Tenant Complaint Response Letters When ChatGPT Wrote the Complaint

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Tenant Complaint Response Letters When ChatGPT Wrote the Complaint.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Tenant Complaint Response Letters When ChatGPT Wrote the Complaint.

Treat a long, AI-written tenant complaint like any other: acknowledge it quickly, strip it down to the specific issues, check each against your records, then answer each one plainly with a date or next step. Tools like ChatGPT can speed the stripping and drafting, but a person must verify the facts and send the reply.

The length is the trap. A 1,800-word letter may hold only three real issues, plus citations, adjectives and threats that do not change what you must do. The process below separates the two, keeps you inside your complaints procedure, and flags the point at which the tenant can take the matter to your redress scheme.

Follow me on Instagram@sagnikteaches

What an AI-written complaint looks like, and what it is not

You will recognise it. Formal headings, numbered grievances, phrases such as "I am writing to formally lodge a complaint", references to statutes and codes of practice, a demand for a response in seven days, and an offer of "remedies" that range from a rent reduction to a regulator referral. It reads as confident, even when the facts underneath are thin or muddled.

Connect on LinkedInSagnik Bhattacharya

Three things to hold onto:

Subscribe on YouTube@sagnikteaches
  • The tool does not change the right. A tenant who used AI to write the letter has the same right to a proper answer as one who wrote it by hand.
  • Confidence is not accuracy. The statements about law, deadlines and entitlements may be wrong. The statements about what happened in the flat are the tenant's account and may be right.
  • Most of the volume is wrapping. Strip out the legal tone and you often find a boiler that failed, a deposit question and a missed callback.

Trade reporting on the trend is thin: I could not find a traceable figure for how many agencies now receive AI-written complaints, so none is given here. What matters is the handling, whichever way the trend goes. The same triage thinking applies to the repair requests covered in AI triage of maintenance requests: sort by what must happen, not by how the message is written.

The Complete Excel Course with AI Integration thumbnail

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

The seven-step process, with time budgets

Illustration: a six-person lettings agency receives a 1,600-word complaint from a tenant about a persistent damp patch, a missed repair appointment, an unanswered email, a deposit deduction threat and a "breach of the Code". The steps below show what happens to it, with rough times.

  1. Log and acknowledge (10 minutes). Open a complaint record, date-stamp receipt, and send a short acknowledgement the same day.
  2. Read once, unaided (15 minutes). Read the whole letter yourself before any AI touches it. You need your own sense of the tone and the risk.
  3. Extract the issues (10 minutes with AI, 10 to check). Produce a numbered list of specific, checkable complaints.
  4. Check each issue against your records (30 to 60 minutes). Repair log, call notes, emails, tenancy file, contractor reports.
  5. Draft the reply (15 minutes with AI, 15 to edit). One section per issue.
  6. Review and send (10 minutes). A second person reads it before it goes, for anything involving fault, money or legal wording.
  7. Close or escalate (5 minutes). Record the outcome, the date and what the tenant was told about their next step.

Total: roughly two to three hours of staff time, most of it in step 4, which is the part AI cannot do for you. Without the process, the same complaint often eats a whole day because someone tries to answer the letter paragraph by paragraph.

Step 1: the acknowledgement that buys time without conceding anything

Send this within one working day, or sooner if your procedure says so. It does not respond to the content.

Subject: Your complaint, reference [ref]

Dear [first name],

Thank you for your message of [date]. We have logged it as a formal complaint, reference [ref], and [name] will look into it.

You will receive our written response by [date]. If we need longer, we will tell you before that date and explain why.

Our complaints procedure is at [link or attached]. If you would like to add anything, reply to this email.

Kind regards,
[your name]

Check your own procedure for the response period before you set the date. The acknowledgement should promise only what your procedure says. If it has a two-stage process, say which stage this is.

Step 3: extracting the real issues with AI, safely

Before pasting anything, check your data rules. A complaint holds names, addresses, sometimes health, immigration or financial information. Use an account that does not train on your content, such as a business plan, redact names and addresses where you can, and keep the original in your case system. Our AI policy for tenant data tutorial covers the policy side. Then use a prompt like this:

You are helping a letting agent triage a tenant complaint. Below is the complaint text, with names replaced by [tenant] and [property].

Task:
1. List every specific complaint as a separate numbered item, one sentence each, using the tenant's own facts (dates, amounts, what happened). Do not add facts.
2. For each item, mark: FACT (something that happened that our records can confirm or deny), REQUEST (something the tenant wants us to do), LEGAL CLAIM (a statement about law or rights), or TONE (opinion or emotion with nothing to check).
3. List every date, amount, and named person or company mentioned.
4. List any threat (regulator, redress scheme, solicitor, press, withholding rent) exactly as written.
5. Say which items you are unsure how to classify.

Do not give legal advice. Do not draft a reply yet.

Complaint:
[paste]

Illustrative output for the damp-patch complaint:

#Issue as statedType
1Damp patch in bedroom first reported on 3 March, still presentFACT
2Contractor appointment on 18 March missed; nobody calledFACT
3Email of 20 March unansweredFACT
4Wants a rent reduction of 25% until fixedREQUEST
5Says the agency is in breach of "Section 12 of the Code"LEGAL CLAIM
6"Utterly negligent and deceitful management"TONE
7Threat to go to the redress scheme within seven daysTHREAT

You now have four things to check, one request to decide, one legal claim to leave unargued, one piece of tone to ignore and one threat to note. Read the list against the original. Did the AI drop anything? Did it merge two issues? A common slip is turning "I think the previous tenant caused it" into a stated fact. If it is not in the letter, it should not be in your list.

Step 4: checking facts against your own records

Build a one-page fact sheet. The AI cannot do this for you, because the evidence is in your system, not in the letter.

IssueWhat our records showSourceAgree / dispute / unknown
Damp first reported 3 MarchPhone note 4 March: "damp bedroom wall"; job raised 5 MarchCall log, repair systemAgree (reported 4 March, one day later than stated)
Appointment 18 March missedContractor note: "no access, tenant not in"; no call to tenant loggedContractor reportDispute on cause; agree no call was made
Email 20 March unansweredReply sent 27 MarchMail archiveAgree it took 5 working days
Rent reductionNot offered; no policy for itTenancy termsDecision for the principal

Notice what this does. It finds that the tenant's dates are slightly off and that the agency is, in part, at fault: nobody phoned about the missed appointment. A reply that denies everything would be wrong, and a reply that accepts everything would be wrong. The fact sheet is what lets you be accurate in both directions.

Mark anything you cannot confirm as "unknown" and do not guess. If a contractor's version is missing, request it today and tell the tenant you are waiting for it.

Step 5: drafting the reply section by section

Give the AI your fact sheet and your decisions, not the full complaint again. The prompt:

Draft a complaint response letter from a letting agency to a tenant. Use plain English, a calm and respectful tone, short paragraphs, and no legal citations.

Structure:
- Opening: thank them, confirm what we looked into.
- One heading per issue, numbered as in the list below. Under each: what our records show, whether we agree or not, what we are doing, and by when.
- Where we were at fault, apologise once, plainly, without excuses.
- Where we dispute something, give our reason and offer to look at any evidence they have.
- Closing: how to reply, and the next step if they are not satisfied, using this wording exactly: [paste your procedure's escalation wording].

Use only the facts below. If a date or decision is marked [CONFIRM], leave it in square brackets.

Facts and decisions:
[paste fact sheet and decisions]

A short extract from an illustrative draft, and what you would fix in it:

2. Missed appointment on 18 March
Our contractor's note says nobody was in when they arrived, and our records show we did not phone you to confirm the visit. We are sorry about that. I have rebooked the visit for [CONFIRM date and time] and asked the contractor to phone you the day before.

Fixes: the first sentence reads as blaming the tenant. Rewrite to "Our contractor recorded that they could not get in at the time of the visit", and confirm the access arrangement. Also check that "we are sorry" appears once in that section, not three times.

Handling the legal claims without arguing them

If the complaint says you have breached a named rule, you have three honest choices: confirm the rule exists and say whether you think it applies; say you are checking and will answer by a date; or answer the underlying practical issue and not mention the citation. Do not let the AI tell you whether the citation is right. It will often sound sure whichever way it goes. Verify against the source document, or ask your trade body, insurer helpline or solicitor.

You raised a concern about Section 12 of the Code. We are checking which obligations apply to the repair timeline and will confirm that to you by [date]. In the meantime, the practical steps above are in motion.

Rent reduction and compensation requests

These are decisions, not drafting questions. Someone with authority, usually the landlord or the agency principal, decides. The AI can set out the options with their consequences (no reduction, goodwill gesture, formal reduction until fixed), but the choice is not its to make. The letter then states the decision and the reason.

Step 6: the review checklist before sending

  1. Every date and amount matches the fact sheet.
  2. Nothing is promised that the landlord has not approved.
  3. No admission of legal liability beyond what the agency principal has agreed.
  4. No statement about law that you have not verified.
  5. Every issue in the numbered list has an answer, even if the answer is "not upheld, and here is why".
  6. The escalation paragraph uses your procedure's wording and the redress scheme's current details.
  7. No tenant data from another tenancy slipped in. This happens when the AI conversation was reused.
  8. The letter reads as written by a person who cares about the outcome, not a template.

Step 7: escalation, and what the redress scheme will look at

A tenant can usually take an unresolved complaint to the redress scheme your agency belongs to once your own procedure has run its course, and often sooner if you stop responding. One scheme's published process, for example, accepts a referral eight weeks after the complaint if no final response has arrived, and expects the referral within 12 months of the agent's final response. Time limits differ between schemes, so read yours directly and quote it in your procedure. Expect the scheme's questions to be practical: did you acknowledge in time, did you investigate, did you reply to every point, did you explain the next step. A file with an acknowledgement, a fact sheet, a reply to each issue and a dated log answers all of those. An AI-written complaint does not make the file any different, other than in how long the letter is.

Escalate internally, before the tenant escalates externally, when any of these appear: an allegation of discrimination or harassment, a health and safety risk to the occupants, a claim involving a landlord's legal duty such as gas, electrical or fire safety, a threat of legal action with a named solicitor, or a safeguarding concern. Those go to a principal or compliance lead the same day, and the AI is not involved in deciding what to do.

Four other versions of the same problem

The repeat complainant

A tenant who sends a long AI-written letter every fortnight. Keep one master case file, ask the AI to compare the new letter with the previous list and flag only what is new, and reply to the new points. State in the reply which issues are already being handled and when they will be closed. That cuts handling time and shows the scheme you did not ignore anything.

The letter with a genuinely serious allegation buried inside

Paragraph 14 of 22 mentions mould spreading to the child's room and a doctor's visit. A summary that drops it is dangerous. Always ask the AI to list every health, safety, safeguarding and discrimination reference separately, then read those passages yourself. The reply must deal with them first, with a date for action, not at the end.

The complaint about a landlord, sent to you

Some issues belong to the landlord, such as a decision to sell or refuse a repair above a limit. Your reply explains who decides, what you have passed on, and when you will update the tenant. Your landlord client needs to know a complaint is open, and the landlord updates and owner statements tutorial has wording for that.

A complaint with fabricated details

Occasionally a letter describes a conversation, a promise or a visit that never happened. Do not call it fabricated. Write "we have no record of this", say what records you did check, and invite the tenant to share a date, a name or a message that would help you find it. Record that you asked.

Before and after: taking the heat out of a draft

AI drafts of replies tend to mirror the tone of the complaint, or swing to over-apologetic. Compare these two versions of the same paragraph (illustrative).

BEFORE (first AI draft)
We sincerely apologise for the profound distress and inconvenience caused by our deeply regrettable failures, and we fully accept that our management has fallen short of every reasonable standard.

AFTER (edited)
We are sorry that the damp has still not been fixed and that nobody phoned you about the missed visit. The repair is now booked for [date], and I will phone you on [date] to confirm it went ahead.

The "before" version admits more than the facts support, uses adjectives the file does not back, and promises nothing. The "after" version says what went wrong, what happens next and who will do it. Use this test on every paragraph: would it still make sense if you deleted every adjective?

A deposit threat, answered from the file

Say a tenant's letter claims, in four paragraphs of formal language, that you plan to take $900 from the deposit unfairly for carpet cleaning. The fact sheet shows the check-out report notes stains in the lounge, the inventory at move-in recorded the carpet as "clean, professionally cleaned", and a cleaning quote of $180 exists. No deduction has actually been decided. The reply should say that no deduction has been decided, set out what the report shows, explain the process and date for any proposed deduction, and invite the tenant to send photographs. Notice how different that is from rebutting "unfair" and "unlawful" line by line. The AI is useful for the structure and the neutral wording; the amount and the decision come from the file and the landlord. If your tenancy deposit rules include a dispute service, quote your own procedure for it.

When the tenant needs a different format

Some tenants use AI because writing in English is hard, or because they are anxious and want to say everything. A formal 1,500-word reply may not serve them. Offer a short phone call after the written reply, send a one-page summary with the three actions and dates, and keep the full letter for the file. If the tenant has told you about a disability or vulnerability, record it and follow your procedure for reasonable adjustments. Using AI to write the original does not mean the person needs less care from you.

Summary for [first name]
1. Damp: contractor visit on [date], morning. We will phone the day before.
2. Rent: the landlord has decided [decision]. 
3. Next: I will call you on [date]. If you are not happy, our letter explains what you can do next.

A quick sum on what the process saves

Illustratively, an agency receiving four long complaints a month spent about six hours on each: reading, circling back, drafting by hand, fretting over wording. With the seven steps it spends about two and a half. That saves around 14 hours a month, which is worth the setup. The saving comes mostly from step 3 (no paragraph-by-paragraph reading loops) and step 5 (a structured first draft), not from replacing judgement. The cost is the discipline of the fact sheet, which takes the time it always took but happens once, not three times.

How to tell the process is working

Track four simple numbers in a spreadsheet for three months: days from receipt to acknowledgement, days to full reply, number of complaints that came back for a second round, and number that went to the redress scheme. A healthy pattern is acknowledgement within one working day, a reply within your procedure's period, and few second rounds because the first reply covered every point. If second rounds keep happening, look at your fact-checking step, not your drafting.

Also audit five closed complaints a quarter. Read the tenant's letter, your fact sheet and your reply side by side. Did the reply answer everything? Was any statement later shown to be wrong? Fix the process, not the person.

A last point on tools. If your complaint log lives in email, an assistant that drafts from the case file is more useful than one that works from a pasted letter, because it sees the repair history. That is a set-up job, and the tenant queries tutorial for letting agents shows where the earlier channel can head off complaints before they start.

Questions about AI-written tenant complaints

Should I tell the tenant I think an AI wrote the complaint?

No. It adds nothing, and you cannot prove it. A tenant may have used AI because writing is hard for them, English is a second language or they are stressed. Treat the content as the complaint. Answer the issues raised, in plain words, whatever tool shaped the wording.

Does a long complaint get a longer deadline?

Not automatically. Your published complaints procedure and your redress scheme's rules set the timetable, and length does not change it. If you need more time because of the number of issues, say so in the acknowledgement, give a date and keep to it. Check the exact wording of your own procedure.

Can I paste the tenant's complaint into ChatGPT to summarise it?

Only if your data policy allows it. A complaint holds names, addresses and sometimes health or financial details. Use a business plan with training off, remove identifiers where you can, and keep the original in your case system. Never rely on a summary without reading the source.

What if the complaint is full of legal citations that look wrong?

Do not argue each citation. Note which points you cannot verify, answer the underlying factual issue, and ask your insurer, trade body or solicitor if a legal claim is genuinely being threatened. AI-written letters often cite rules that do not exist or do not apply, but the underlying problem may be real.

Further reads

Sources: one property redress scheme's published complaint process (referral timings); OpenAI and Anthropic pages on business-plan data use; general complaint-handling practice for letting agencies. Read your own redress scheme and trade body guidance directly for your procedure. No statistics are cited; examples are illustrative.

Want a complaints workflow your team can actually run?

On a 1:1 call we'll map how complaints reach your agency today, decide which steps AI can speed up safely, and set a short checklist your staff can follow every time.

Book a 1:1 call with me