The problem
Receiving a complaint is stressful, even when you are confident you acted correctly. The process of responding properly is also time-consuming and requires care. The FCA's complaint handling rules are specific: complaints must be acknowledged promptly, investigated thoroughly, and responded to within defined timescales. The response letter itself must address every point raised by the complainant, explain the investigation undertaken, state the firm's conclusion, and set out the client's right to escalate to the Financial Ombudsman Service if they remain dissatisfied.
Getting this wrong has real consequences. A response that does not address all the points raised looks dismissive and is more likely to result in an FOS referral. A response that is overly defensive or combative can inflame a situation that could have been resolved. A response that inadvertently admits liability where none exists can create problems the original complaint never would have. Writing good complaint responses requires balancing empathy, factual accuracy, regulatory compliance, and legal caution simultaneously, which is genuinely difficult under time pressure.
Many smaller firms handle complaints infrequently enough that there is no well-worn process or template library to draw on. Each complaint arrives as a novel situation requiring careful thought. AI cannot replace that thought, but it can provide the structure and drafting support that makes the response faster to produce and more thorough in its coverage.
The system
Step 1: Analyse the complaint and identify all points raised (Claude)
Before drafting anything, use Claude to ensure you have identified every discrete point in the complaint. Clients often raise multiple issues in a single letter, and missing any of them is a compliance failure.
Prompt example: "I have received a complaint letter from a mortgage client. Please read the following letter and: (1) identify every distinct complaint point raised, (2) categorise each point by type (e.g. advice quality, communication, process delay, product outcome, charges), (3) flag any regulatory or legal issues I should be aware of, and (4) suggest what evidence or documentation I should review before responding. Here is the complaint letter: [paste letter]"
This analysis should be reviewed carefully. Add any points you believe Claude has missed and remove any that are not genuinely raised in the complaint.
Step 2: Document your investigation (Claude)
Before drafting the response, use Claude to help structure your investigation notes. This ensures your response is grounded in the facts you have actually reviewed.
Prompt example: "I have reviewed the client file in response to this complaint. Please help me structure my investigation notes in a format suitable for the complaint file. The key facts I have established are: [paste your notes]. Please organise these by complaint point, noting for each: the factual position, the evidence reviewed, and my conclusion on whether the complaint point is upheld, partially upheld, or not upheld."
Step 3: Draft the response letter (Claude)
With your investigation documented, use Claude to draft the full response letter. Be specific about the regulatory requirements the letter must meet.
Prompt example: "Please draft an FCA-compliant complaint response letter based on the following information. The letter must: address every complaint point identified, confirm whether each point is upheld or not upheld and give clear reasons, explain the investigation undertaken, be empathetic in tone without being defensive or admitting liability where none exists, include the standard FOS referral information and the correct time limit for the client to refer to FOS, and conclude with a clear statement of any remedy offered (if applicable) or explanation of why no remedy is offered. Here are the complaint points, investigation notes, and conclusions: [paste your structured notes]"
Review the draft carefully. Ensure it accurately reflects your investigation and conclusions. Add or amend anything that does not match the facts of the case. Check that the FOS referral language matches current FCA requirements.
Step 4: Review for tone and compliance (Notion AI)
Copy the draft into Notion and use Notion AI to review the tone and check for any language that could be problematic.
Prompt example: "Please review this complaint response letter and flag: (1) any language that could be read as admitting liability unintentionally, (2) any sections where the tone is defensive rather than empathetic, (3) any complaint points from the original letter that do not appear to be addressed, (4) any regulatory information that appears to be missing or incomplete. Please suggest specific rewrites for any flagged sections."
Step 5: Build a template library (Notion AI)
Over time, build a library of response components in Notion for the most common complaint types you encounter: delays in processing, advice suitability concerns, rate expectations not met, and so on. Use Notion AI to help draft and refine these components so future responses can be assembled faster.
The results
Before: Drafting a thorough complaint response from scratch typically took two to three hours, including reviewing the file, structuring the response, and checking the regulatory requirements. Under time pressure, responses were sometimes less thorough than they should have been, and firms occasionally received FOS referrals on complaints that could have been resolved at firm level with a better initial response.
After: The structured AI-assisted approach produces a more thorough and compliant response in 45 to 60 minutes. One firm reported that their FOS referral rate fell noticeably after implementing a structured complaint response process, attributing the improvement to higher quality initial responses that addressed all points raised and were more empathetic in tone. The time saving is approximately 90 minutes per complaint, and the quality improvement carries its own significant value.