The first answershould not be the tool.
Technology becomes a useful decision when the business can explain the result it wants, what may be preventing that result, and what evidence could alter its view. Perspective is the work of making those conditions visible.
See what perspective changesA technology request is already a conclusion.
“We need an AI assistant” sounds like a starting point. It already assumes that the problem is suitable for AI, that assistance belongs in the current workflow, and that the surrounding information is reliable enough to support it.
Perspective temporarily reopens those assumptions. It considers the intended business difference alongside the work, information, systems, people and uncertainty involved. Research on information-technology investment supports examining organisational changes and complementary capabilities together with the technology itself.1
- Request
- Introduce an AI assistant.
- Perspective
- What is limiting the quality, speed or reliability of this work?
- Decision
- Which response fits the cause, the consequence and the organisation’s capacity?
Open the interpretation before narrowing the response.
A proposed tool enters from the left. Perspective examines what the proposal assumes. What leaves the right side is not a product specification; it is a performance question that can guide the technical decision.
A product category has arrived before the problem has been adequately framed.
Which difference matters, and why now?
What is observed, and what remains assumed?
Where do records, rules and responsibilities connect?
Which bounded task might benefit—and where can it fail?
How will people understand, correct and complete the work?
The outcome can now be examined without committing prematurely to one technical response.
Perspective defines the change. Performance tests whether it happens.
The two inquiries meet at a practical handover. Perspective has done enough when the organisation can describe the intended difference, the plausible causes of the present condition and the uncertainty that still matters. It can then hand Performance a testable proposition: how would that difference appear in real work?
What deserves to change?
It widens interpretation before selection and keeps alternative responses available.
- The outcome worth pursuing
- The conditions that may shape it
- The assumptions still in the way
What would better work look like?
It defines observable conditions and examines benefits, costs and displaced effort.
- The signals of useful change
- The people and work affected
- The evidence for continuing or adapting
One request can conceal different problems.
The example supports the argument; it is not the argument. Nothing here reports measured customer results.
“Customers keep returning. Let AI answer them faster.”
Speed may not be the constraint.
Customers may return because records conflict, the service rule is unclear or the first reply does not resolve the need.
Why does resolution fail?
Trace the enquiry, source information, handovers, exceptions and points where a person must judge.
Define useful change.
Examine resolution, repeat contact, review effort and exceptions before deciding whether AI belongs in the response.
Each concern can change the decision.
They do not take turns. They meet around the same business question and reveal different conditions that a technology choice must respect.
Strategic intelligence
Which difference is worth pursuing?
Connect internal evidence and external change to priorities, capability and accountable choice.2
Technology can enable a direction. It cannot decide which direction deserves commitment.
Business intelligence
What does the evidence show?
Define the event, measure, population and time window needed to see the work accurately.
A pattern can guide inquiry without proving its cause.
Information systems
What must remain connected?
Follow information quality, system quality, service quality and use across records, roles and handovers.3
Fast integration can still preserve ambiguity about which record is correct.
Artificial intelligence
Which task could benefit, within what limits?
Separate classification, retrieval, summarisation and generation; each needs its own inputs, evaluation and response to failure.4
Fluent output can still be unsupported, outdated or wrong for the case.
User interface & experience
Can people understand and complete the work?
Consider the whole journey: what a person sees, what the system withholds, and how correction, escalation and recovery happen.
A clear screen cannot repair missing information, but it can expose uncertainty and support responsible action.
Do not ask technology to define the progress it is meant to serve.
Keep the response open until the intended difference, the relevant conditions and the remaining uncertainty can be stated together. Then move to Performance: define what improvement would look like in the work, including its costs and consequences.
- The intended difference
- What should become possible?
- The present condition
- What may be preventing it?
- The consequential uncertainty
- What evidence could change the decision?
Sources and notes
The example is hypothetical. The five concerns are an editorial guide; the sources support particular ideas, not a validated five-part model.
- Brynjolfsson & Hitt, 2000 — Beyond Computation. Supports examining organisational changes and complementary capabilities alongside information-technology investment; it does not validate this page’s structure.↩
- Pellissier & Kruger, 2011 and Maccoby, 2015 provide business and leadership formulations of strategic intelligence. The definition here is an editorial synthesis.↩
- DeLone & McLean, 2003 — Information Systems Success. Supports examining several aspects of information-systems success, including quality, use and user satisfaction.↩
- NIST, 2023 — AI Risk Management Framework 1.0. Supports examining context, evaluating systems, managing risks and considering non-AI alternatives.↩