Skip to main content
ReportSwift home

Tailored AI systems for organisational psychology consultancies

Connect the work from assessment data to professional insight to finished report.

What a reliable reporting system must control

A reliable report depends on a defensible production process and sound use of the evidence. Both matter to client trust and professional accountability.

Question 1 of 2

Can you defend how the report was produced?

01 —

Data Governance & Confidentiality

Sensitive assessment data requires a workflow whose hosting, access, retention, and deletion controls match the consultancy’s client and professional obligations.

Read detail

General-purpose AI services differ materially in how they store conversations, retain files, process data, and expose administrative controls. Suitability depends on the selected service, contract, configuration, hosting region, subprocessors, and the consultancy’s own obligations.

When identifiable assessment data is used without those conditions being defined, the consultancy may lose control over where the data is held, who can access it, and when it is deleted. The risk comes from the implementation—not from the product category alone.

02 —

Method Control

Scoring rules, construct definitions, comparison logic, and interpretation boundaries need to be explicit, testable, and carried through the reporting workflow.

Read detail

Some reporting decisions have one correct result. Scores, lookups, thresholds, transformations, and response counts should be calculated with fixed rules so the same input produces the same output.

Other decisions require professional interpretation. A language model can work within agreed boundaries, but a prompt does not automatically make those boundaries explicit or testable. The reporting method may begin as an established approach or take shape as the consultancy’s requirements are clarified. In either case, the system needs a clear account of what each result can support.

The production test is concrete: where does each rule or boundary live, how is it tested, and what happens when it is not met?

03 —

Evidence Traceability

Each material conclusion needs a reviewable path through its supporting insight and source-linked facts to the approved evidence behind it.

Read detail

A finished paragraph does not show how the report moved from scores and comments to a conclusion. Without a retained path, a reviewer has to reconstruct that movement from the original material.

Traceability keeps the relationship between the narrative, its supporting insight, the contributing facts, and the original sources available for inspection. An investigation can begin with the conclusion in question instead of rereading the entire source set.

This path does not prove that the conclusion is correct. It makes the basis of the conclusion visible so its support and interpretation can be examined.

Why both questions matter

Even when the process is controlled and traceable, the report can still omit qualifying evidence, misread the context, or express a conclusion more strongly than the evidence supports.

Question 2 of 2

Does the report make sound use of the evidence?

04 —

Evidence Coverage & Conflict

A report can use genuine scores and comments while omitting qualifying evidence, flattening disagreement, or making a conclusion broader than the evidence supports.

Read detail

Fluency can make these failures difficult to see. A paragraph may contain no invented information yet give one respondent group too much weight, lose a material conflict, or turn one observation into a stable behavioural pattern.

The workflow therefore needs explicit coverage and materiality checks: what evidence supports the conclusion, what qualifies it, what disagrees, and whether the interpretation is ready for narrative use.

05 —

Contextual Judgement

Assessment results need to be interpreted in relation to the report’s purpose, role, audience, organisational setting, and comparison logic.

Read detail

A score or comment does not have one fixed meaning outside its use. The same pattern may carry different implications in selection, development, or coaching, and its importance may change with role demands and the surrounding evidence.

Supplying contextual information to a language model does not ensure that it will be applied correctly. If the purpose, role demands, comparison logic, or interpretation boundaries are absent or loosely defined, the system may draw a conclusion that the evidence does not support in that setting.

The workflow needs those contextual anchors to govern how facts become insights for this report.

06 —

Professional Voice

Professional voice governs how supported facts and insights are expressed: what receives emphasis, how certainty is qualified, and how findings are translated into practical meaning.

Read detail

Professional voice is not a list of preferred words or a request to imitate earlier reports. It is the consultancy’s recurring way of representing facts, forming insights, and expressing its professional position.

A language model can reproduce a familiar tone while losing the reasoning behind a consultant’s emphasis, qualification, or level of certainty. The result may sound right while changing the report’s professional position: its balance, certainty, and practical implication.

A controlled workflow turns consultant feedback into reusable guidance across fact, insight, and narrative decisions, then tests that guidance on new cases within the defined report type, purpose, and methodology.

Fluent output alone cannot resolve these risks. They require explicit controls across the full workflow.

Reporting systems in production

Franchise Relationships Institute and Lixivium Consulting use custom reporting systems designed and built by ReportSwift to produce assessment reports for client delivery.

Client system 01

Franchise Relationships Institute (FRI)

What it produces
An automated reporting workflow for Franchise Mentor surveys that turns multi-rater data into three connected reports: the franchisee’s self-rating report, the franchisor’s rating of the franchisee, and a comparison report that brings the two perspectives together.
Where AI adds value
Report templates present the scores and fixed content. AI synthesises the survey evidence into report-specific summaries, practical tips, and coaching considerations.
Evidence and review
The system maintains links between the source data and the generated narrative. A dedicated audit step compares the narrative with the source evidence before the reports proceed to client delivery.

Client system 02

Lixivium Consulting

What it produces
A custom 360-degree feedback platform connecting participant and rater administration, data collection, analysis, and PowerPoint report production.
Where AI adds value
Established templates present quantitative scores, charts, and fixed content. AI combines the rating results with written feedback to generate the analysis and narrative for Areas of Strength and Areas for Development.
Evidence and review
The platform prepares the complete report draft and retains a deliberate consultant review stage. The consultant reviews the generated analysis alongside the source data before using the report in the participant feedback session.

Client perspective

“He quickly understood the practical requirements of our work, including the need to manage large volumes of qualitative and quantitative feedback, produce clear and accurate outputs, and create a system that is both efficient and easy to use.

Ding brought strong technical capability, thoughtful problem solving and a genuine interest in understanding the end user experience. The platform has significantly streamlined the way 360 feedback is processed, reducing manual effort while improving consistency, speed and quality.”

Warren Senn Director, Lixivium Consulting

Discuss whether ReportSwift fits your workflow.

Contact Ding

Prototype recognition

Lucid — Overall Track Prize, 2025 Google DeepMind Gemini 3 Hackathon

Lucid, a ReportSwift prototype, received an Overall Track Prize in the 2025 Google DeepMind Gemini 3 Hackathon.

The prototype explored how evidence from multiple sources could remain linked through AI analysis and conflict review. ReportSwift has continued this evidence-led design direction in its later Evidence Chain work.

Featured by FRI

FRI featured Lucid’s Google DeepMind recognition in its August 2026 newsletter, naming Ding Wang as the developer behind the software and an FRI collaborator.

View the award announcement

How the architecture implements these controls

The Anchored Narrative Engine uses the report’s purpose, role, audience, and methodology to constrain narrative decisions before writing begins. Within that approach, the Evidence Chain preserves evidentiary support and Adaptive Voice Calibration guides professional expression. Fixed rules handle scoring and comparison calculations; AI supports interpretation and writing within defined boundaries.

Method 01

The Evidence Chain

The Evidence Chain carries authorised source material through fact extraction and insight formation before narrative generation begins. It keeps supporting evidence, disagreement, and qualification attached to the material available for writing, while preserving a path back to the source when a claim needs investigation. The diagram shows the control design; its effectiveness depends on implementation and testing in the reporting context.

Explore the Evidence Chain Show diagram and stage detailsHide diagram and stage details
Conceptual Evidence Chain: authorised sources feed fact extraction; report purpose, role, audience, and methodology govern insight formation. Facts retain source references and context. Insights proceed to writing only when support and interpretation criteria are met, with qualifications retained; unresolved insights return for investigation. A separate audit compares the draft and evidence path with sources retained from the assessment dossier. Release rules then determine proceed or hold. Held drafts return to the affected stage for correction, regeneration, and another audit before a new release decision.

Sources authorised for use

Assessment dossier

Scores, qualitative evidence, benchmarks, and relevant context accepted for use in the report.

Feeds Fact Extraction. A source copy is retained for the audit.

Anchored Narrative Engine

Context & method

Report purpose, role, audience, and methodology define the interpretation boundaries.

Governs insight formation and narrative use

01 Source-linked facts

Fact Extraction

Record source-linked facts with attribution and context.

Works from: authorised source material

Controlled handoff

Source-linked facts proceed

Context and source references retained

02 Evidence-linked insights

Insight Formation

Form candidate interpretations, retaining disagreement and qualification.

Works from: source-linked facts, context, and method

Narrative-use gate

Support and interpretation criteria met?

Yes: carry facts and qualifications forward.

No: hold insight and return to Insight Formation for investigation.

03 Evidence-linked claims

Narrative Generation

Write from cleared insights with evidence links and qualifications intact.

Works from: insights ready for use and supporting facts

Reviewable evidence path

Authorised sourceSource-linked factEvidence-linked insightMaterial report claim

Separate checking stage

Traceability & Coverage Audit

Compare the draft with its evidence path and retained sources. Record findings without rewriting the report.

From the assessment dossier

Audit copy

Authorised sources retained separately from the generated facts, insights, and prose.

Material under review

Draft + reviewable evidence path

Material claims are checked for source support, contradiction, omission, and required coverage.

Output: findings on source support, attribution, lost qualifications, contradictions, and omissions, linked to the affected claims.

Assurance decision

Apply release rules

Use audit findings, required checks, and escalation criteria to record whether the draft may proceed or must be held.

Proceed

No issue requiring a hold detected

The draft meets the configured conditions for the next step.

Next configured step

Proceed under the workflow’s release rules

Each implementation defines consultant review, escalation, and release responsibilities. A draft may continue automatically only within an approved and tested boundary.

Hold

Investigation required

A finding or a case outside the configured boundary requires correction or professional escalation.

Held from client release

Correction route

Return to the affected stage
  1. Investigate the flagged evidence path and resolve the issue or obtain professional direction.
  2. Correct the affected facts or insights, then regenerate dependent prose; wording-only corrections stay at the narrative stage.
  3. Return to Traceability & Coverage Audit, then apply release rules again. Unresolved cases remain held.

Evidence Chain stage details

01 — Fact Extraction
Record what each score showed or source reported, keeping attribution and context attached.
02 — Insight Formation
Form candidate interpretations from the facts, retaining disagreement and qualification within the defined context and method.
03 — Narrative Generation
Write from insights cleared for use, preserving their evidence links, disagreement, and qualification.

Method contribution

The method constrains what the AI writing step can use before prose is generated. The same evidence path supports a later audit and targeted investigation. When new context changes an interpretation, the affected facts, insights, and narrative need to be updated through that path.

Source-linked facts record what sources reported, rather than proving reported behaviour. Traceability and a separate audit do not by themselves establish interpretive validity or independent assurance.

Method 02

Adaptive Voice Calibration

Adaptive Voice Calibration turns consultant feedback into reusable guidance across fact representation, insight formation, and narrative expression. The guidance addresses how the consultancy balances evidence, certainty, and practical meaning—not only vocabulary and tone—and is tested on new cases within a defined report type, purpose, and methodology.

Explore Adaptive Voice Calibration Show diagram and stage detailsHide diagram and stage details
Adaptive Voice Calibration begins with existing reports or controlled alternatives, compares controlled report samples, uses consultant feedback to guide internal refinement across fact representation, insight formation, and narrative expression, and adopts the guidance as the workflow standard only after consultant acceptance criteria are met on new cases.

Calibration starting point

Existing reports or controlled alternatives

Established workflows can use previous reports as evidence of the consultancy’s existing judgement. New workflows can use controlled alternatives while keeping the evidence, report purpose, and methodology stable.

Guidance development

Trace feedback to its root cause

Consultants describe the professional position the report should carry. ReportSwift traces their feedback to fact representation, insight formation, or narrative expression and identifies the change required.

Reusable guidance

Professional voice guidance

Fact representationEvidence balanceCertaintyNarrative expression

Iterative calibration loop

Compare, diagnose, and refine

Each iteration keeps the reference conditions stable while consultant feedback guides internal diagnosis and refinement across fact representation, insight formation, and narrative expression.

01 — Reference cases

Stable comparison conditions

Keep evidence, report purpose, and methodology stable.

02 — Sample generation

Controlled report sample

Apply current guidance to produce a controlled report sample.

03 — Consultant review

Professional feedback

Consultants explain the professional direction and reasons for change.

04 — Internal refinement

Stage-specific refinement

ReportSwift diagnoses the cause and refines the relevant guidance.

Assessment scoring and any fixed methodological weights remain unchanged.

Professional feedback informs stage-specific refinement before the guidance is tested in the next controlled report sample.

New-case review

Consultant acceptance criteria

The guidance is applied to cases not used during its development. Consultants confirm whether the reports carry the intended professional voice within the defined workflow. This acceptance concerns voice; it does not validate the assessment instrument, interpretation, or wider reporting system.

Further refinement

Acceptance criteria not yet met

Consultant feedback returns to stage-specific refinement. The revised guidance is then tested on another case not used during its development.

Guidance not yet adopted

Accepted for defined workflow

Acceptance criteria met

The guidance becomes the standard starting point for the defined report type, purpose, and methodology.

Workflow standard

Accepted voice guidance

Reusable guidance within the calibrated workflow, reconsidered when the methodology, report purpose, or available evidence changes.

Applied through the Evidence Chain

Fact, insight, and narrative stages

Calibration loop details

01 — Reference cases
Representative cases reflect the defined reporting workflow, including mixed-feedback and boundary cases where relevant. Evidence, report purpose, and methodology remain stable across comparison rounds.
02 — Sample generation
The current guidance is applied to each reference case, producing a report sample for comparison with the intended professional voice.
03 — Consultant review
Consultants assess the sample’s evidence balance, certainty, practical meaning, and narrative expression against the intended professional voice. Their feedback explains the direction and reasons for any required change.
04 — Internal refinement
ReportSwift uses internal rounds to diagnose the cause and refine the relevant guidance before returning a complete report sample for consultant feedback. The guidance captures the reason for the change, with minimal examples.

Method contribution

The method turns consultant feedback into reusable guidance across fact representation, insight formation, and narrative expression. Within the calibrated workflow, it supports a consistent professional voice and is designed to reduce repeated correction of the same underlying issues.

Primarily supports Professional Voice and contributes to Contextual Judgement without changing the underlying Method Control.

Founded on over a decade of reporting systems experience

ReportSwift founder Ding Wang has spent more than ten years building reporting systems for organisational psychology research and consulting. His background combines a master’s in psychology with a bachelor’s in information and computing science.

Ding Wang, founder of ReportSwift

Ding Wang

Founder and systems developer, ReportSwift

Reporting systems for research and consulting

From assessment data to reports used in practice

Ding built reporting systems within the Centre for Transformative Work Design at the University of Western Australia and the Future of Work Institute at Curtin University. As the sole developer supporting multiple research and consulting programs, he translated their reporting requirements into working software.

The work required systems that followed established assessment methodologies, handled complex evidence, and produced reports for professional review against project deadlines. Across this work, his systems produced more than 30,000 psychometric reports and delivered nearly 14,000 dashboards.

Psychometric reports
30,000+
Dashboards delivered
Nearly 14,000
Research contribution
Reporting systems used in doctoral research

Academic qualifications

Psychology and mathematical computing

Ding holds a Master’s in Psychology from the Institute of Psychology, Chinese Academy of Sciences, with a foundation in psychological research and evidence interpretation.

He also holds a Bachelor’s in Information and Computing Science from Beijing Jiaotong University, with a foundation in mathematics and computing.

The report and dashboard totals reflect Ding’s reporting systems experience; they are not measures of ReportSwift’s AI performance. To assess whether ReportSwift fits your consultancy, start with your methodology and recurring reporting needs.

Who this is built for

ReportSwift builds tailored AI reporting systems for organisational psychology consultancies whose recurring assessment work requires substantial analysis and report preparation. The starting point can be an existing reporting process or a new workflow developed with your team.

Fit criteria

Indicators of a good fit

  • You produce assessment or 360-degree feedback reports regularly, or are developing a service that will.
  • Analysis, writing, or report assembly takes substantial consultant time or limits how much work your team can deliver.
  • You need a system that follows your assessment methodology, keeps conclusions linked to evidence, and handles sensitive data under agreed controls.
  • Your consultants can help define the reporting requirements, review sample reports, and refine how the system expresses their professional judgement.

Other options

When another option may suit you better

  • You need occasional writing help or a one-off report, with little recurring work to automate.
  • A ready-made reporting product or template already meets your requirements.
  • You need an immediate solution and cannot commit time to defining requirements and reviewing report samples.

Commercial model

How pricing works

Pricing combines a one-off customisation fee with a per-report fee. Both reflect your agreed scope and reporting requirements.

One-off customisation fee
Covers defining your reporting requirements, mapping your assessment methodology, agreeing data controls, and building the workflow. Your consultants’ feedback on sample reports helps refine how the system expresses their professional judgement.
Per-report fee
Includes AI processing, hosting, and all system maintenance. You pay only when a report is generated, at a rate that reflects your data complexity and reporting requirements.