Customer Education | Dialogue API

Most financial services education content is written around what the product team thinks customers should know, not what customers are actually confused about. The result is generic jargon busters and FAQs that miss the real gaps.

The Dialogue corpus contains thousands of real consumer comments where people ask questions, express confusion, repeat myths, and reveal fundamental misunderstandings about financial products. Mining these systematically gives you a content strategy built on genuine need rather than assumption.

Step 1: Find what people genuinely don’t understand

Start by searching for confusion, surprise, and realisation — the moments where people discover something they didn’t know.

Things people misunderstood about pensions:

$ POST /api/v1/search
$ {
>   "query": "didn't realise didn't know thought pension worked differently",
>   "limit": 15,
>   "threshold": 0.4,
>   "product": "pension_sipp"
> }

Confusion about life insurance:

$ POST /api/v1/search
$ {
>   "query": "confused about life insurance how it works what it covers",
>   "limit": 15,
>   "threshold": 0.4,
>   "product": "life_insurance"
> }

Mortgage misconceptions:

$ POST /api/v1/search
$ {
>   "query": "thought mortgage worked that way didn't understand",
>   "limit": 15,
>   "threshold": 0.4,
>   "product": "mortgage"
> }

These comments show you where the real knowledge gaps are — often very different from what the compliance team assumes needs explaining.

Step 2: Find the questions people are asking

Questions with high like counts are particularly valuable — they represent a gap that thousands of people share but only one person was willing to ask.

$ POST /api/v1/search
$ {
>   "query": "can someone explain why how does what happens to",
>   "limit": 20,
>   "threshold": 0.35
> }

$ POST /api/v1/search
$ {
>   "query": "is it true that am I right in thinking",
>   "limit": 20,
>   "threshold": 0.35
> }

Filter these by product to get topic-specific question banks:

$ POST /api/v1/search
$ {
>   "query": "question about state pension how much when can I",
>   "limit": 15,
>   "threshold": 0.4,
>   "product": "state_pension"
> }

Step 3: Identify myths being spread — especially popular ones

High-engagement comments containing misinformation are particularly important: they represent false beliefs that have been actively validated by hundreds of people hitting like. These are the myths most worth debunking.

$ GET /api/v1/quotes?product=life_insurance&emotional_register=neutral&min_likes=50&limit=20

Read through these for factually incorrect claims that attracted significant engagement. Common examples in personal finance: “pensions are a scam”, “you lose everything if you die before retirement”, “life insurance doesn’t pay out”, “renting is always cheaper than buying”.

Search for specific myths you want to pressure-test:

$ POST /api/v1/search
$ {
>   "query": "pensions don't pay out won't get it back not worth it",
>   "limit": 15,
>   "threshold": 0.35
> }

$ POST /api/v1/search
$ {
>   "query": "life insurance never pays out claim rejected loophole",
>   "limit": 15,
>   "threshold": 0.35
> }

Step 4: Segment misconceptions by life stage

Different life stages have different knowledge gaps. A first-home buyer’s mortgage misconceptions are completely different from a pre-retiree’s pension ones. Filter by life stage to build targeted education for each segment.

What first-home buyers misunderstand:

$ POST /api/v1/search
$ {
>   "query": "didn't know realise understand about buying first home mortgage deposit",
>   "life_stage": "first_home_buyer",
>   "limit": 15,
>   "threshold": 0.4
> }

What pre-retirees get wrong about their options:

$ POST /api/v1/search
$ {
>   "query": "thought I had to annuity drawdown didn't know options at retirement",
>   "life_stage": "pre_retirement",
>   "limit": 15,
>   "threshold": 0.4
> }

What younger people misunderstand about why to start early:

$ POST /api/v1/search
$ {
>   "query": "wish I had started earlier compound interest pension young",
>   "life_stage": "young_pre_property",
>   "limit": 15,
>   "threshold": 0.4
> }

Step 5: Generate a content brief with the AI analyst

Once you’ve gathered the raw material, use the AI analyst to synthesise it into a structured content strategy.

Full misconception audit:

$ POST /api/chat
$ {
>   "messages": [
>     {
>       "role": "user",
>       "content": "Audit the corpus for misconceptions and knowledge gaps around pension saving. Identify the top eight most common misunderstandings — for each one, describe the misconception as consumers express it, explain why it matters, quote the most resonant example from the corpus, and suggest a content format and key message to address it. Focus on misconceptions with evidence across multiple comments."
>     }
>   ]
> }

Life stage specific education plan:

$ POST /api/chat
$ {
>   "messages": [
>     {
>       "role": "user",
>       "content": "Build an educational content plan for pre-retirement consumers aged 55-65. Based on the corpus, what are the five topics they most need educating on? For each topic: describe the knowledge gap, give two verbatim examples of the confusion, and recommend a content format, title, and the single most important thing the content needs to communicate."
>     }
>   ]
> }

Content calendar brief:

$ POST /api/chat
$ {
>   "messages": [
>     {
>       "role": "user",
>       "content": "Based on the most common questions and misconceptions in the corpus about life insurance, produce a 12-topic educational content calendar. For each topic include: the consumer question or misconception it addresses, a suggested article title written in plain consumer language, the emotional register to adopt, and the life stage it's most relevant to."
>     }
>   ]
> }

Using the MCP server

In Claude Desktop, education content research is particularly well-suited to a conversational workflow:

“Look through the corpus for misconceptions about pension drawdown. What do people commonly get wrong, and what do they wish someone had explained to them earlier?”

“Now find the questions about drawdown that got the most likes — these are the shared confusions people were afraid to ask. List them with the like count.”

“Use those to write five article titles for an educational series, written in the language real people use, not financial jargon.”

Tips

High-liked questions are your best content briefs — a question with 200 likes is a confirmed content gap. Run GET /api/v1/quotes filtered by a question-style semantic search and sort by like_count.
Use anger_cynicism comments as a myth detection signal — cynical comments often contain misconceptions presented as fact (“they’re all scams”, “it never works out”). These are the beliefs your education content most needs to counter.
Match reading level to the confusion — if the corpus shows people confused by basic terminology, the content needs to start further back than you’d assume. Use the consumer language in comments as a ceiling, not a floor.
Reddit data is especially rich for misconceptions — community platforms like UKPF surface raw, unfiltered questions and debate that YouTube comments rarely do. Filter by channel=ukpf to find the most candid expressions of confusion.