The Repeated Question Audit: 30-Min Checklist for UK SMEs

---

Repeated questions are not a people problem. They are a system problem.

In a 10–100 person SME, the same handful of questions quietly drives dozens of daily interruptions: “Where’s the latest template?”, “Who approves this?”, “How do we handle this edge case?”. Each one only takes a minute or two, but the context switching destroys deep work. Across a London‑based SME, that easily adds up to a day a week of lost capacity in operations and leadership (rough estimate based on our client assessments).

Most owners respond by telling people to "check the shared drive" or "read the SOP". It rarely works. The underlying issue is that the answers are hard to find, inconsistent, or live in one person’s head.

This is where a repeated question audit helps. Spend 30 minutes mapping the patterns and you get:

• A concrete knowledge gaps audit
• A shortlist of internal FAQ automation candidates
• A realistic scope for an AI internal assistant that actually answers questions instead of becoming yet another unused tool

Below is the exact checklist we use at SIMARA AI with UK SMEs when we want to turn repeated questions into a roadmap for capacity gains.

---

1. List your top 10 repeated questions (not the ones you wish people asked)

What it is
Capture the 10 questions your team has asked most often in the last month. Focus on real interruptions, not theoretical “important” questions.

Why it matters
If you start from processes or org charts, you will miss where attention actually goes. Repeated questions are a live signal of where your operating model is unclear. In our AI Readiness Scorecard, this is the practical test for Process Clarity – if people keep asking, the process is not truly documented, regardless of how many SOPs exist.

Actionable step

• Block 10 minutes with 3–5 people who get pinged constantly: ops manager, team leaders, senior specialists.
• Ask: “What do you get asked at least once a day that you could answer in your sleep?”
• Write each question on a single line in a shared doc or whiteboard.
• If you reach more than 10 easily, you already have a material SME capacity leak.

---

2. For each question, identify who is usually asked

What it is
Map the "answer owner" for each repeated question – the person or small group who gets interrupted.

Why it matters
If the same names appear repeatedly, you have key‑person bottlenecks and concentration risk. According to UK SME data, staff turnover in admin and co‑ordination roles is roughly 15–20% annually in London [FSB, 2024 – indicative]. If those people hold critical know‑how, every resignation resets your operational memory.

Actionable step

• Add a simple column next to each question: “Who do you normally ask?”
• Tally how many questions each name appears on.
• If any individual appears on more than 3 of the 10 questions, mark them as a knowledge bottleneck.

---

3. Mark how often each question is asked per week

What it is
Estimate the weekly frequency for each question type: daily, several times a week, weekly, monthly.

Why it matters
Our Process Priority Matrix says: daily, high‑impact items get automated first. The same logic applies to knowledge. A question asked daily by three people is a better candidate for internal FAQ automation than a complex question asked monthly.

Actionable step

• For each question, ask the answer owner: “Roughly how many times a week do you get this?”
• Categorise:
  • 1 = Monthly or less
  • 2 = Weekly
  • 3 = Several times a week
  • 4 = Daily
• Circle anything scored 3 or 4 – these are your primary repeated questions.

---

4. Capture where the current answer actually lives (if anywhere)

What it is
For each question, note the real source of truth today: in someone’s head, in a PDF, buried in Teams, in a CRM note, in an SOP, or nowhere.

Why it matters
Most SMEs assume the answer “is in the system”. When we run audits, we typically find:

• 30–50% of answers live only in people’s heads (rough estimate from SIMARA assessments)
• 20–30% exist in out‑of‑date documents
• The rest are scattered across emails, chat, and legacy files

This directly hits the Data Accessibility and Process Clarity dimensions of our AI Readiness Scorecard.

Actionable step

• Add a column: “Where is the answer today?”
• Use strict options: Head only / Scattered messages / Old doc / System field / Up‑to‑date doc.
• Highlight anything that is Head only or Scattered messages as a knowledge gap, even if people “sort of know” the answer.

---

5. Estimate the cost of each question type in minutes and £

What it is
Attach a rough time and cost to each repeated question – including the interruption cost.

Why it matters
You cannot justify an AI internal assistant or internal FAQ automation without numbers. Using our ROI calculator template, a 2‑minute question that happens 20 times a week at a blended cost of £35/hour (London operations staff fully loaded) quickly becomes meaningful.

Actionable step

• For each question, estimate:

  • Time to ask + answer (include context switching): 2–10 minutes
  • People involved: asker + responder

• Use a simple formula:

  Weekly minutes lost ≈ frequency × minutes × number of people involved

• Convert to cost using a rough hourly rate (e.g. £30–£50/hour for many London knowledge workers).

• Flag any question costing more than £150/month (rough threshold) as high‑impact.

---

6. Classify the question type: “how do I…?”, “where is…?”, or “what’s our rule on…?”

What it is
Group each repeated question into one of three patterns:

• How do I…? (process/how‑to)
• Where is…? (location of information/assets)
• What’s our rule on…? (policy/decision criteria)

Why it matters
Each type lends itself to different automation patterns:

• "How do I…?" → step‑by‑step SOPs, short how‑to videos, AI assistants that retrieve procedural steps.
• "Where is…?" → better search, consistent naming, index pages, tools like Notion or Confluence.
• "What’s our rule on…?" → clear policies, decision trees, governance rails.

This mirrors how we design knowledge flows in our internal wiki work.

Actionable step

• Add a “Question type” column with the three options above.
• Count how many fall into each. If more than 50% are "Where is…?", you have a navigation and indexing problem. If "What’s our rule on…?" dominates, you have a governance and decision clarity gap.

---

7. Score the change risk if the answer is wrong or outdated

What it is
Assess what happens if the answer given is incorrect or based on old information.

Why it matters
Not all repeated questions are equal. A wrong answer on “how to reset my password” is annoying. A wrong answer on “how we handle client refunds” can cost thousands or breach UK consumer law.

Actionable step

• For each question, score impact of a wrong answer on a 1–5 scale:
  • 1 = Mild frustration
  • 3 = Delays, internal rework
  • 5 = Direct financial, legal or reputational risk
• Prioritise FAQs and automation for anything scored 4 or 5. These belong in a governed, reviewed knowledge base, not just someone’s memory.

---

8. Check if the answer is standard (always the same) or conditional (depends on context)

What it is
Decide whether each question has a single, stable answer or if it changes based on client, product, amount, date, etc.

Why it matters
Standard answers are well suited to static internal FAQ pages, pinned posts, or simple chatbots. Conditional answers often justify an AI internal assistant that can interpret context from your CRM, ticketing system, or finance tools.

Actionable step

• Add a column: Standard / Conditional.
• Mark as Conditional if you hear phrases like "it depends", "if it's over £X", or "for that client".
• Conditional, high‑frequency questions are prime candidates for AI‑augmented workflows (e.g. pulling data from Xero or HubSpot to answer "what’s the current status for this client?").

---

9. Map where the question is usually asked (email, Teams/Slack, WhatsApp, in person)

What it is
Identify the channel where repeated questions show up: internal chat, email, ticketing, phone, in person.

Why it matters
Your AI internal assistant or internal FAQ automation needs to live where the questions already are. For many SMEs, that is Microsoft Teams or email, not a new portal. Tools like Microsoft Copilot or Slack’s AI features only create value if your knowledge is accessible from those channels.

Actionable step

• For each question, log the typical channel.
• Rank channels by volume of repeated questions.
• Decide a primary “home” for self‑serve answers (e.g. a Teams app backed by SharePoint + AI, or a Slack bot backed by Confluence).

---

10. Note whether the answer is already documented somewhere, even if poorly

What it is
Check if a version of the answer already exists in any document: SOP, wiki, onboarding pack, PDF, email thread.

Why it matters
Starting from scratch is slow. Most SMEs have 60–80% of the necessary content scattered across archives. Part of our methodology is to use what already exists and normalise it into an AI‑ready wiki, as we outline in our playbook on turning tribal knowledge into an internal wiki.

Actionable step

• Add a simple flag: Documented (yes/no).
• For "yes", keep a link or path – even if it is old. This becomes input for your knowledge consolidation phase.
• For "no" on high‑risk or high‑frequency questions, mark these as top content creation priorities.

---

11. Identify owner and review cadence for each answer

What it is
Assign a named owner responsible for keeping each answer correct and decide how often it should be reviewed.

Why it matters
An FAQ without ownership rots. In our AI‑ready knowledge work, version control and ownership are non‑negotiable. This also matters for GDPR‑related topics, where the ICO expects clear accountability for data handling guidance.

Actionable step

• Add two columns: “Content owner” and “Review frequency” (quarterly, bi‑annually, annually).
• For policy or compliance‑related questions, default to quarterly reviews.
• Do not assign more than 10–15 questions to any single owner initially – or nothing will actually be reviewed.

---

12. Decide the best answer format: text, screenshot, short video, or decision tree

What it is
Select the format that will make the answer quickest to consume and least likely to be misunderstood.

Why it matters
Wall‑of‑text SOPs are rarely read. Many "How do I…?" questions are best handled with a 60‑second Loom‑style video or a simple screenshot with arrows. "What’s our rule on…?" questions often need a clear decision tree.

Actionable step

• For each question, label preferred format:
  • Short text (≤200 words)
  • Step‑by‑step checklist
  • Screenshot/diagram
  • 1–2 minute screen‑recording
  • Decision tree flow
• Plan to build content in that format first – you can always add a text transcript later for AI indexing.

---

13. Tag which questions are suitable for static FAQs vs an AI internal assistant

What it is
Split your list into two buckets:

• Static FAQ → single, simple answer, low variation
• AI internal assistant → needs context from systems or flexible natural‑language answers

Why it matters
You do not need AI everywhere. We often see more ROI from 10 well‑designed static FAQs than from a rushed chatbot. AI shines when:

• The answer needs live data ("What’s the client’s current payment status?")
• The same concept is asked in many different ways
• The question requires combining several sources (CRM + finance + project tool)

Actionable step

• Mark each question as FAQ or AI‑assistant.
• If you end up with more than ~30% in the AI‑assistant bucket, that is a sign your processes and data are fragmented – AI will need a careful integration design.

---

14. Check your data accessibility for AI‑assistant candidates

What it is
For questions you tagged as AI‑assistant, assess whether the underlying data is machine‑readable and reachable via API or export.

Why it matters
An AI internal assistant is only as good as the data it can see. If crucial answers live in PDFs, scanned forms, or someone’s inbox, you need a data accessibility step first. This directly ties to the Data Accessibility dimension of our AI Readiness Scorecard.

Actionable step

• For each AI‑assistant question, list the systems involved (Xero, HubSpot, Microsoft 365, Trello, etc.).
• For each system, mark:
  • API / Integrations available
  • Export only
  • Manual only
• If most relevant systems are Manual only, delay AI‑assistant ambitions and start with document processing or light integration (e.g. using Power Automate, Zapier or Make) to get data out of silos.

---

15. Apply a mini ROI check to 3–5 top candidates

What it is
Run a rough ROI calculation on the highest‑frequency, highest‑cost repeated questions.

Why it matters
Without numbers, knowledge projects get deprioritised. Our typical SME ROI patterns show that even modest automations can return investment in 6–12 months. The same is true for answering repeated internal queries.

Actionable step

• Pick 3–5 question types with:
  • High frequency (3–4 on your earlier scale)
  • High cost (≥£150/month each)
• Estimate:
  • Weekly hours lost = weekly minutes ÷ 60
  • Monthly savings if 70% is automated (rough coverage for good internal FAQ + AI assistant)
• Compare against an estimated implementation cost (often £5,000–£15,000 for a first internal assistant workflow based on our projects). If payback is under 12–18 months, they qualify as phase‑one candidates.

---

16. Spot the onboarding hotspots: which questions new starters ask repeatedly

What it is
Identify which of your repeated questions are disproportionately asked by new hires in their first 90 days.

Why it matters
Cutting ramp time in half often delivers more value than small efficiency wins across the board. We discuss this in more depth in our guide to AI‑supported onboarding and cross‑training. If new starters ask the same logistical and process questions, you have a clear case for structured onboarding packs plus an AI internal assistant tuned for “day‑1 to day‑90” queries.

Actionable step

• Add a column: Frequently asked by new starters? (yes/no).
• Ask your last 2–3 hires which questions they kept asking.
• Elevate any questions marked "yes" into a dedicated Onboarding FAQ set.

---

17. Identify cross‑team confusion (the same question asked differently by sales, ops, and finance)

What it is
Look for repeated questions that appear in multiple teams with slightly different wording or assumptions.

Why it matters
These are not just knowledge gaps – they are alignment gaps. For example, different interpretations of "qualified lead" or "priority client" cause reporting errors and misaligned expectations. We see this repeatedly in our work on cutting decision cycles from 30 days to 3 days.

Actionable step

• For each question, note which teams ask it.
• If a question spans two or more teams, mark it as cross‑functional.
• Prioritise these for leadership review and clear, shared definitions before you automate; otherwise your AI assistant will simply repeat conflicting logic.

---

18. Decide what should not be automated (for now)

What it is
Explicitly ring‑fence questions that must stay human‑mediated due to sensitivity, nuance, or cultural impact.

Why it matters
Not all repeated questions should be answered by a bot or static FAQ. Performance, pay, grievances, and one‑off client escalations often need a human ear. Knowing where not to use AI is part of building trust.

Actionable step

• Tag each question as one of:
  • Safe to automate
  • Assist only (AI draft, human final)
  • Human only
• Exclude "Human only" items from your initial automation scope, but still document guidance for managers.

---

19. Group your top 10–20 questions into 3–5 internal FAQ themes

What it is
Cluster related questions into themes such as "Client delivery", "Finance & billing", "HR basics", "Tools & access", "Policies & compliance".

Why it matters
People search by theme, not by perfect wording. Grouping helps you design intuitive navigation in a wiki or internal portal, and improves recall for staff. It also aligns with how AI retrieval systems (like those behind tools such as Microsoft Copilot or Notion AI) benefit from well‑structured, tagged content.

Actionable step

• Create 3–5 buckets and drag questions into them.
• Ensure each bucket is small enough to be surfaced as a single FAQ page or section (ideally 5–10 questions each).
• These become the backbone of your internal FAQ automation roadmap.

---

20. Pick your first 3 automation moves and owners

What it is
Convert the audit into a concrete, small‑scale plan with owners and dates.

Why it matters
Without a clear first move, audits die in shared drives. Our three‑phase implementation model always ends phase one with a specific pilot. Treat repeated questions the same way.

Actionable step

• From your prioritised list, select:
  • 1–2 static FAQ bundles (e.g. “Client delivery basics”)
  • 1 AI‑assistant candidate (e.g. "What’s our rule when X happens?")
• For each, define:
  • Owner
  • Target go‑live date (within 4–8 weeks)
  • Success metric (e.g. “Reduce pings to ops channel about delivery steps by 50% within 2 months”).
• Consider a lightweight pilot using existing tools: for example, a SharePoint FAQ surfaced via Microsoft Teams with a simple Q&A bot, or a Slack bot backed by a Notion knowledge base.

---

Final review / summary

If you have followed this checklist, you now have:

• A concrete list of your top repeated questions and who they interrupt
• A quantified view of where your SME capacity leak actually sits
• A first pass knowledge gaps audit tied to real costs and risks
• A short, ranked list of FAQ and AI‑assistant opportunities that are technically feasible

The temptation is to jump straight to buying an "AI assistant". Our advice is stricter: fix the top 10–20 questions, prove you can maintain accurate answers, then layer AI on top so staff can ask in natural language and get the same governed answer every time.

Run this repeated question audit quarterly. It is a 30‑minute discipline that exposes where your business model is frictional – and where AI‑supported knowledge management can safely remove that friction without adding yet another system.

Ready to turn this audit into a concrete automation roadmap? → Book a consultation

---

Sources & further reading

• Federation of Small Businesses (FSB), "UK Small Business Statistics" (2024 overview).
• ICO, "Guide to the UK General Data Protection Regulation (UK GDPR)" – practical guidance for UK organisations handling personal data.
• Microsoft, "Microsoft 365 Copilot for business" – documentation on integrating AI with Teams and SharePoint.
• Notion, "Using AI as your team’s knowledge assistant" – examples of internal Q&A over a shared wiki.

---

For most SMEs, quarterly is enough. Run it more frequently if you are growing fast, changing services, or onboarding several people at once. The aim is to keep your internal FAQs and AI internal assistant aligned with how people actually work, not how the org chart says they work.

Do we need an AI internal assistant from day one?

No. Start with static, well‑structured FAQs and playbooks for your top repeated questions. Once you have 50–100 high‑quality answers in one place, layering an AI assistant on top becomes far more effective and lower risk. We generally see better ROI when AI is phase two, not phase one.

Which tools work best for internal FAQ automation in a small UK SME?

Most SMEs already have suitable platforms: Microsoft 365 (SharePoint + Teams), Google Workspace (Drive + Chat), or tools like Notion. Products such as Confluence are common in tech‑leaning teams. The choice matters less than being consistent and making sure the knowledge is searchable and permissioned correctly.

How do we avoid exposing sensitive data when using AI internally?

Keep personal and sensitive data in your core systems, not in the training data for your assistant. Use tools that support data residency and enterprise controls, ensure data processing agreements are in place, and only allow the AI to retrieve from existing documents and systems under your usual access rules. Follow ICO guidance on lawful bases and data minimisation.

What if our repeated questions are all about messy exceptions rather than simple rules?

That usually indicates underlying process design issues, not just knowledge gaps. Use the audit to identify patterns in exceptions and consider whether policies or product rules need simplifying. An AI assistant can help surface past decisions, but it should not be used to paper over fundamentally inconsistent rules.

---

Find 3 hidden efficiency gains in 30 minutes → Book a consultation

---