Running the KCS loop in a contact centre

What is KCS?

KCS, short for Knowledge-Centered Service, is a knowledge management methodology developed and maintained by the Consortium for Service Innovation, a non-profit group of support organisations. Its central idea is that knowledge is a by-product of solving problems. The people who answer customers capture what they learn while they work, reuse what others have captured, and improve it every time they use it. Knowledge is never "finished"; it is continuously refined by demand.

The methodology describes two loops. The solve loop is what happens in each interaction. The evolve loop is what the organisation does over time to keep the system healthy: content health, process integration, performance assessment, and leadership and communication. Most of the friction in contact centres sits where the two meet.

This guide doesn't restate the methodology. The Consortium publishes the practices guide and adoption material in full. It looks at what happens when you run the loop in a busy contact centre, and how conversation evidence can carry the parts people struggle to do at volume. Terms such as coverage rate and repeat contact are defined in the knowledge management glossary.

What are capture, structure, reuse and improve?

Each practice is sound. The problem is that all four assume an agent with time, a habit of searching, and a reason to write things down. Contact centres are designed to minimise all three.

Why is KCS harder in a contact centre?

KCS grew up in technical support, where cases are longer, engineers already write case notes, and one resolution often serves many future customers. A contact centre is different in ways that hit the solve loop directly:

• Volume and pace. Short interactions, tight handle time targets and back-to-back calls leave little room for writing.
• Voice. Much of the hardest demand arrives by phone, where nothing is typed by default. The customer's words exist only in the recording or transcript.
• Turnover. New agents are the heaviest knowledge users and the least able to judge whether an article is right. Experienced agents are the best judges and the least likely to search.
• Many small questions. A large share of demand is short, repetitive questions on billing, policy and process. Each one feels too small to write up, and together they are most of the cost.
• Several channels. The same question arrives by voice, chat, email and messaging, handled by different teams, often with different articles.

Where does the loop break in practice?

In most contact centres the loop is running, slowly, in a way that hides its own gaps. Four breaks are common.

Capture depends on someone noticing. An article gets written when an agent or team leader decides a question deserves one. Questions handled by improvisation, transfer or a callback rarely trigger that decision, because the agent who struggled has already moved on to the next call.

Reuse is measured by searching. Link rate and article views only count interactions where an agent searched and linked. Everything answered from memory is invisible, so the metrics flatter the topics agents search for and miss the rest. This is the same blind spot that affects search-log gap reports, covered in failed searches vs knowledge gaps.

Improve stalls in a queue. Flags and feedback arrive faster than a small knowledge team can process them, without a sense of which matter most. The oldest flag gets worked, not the most expensive one.

Content health falls behind. Under pressure, agents create a new article instead of finding the old one. Duplicates appear, then drift apart until they contradict each other. See duplicate and contradicting articles for how that happens and how to unwind it.

How does conversation evidence fix capture?

Treat the conversation as the capture. Every interaction already records the customer's question in their own words, what the agent tried, and whether it worked. That is the raw material KCS asks agents to write down. Reading it automatically moves capture off the agent's desk without losing the "customer's words" principle.

In practice, conversation-based gap detection does this for each interaction:

1. Extracts what the customer wanted, as one or more plain-language intents.
2. Places each intent in a taxonomy built around how customers ask.
3. Checks whether an existing article answers it, partly answers it, or doesn't.
4. Groups unanswered intents into gaps and ranks them by impact: volume, weighted by signals of difficulty such as negative sentiment, unresolved outcomes and callbacks.

The knowledge team then works a ranked list instead of a flag queue. Each gap carries the conversations behind it, so the reviewer sees the exact phrasing customers use and the point where agents got stuck. Where an article already exists, the right move is to link or update it, which keeps to the KCS habit of improving what is there before creating something new.

Drafting can use the same evidence. In Knowledge Quest, Knowledge Copilot drafts a new article or an update from the contributing conversations and your authoritative documents, following your template, and records where each draft came from. A subject-matter expert reviews it before anything is published. Structure comes from the template; judgement stays with people.

How should you measure reuse and improvement?

Measure from the demand side. Agent search behaviour is a weak proxy for whether knowledge served the customer. Three measures work better in a contact centre:

• Coverage. The share of customer intents an article answers. Track the change from one period to the next and which topics drag it down. Chasing 100% is pointless; the trend is the signal.
• Gaps closed. How many identified gaps were resolved with a new or updated article in the period. A quiet week at zero is fine. A quarter at zero means authoring isn't keeping up.
• Outcomes on fixed topics. Transfers, callbacks and repeat contacts on a topic before and after its article changed. If they don't move, the fix didn't work, however many views it gets. Measuring knowledge article impact covers how to run that comparison fairly.

These measures don't depend on agents searching or linking, so they are honest about voice interactions and experienced agents. They also turn the improve practice into something you can show a contact centre manager: this topic drove this many repeat contacts last month, we changed this article, and they fell.

How does this support the evolve loop?

The evolve loop is where KCS programmes usually lose momentum, because content health and performance assessment are ongoing work with no natural trigger. Conversation evidence gives them one.

• Content health becomes a scheduled set of checks, not an annual audit. Readability, validity against your source documents, brand alignment, complexity, duplicates and contradictions can all be checked continuously. Our guide to knowledge base content health explains each check.
• Process integration gets easier when capture doesn't add steps to the agent's workflow.
• Performance assessment can use coverage and outcomes per topic, not individual link counts that reward volume over quality.
• Leadership and communication have a weekly story to tell: what customers asked that we couldn't answer, what we fixed, and what it changed.

Emerging topics belong here too. New products, outages and policy changes create questions that didn't exist last month. Spotting a topic while it is still growing lets the team write the article before it becomes the busiest queue in the building.

How do you start without a big programme?

1. Pick one knowledge area. Choose one with high volume and a clear owner, such as billing or claims.
2. Read a month of conversations. Across all channels, including voice. Let the gaps and their impact set the first priorities.
3. Agree a weekly rhythm. One hour a week to work the ranked list: link, update, create or dismiss with a reason.
4. Keep subject-matter review. Every draft is approved by someone who knows the answer is right.
5. Report coverage and outcomes monthly. Show what changed on the topics you fixed.

None of this needs a new knowledge base. Knowledge Quest reads conversations from your contact centre platform, finds and ranks the gaps, drafts fixes with their evidence, and sends approved articles to the knowledge base you already run. It helps the KCS loop keep pace with contact centre volume, and the method stays yours. For the analysis method in full, see knowledge gap analysis.