Case study: Agent Assist

Agents got the full picture. Customers felt the difference.

Support agents juggle long ticket histories, processes that change often and high volumes, every day. We built Digital Agent Assist with them so every ticket starts with the full picture, straight from Zendesk. Agents reply with confidence and spend their time on the customer, not the backstory.

+10.3pts

CSAT against the pre-launch baseline, four weeks after go-live

50k+

Tickets analysed in the first five weeks

180+

Agents using it weekly, adopted from day one

10%

Of the market price for equivalent solutions, with the Zendesk integration built at no cost

How it works

One ticket number, the whole story

The agent enters a ticket number. Digital Agent Assist reads the full thread in Zendesk and checks it against the current process, so the agent can go straight to the reply. Interactions per hour went up as a result.

SummaryWhere the ticket stands, in a few lines.
Next stepsWhat to do, based on the current process.
WarningsRisks flagged up front, such as legal action.

Ticket

#48213

Summarise

Zendesk 9 messages 3 previous agents

Possible legal action

Customer says they have spoken to a solicitor (message 7). Follow the legal escalation process before replying.

Summary

Customer was charged twice in August. A refund was promised on 19 Aug but has not reached their account. This is their third contact about the same issue and they are now asking for compensation.

Next steps

Check the refund status in billing and note the reference.
Raise to the complaints team under the legal escalation process.
Reply confirming the escalation and when they will hear back.

How we built it

Started on the floor, earned every stage

We did not arrive with a product. We spent time with the operation, watched how tickets were really worked and found where the time went. We built it around those pain points, rather than retrofitting an existing tool onto the floor.

What we saw on the floorWhat we built in

Processes kept changing

Every ticket is checked against the current version of the process, so nobody has to keep every change in their head.

Constant context switching

Every ticket is a new context. Digital Agent Assist gets the agent up to speed in one step, instead of piecing it together from the thread.

High volumes, every day

Digital Agent Assist does the reading, so agents can put their time and judgement into the reply.

Tickets changed hands

Whoever picks a ticket up next gets the story so far before they reply, without reading the whole thread.

Then we widened it, one stage at a time

Each stage ran on live tickets, and each one had to earn the next.

A team leader
Set the standard for a good brief, then became the super user who helped everyone adopt it
Team leader and their agents
Feedback from the people answering tickets all day
40 agents and team leaders
Proved it held up at volume
Everyone
180+ weekly users, from day one

Consistent use, week after week

Use didn't spike at launch and fade. Weekly users held above 160 from the first week and grew to 180+, with more than 10,000 tickets analysed every week after launch.

17 Aug
24 Aug
31 Aug
7 Sep
14 Sep

Weekly active users, from go-live

At every stage, the client's CI team checked accuracy. Their continuous improvement team analysed its output before each step up, so the next group only got it once it was right.

Security and compliance

Customer data stays where it belongs

Ticket data not stored

Digital Agent Assist reads the ticket when the agent asks for it. Ticket content is not kept.

All processed in‑region

Every request is processed within the region, so data never leaves the jurisdiction.

No AI models trained

No customer or ticket data is used to train any AI model.

Audit-ready too. Every answer carries the date it was generated, knowledge keeps a full version history, and usage is logged in detail.

Want Digital Agent Assist
for your queue?

We start on your floor and build around what your agents need.

Book a demo