Handling AI-Assisted Complaints and Correspondence

By /

SWANK AI Guidance Note 01
AI Governance · Communication · Administrative Systems

Core Standard

SUBSTANCE OVER TOOL

The use of artificial intelligence to assist with writing should not, by itself, determine whether a complaint or communication is credible, reasonable, repetitive or worthy of response.

The relevant questions are:

What does the communication actually say?

What evidence supports it?

What action does it require?

Purpose

Generative AI is increasingly used to:

  • organise information;
  • improve written communication;
  • formulate questions;
  • summarise complex material;
  • structure complaints or representations;
  • and navigate complicated administrative systems.

AI assistance can make communication more structured, detailed or polished.

That does not establish whether the underlying communication is accurate or inaccurate.

It does not establish whether the person raising the issue is reasonable or unreasonable.

And it does not establish whether the issue has already been resolved.

The appropriate institutional response is to assess the substance of the communication and the evidence relevant to it.

Substance Over Drafting Method

Organisations should avoid substituting assumptions about drafting technology for substantive review.

Where correspondence has been prepared with AI assistance, the operational questions remain:

  • What material issue is being raised?
  • Has that issue previously received a substantive response?
  • What remains unresolved?
  • What information requires verification?
  • What action, if any, is required?
  • Who is responsible for that action?
  • What is the appropriate response timeframe?

The presence of AI does not answer any of those questions.

Repetition Is Not the Same as Unresolved Engagement

Administrative systems should distinguish between:

repetition after an issue has received a substantive response

and

continued correspondence because an issue remains unanswered, partially answered or unclear.

These are operationally different conditions.

A communication may repeat information because the underlying matter has genuinely been resolved.

It may also repeat information because the sender is still trying to obtain an answer to a question that has not been substantively addressed.

Treating both situations identically can obscure the actual state of the administrative process.

AI and Accessibility

Generative AI may also function as an assistive communication technology.

It may support people who experience difficulties involving:

  • speech;
  • writing;
  • language;
  • cognitive load;
  • fatigue;
  • information organisation;
  • or navigating complex administrative processes.

For some people, AI assistance may make participation substantially easier.

Organisations should therefore avoid assuming that correspondence is inherently problematic because it is:

  • highly structured;
  • unusually detailed;
  • formally written;
  • grammatically polished;
  • or clearly produced with technological assistance.

The relevant question remains whether the communication contains material issues requiring consideration.

Accuracy Still Matters

AI assistance does not remove the need for verification.

Where a communication contains material:

  • factual claims;
  • legal propositions;
  • technical information;
  • statistics;
  • quotations;
  • citations;
  • or other consequential assertions,

those matters should be verified where necessary before reliance.

The correct principle works in both directions:

AI assistance does not make a communication inherently unreliable.

But:

AI assistance does not make its factual contents inherently reliable either.

Evidence remains necessary.

Administrative Triage

Where organisations receive large volumes of correspondence, AI may assist with administrative functions such as:

  • identifying subjects;
  • detecting deadlines;
  • routing communications;
  • recognising missing information;
  • grouping related correspondence;
  • or flagging material requiring human review.

That is different from allowing AI to determine the substantive merit of what has been raised.

Triage may help answer:

What requires attention?

It should not silently answer:

What is true?

Human Review

Where correspondence could materially affect:

  • access to services;
  • safeguarding;
  • education;
  • healthcare;
  • complaints outcomes;
  • legal rights;
  • disciplinary processes;
  • or other consequential interests,

meaningful human review should remain available.

A responsible reviewer should be capable of:

  • identifying the substantive issue;
  • locating relevant source material;
  • distinguishing evidence from interpretation;
  • recognising unresolved questions;
  • correcting inaccurate summaries;
  • and determining what action is actually required.

Human review should involve judgment rather than automatic acceptance of an AI-generated classification.

Record Integrity

If AI is used internally to summarise correspondence, organisations should preserve the relationship between the summary and the original source.

A later reviewer should be able to determine:

  • what the person actually communicated;
  • what information was summarised;
  • what was omitted;
  • whether anything remained disputed;
  • whether uncertainty was preserved;
  • and whether the summary accurately reflects the underlying communication.

A summary should assist access to the source.

It should not replace the source where context matters.

SIAAF Relevance

This Guidance Note principally relates to:

Domain 03 — Evidence & Traceability

Can material conclusions be traced back to the communication and evidence from which they arose?

Domain 04 — Communication & Feedback Integrity

Does the institution accurately receive, interpret and respond to information?

Domain 05 — Escalation, Challenge & Contestability

Can an individual clarify, challenge or correct an institutional interpretation where necessary?

It may also engage Domain 07 — AI Literacy & Organisational Readiness where staff require sufficient understanding to distinguish legitimate AI assistance from unreliable AI-generated content.

Questions for Organisations

Where AI-assisted correspondence is received, organisations may ask:

  1. What substantive issue is being raised?
  2. What evidence supports it?
  3. Has the issue actually been answered?
  4. What remains unresolved?
  5. Does any material claim require verification?
  6. Is AI assistance being confused with substantive unreliability?
  7. Could accessibility needs explain the use of AI?
  8. Has internal summarisation preserved the original meaning?
  9. Can inaccurate information be corrected?
  10. Is meaningful human review available where consequences are significant?

SWANK AI Standard

SUBSTANCE OVER TOOL

Assess what the communication says, what evidence supports it and what action it requires.

Do not substitute assumptions about the technology used to draft a communication for substantive review.

AI assistance neither removes the need for evidence nor makes the communication inherently unreliable.

Related SWANK AI Guidance

Guidance Note 02 — AI Summarisation and Record Integrity

Guidance Note 07 — Public-Sector AI and Accessibility

Guidance Note 12 — AI and Administrative Fairness

Guidance Note 16 — AI Triage and Prioritisation

Guidance Note 19 — Correction Rights in AI-Assisted Systems

SWANK AI

Independent AI & Institutional Assurance

We do not just review AI. We review the institutional systems responsible for governing it.

Scroll to Top