Readiness

This measures how prepared the organisation is to test or implement the initiative responsibly. It might include:

access to suitable data, documents and permissions;

stability and clarity of the underlying workflow;

availability and reliability of the required technology;

integration, security and operational support requirements;

the presence of an accountable owner, human review and governance route.

The distinction matters. A highly valuable initiative may be unready because the workflow is unstable, the data is inaccessible or nobody owns the consequences. A low-value initiative may be easy to deploy because the process is simple and the risk is modest. Ease is not importance, and importance is not readiness.

On a narrow screen, each row is shown as a separate card.

Illustrative action map: Organisational Value x Delivery Readiness. The labels are intentionally independent of Gartner's proprietary quadrant terminology.

The most important quadrant may not be Commit. In some organisations, Prepare contains the initiatives that deserve the most senior attention because their value is high but the conditions for responsible delivery are missing. The matrix is useful precisely because it separates what deserves investment from what is merely easy to do.

How to build a two-axis model that survives challenge

1. Write the decision question in one sentence

'Which AI initiatives should enter a controlled pilot this quarter?' is a decision question. 'What do we think about AI?' is not. The model should support one decision, not become a general picture of everything the organisation cares about.

2. Define the population

Only compare items that belong in the same decision set. A vendor, an internal workflow, a training programme and a governance policy may all matter, but they are not necessarily comparable on the same axes. If the items require different evidence or lead to different types of action, create separate models.

3. Define the evidence universe

Before choosing the axes, list the information the organisation can genuinely obtain: external research, internal performance data, workflow records, user evidence, technical assessments, risk registers and implementation experience. A model does not become more relevant merely because it is customisable. It becomes more relevant when the customisation is supported by better local evidence.

4. Choose and approve axes before scoring

The axes should be consequential and distinct. 'Commercial attractiveness' and 'revenue potential' may measure much the same thing. 'Strategic value' and 'delivery readiness' create a more useful tension because an initiative can be strong on one and weak on the other. Approve the axes, their rationale, their exclusions and the intended quadrant actions before scoring begins.

5. Define what low, medium and high mean

Give each axis a short set of criteria and anchored descriptions. A five-point scale can be useful, but the number is an index for structured discussion, not a claim of objective measurement. Record the evidence behind the score and the confidence attached to it.

6. Separate evidence from preference

Subject-matter experts should assess what the evidence says about an option's performance. Decision-makers should make the value judgement about which outcomes matter and how much. Combining those roles without discussion allows personal preference to masquerade as technical fact. Government MCDA guidance makes a similar distinction between experts assessing option performance and decision stakeholders expressing preferences. ³

7. Plot uncertainty, not just position

A precise dot can imply more certainty than the evidence supports. Use a confidence label, a range, or a larger circle for uncertain placements. An initiative near a boundary may need more evidence before the quadrant label is allowed to drive action.

8. Attach an action to every quadrant

If every quadrant leads to 'consider further', the model has not helped. Define the default action before plotting the items: commit, prepare, use selectively, park, investigate, escalate, or another decision appropriate to the problem.

9. Stress-test the result

Change the thresholds, weights or disputed assumptions. If a small adjustment moves an initiative from Park to Commit, the decision is fragile and should be presented as such. Sensitivity analysis is a central feature of robust multi-criteria decision work because it shows how dependent the result is on particular preferences and assumptions. ⁴

Where the model becomes management theatre

The axes are vague. If two managers cannot explain what 'high value' means in approximately the same terms, the matrix is not ready to use.

The axes are chosen after the answer is visible. Once the team knows which option it wants, it defines value, readiness or risk so that the preferred item lands well. This is the central danger of a custom model. Freeze the axis definitions, evidence rules and quadrant actions before scoring.

The axes overlap. If implementation cost reduces the readiness score and then reduces the value score again, the same weakness has been counted twice. Overlap can push an option away from the desirable quadrant without anyone noticing why.

Everything drifts towards the top right. Sponsors know which area looks desirable. Without anchored scales and independent challenge, optimism becomes a positioning strategy.

The chart hides uncertainty. A dot is treated as a fact even though the placement rests on weak evidence, a disputed forecast or a vendor claim. Confidence should travel with the position.

The picture is mistaken for the business case. A two-axis model is a light-touch sorting and discussion instrument. It is not a full multi-criteria analysis, cost-benefit analysis, procurement exercise or risk assessment. HM Treasury guidance explicitly distinguishes robust MCDA from simple scoring and weighting, and warns against using MCDA as a substitute for detailed economic appraisal. ⁴

No decision follows. The team spends an hour debating placement, admires the finished chart and returns to the same portfolio. A decision model earns its place only when it changes what happens next.

The model becomes an alibi. The matrix is presented as objective even though the axes, weights and thresholds were chosen to protect a preferred answer. The remedy is not more decorative precision. It is an inspectable record of who defined the model, what evidence was used, what assumptions were disputed and when the decision will be revisited.

Let AI support the analysis, not choose the values

AI can make the method faster and more rigorous without becoming the decision-maker. A document-aware system can review business cases, pilot reports, process maps, risk registers and meeting notes; extract evidence against defined criteria; flag contradictions; identify missing information; and generate an auditable rationale for each provisional placement. That can make local evidence more usable, but it does not turn a small or biased evidence base into a large one.

It can also create alternative maps. A leadership team might first compare Organisational Value with Delivery Readiness, then test Consequence of Failure against Evidential Confidence. The second view may reveal that an apparently attractive initiative is too consequential to proceed on weak evidence.

AI is particularly useful for sensitivity testing. It can recalculate placements under different weights, thresholds or confidence assumptions and show which decisions remain stable. It can track movement over time, distinguishing an initiative that is genuinely becoming more ready from one that has simply acquired a more persuasive sponsor.

But AI should not quietly decide that cost matters twice as much as service quality, or that speed matters more than accountability. Those are organisational value judgements. The system can expose their consequences; leadership must own them.

When two axes are not enough

The simplicity of a two-by-two matrix is its advantage and its limit. The University of Cambridge's Institute for Manufacturing describes the format as useful for initial sorting of qualitative data and for helping groups build a visible common understanding. ⁵ That is a good description of its proper role.

If the decision has several conflicting objectives, serious financial consequences, regulated risks or material effects on service users, the organisation may need a fuller multi-criteria process, cost-benefit analysis, formal procurement evaluation or specialist review. A more complex method is justified when the consequences of a poor decision make a simple model indefensible. ³

The two-axis model should help a team decide what deserves deeper work. It should not create false confidence that the deeper work has already been done.

The manager's test

Before accepting any internal quadrant, ask eight questions:

Do the axes measure genuinely different things?

Were the axes, evidence rules and quadrant actions approved before the candidates were scored?

Can the high and low ends be described using observable evidence?

Does the model use local evidence that an external market view would not capture?

Would two informed people place the same item in approximately the same area?

Does each quadrant imply a different management action?

Is uncertainty visible, or has it been converted into a falsely precise dot?

Would the model still be accepted if a favoured initiative landed in an unfavourable position?

That final question is the hardest. If the answer is no, the matrix is probably advocacy dressed as analysis.

The four boxes are the easy part

The Magic Quadrant is powerful as a visual idea because it gives people a shared language for comparing different forms of strength across a market. An internal two-axis model can do something no external analyst can do on an organisation's behalf: define the decision around its own workflows, constraints, risk appetite and capacity to act.

The best internal model does not compete with Gartner on breadth. It uses a smaller but more relevant evidence universe to answer a narrower question. Its value comes from defining the decision, choosing and approving defensible axes before scoring, making the evidence inspectable, showing uncertainty and attaching action. The quadrant should not end debate by pretending to be objective. It should make the organisation's reasoning visible enough to challenge.

External intelligence helps an organisation understand the market. Internal modelling helps it decide what that market means here.