When a customer asks something your website doesn’t cover, one of two things usually happens. They give up, or they ring your team. Either way, nobody writes the question down, and the next person asks it again tomorrow.

A Digital Assistant can do better than that. Not because it knows everything, but because it can tell you exactly what it didn’t know.

The knowledge gap register

Every Sum Vivas Digital Assistant reports to Cortex, our service intelligence layer. When the Assistant can’t answer a question from your approved knowledge, Cortex records it in the knowledge gap register, along with:

  • how many times it has been asked;
  • which channels it came from;
  • the conversations behind it.

Your team writes the answer once, on the same page, and from then on the Assistant answers it on every channel.

That turns “the AI got it wrong” into a weekly list of small, specific fixes, ordered by how much they matter.

Every conversation, reviewed

The gap register is one part of a broader review. After each conversation ends, Cortex assesses it and records:

  • what the person wanted;
  • whether it was resolved;
  • how they felt, and whether that improved;
  • the quality of the answers;
  • why it failed, if it did.

You can define your own scorecard too, so the review records what matters to your organisation, such as “booking made” or “access request raised”.

Dashboards you can click through

Cortex’s charts open the conversations behind them. If resolution dips on Saturday mornings, you can see the exact conversations that caused it. If one topic starts trending, you can read what people are asking.

A report you can share

Cortex produces a one-click Word or PDF report with a plain-English summary, so the people who never log in still see how the service is performing.

Estimates labelled as estimates

Cortex can estimate the staff time conversations would have taken. Where a figure is an estimate, it’s labelled as one, and you can see how it’s calculated. We’re developing extended value reporting, such as hours avoided, equivalent human cost and ongoing return, with customer-configurable assumptions. We will label it as in development until it’s released.

Why it matters

An assistant that fails quietly gets worse in the eyes of your customers. One that shows you its gaps gets better every week, because your team can see exactly where to improve.

Talk to us to see Cortex with a demo account.