Practical AI adoption insight, every week. Subscribe to aiblBRIEF
aibl
  • Practical AI Insights
    • The research
      • Benchmark your AI ROI (Beta)
      • State of UK AI Adoption 2026
      • What our research says
      • AI analysis
      • Leadership Series report
      • Newsletter archive
    • Learn and apply
      • AI explained
      • AI case studies
      • AI build stories
      • AI playbooks
      • AI workshop guides
      • Boardroom findings
      • AI leader interviews
  • aiblLIVE
    • 2026 session library
    • aiblLIVE 2027
  • Find AI enablement partners
    • AI Enablement Directory
      • Search the directory
      • Register your business
    • By Category
      • AI Strategy & Readiness
      • AI Implementation & Deployment
      • AI Training & Capability Building
      • AI Agents & Agentic Workflows
      • AI Governance & Risk
      • Generative AI Applications
      • Data & Analytics
      • Process & Document Automation
  • Partner with aibl
    • For enablement partners
    • For AI platforms and vendors
    • Insights for vendors
    • Market insights reports
  • About
    • About Us
    • AI For business leaders
Subscribe to aiblBRIEF
  • Practical AI Insights
    • The research
      • Benchmark your AI ROI (Beta)
      • State of UK AI Adoption 2026
      • What our research says
      • AI analysis
      • Leadership Series report
      • Newsletter archive
    • Learn and apply
      • AI explained
      • AI case studies
      • AI build stories
      • AI playbooks
      • AI workshop guides
      • Boardroom findings
      • AI leader interviews
  • aiblLIVE
    • 2026 session library
    • aiblLIVE 2027
  • Find AI enablement partners
    • AI Enablement Directory
      • Search the directory
      • Register your business
    • By Category
      • AI Strategy & Readiness
      • AI Implementation & Deployment
      • AI Training & Capability Building
      • AI Agents & Agentic Workflows
      • AI Governance & Risk
      • Generative AI Applications
      • Data & Analytics
      • Process & Document Automation
  • Partner with aibl
    • For enablement partners
    • For AI platforms and vendors
    • Insights for vendors
    • Market insights reports
  • About
    • About Us
    • AI For business leaders
  • Subscribe to aiblBRIEF

AI agent monitoring at Sierra: the agent that improves agents

Downloads | AI case studies

Sierra handles AI agent monitoring with another agent. Its tool, Ghostwriter, analyses, edits and tests customer service agents, and now watches over them from inside customers’ team channels. Rebecca Okin, Solutions Engineer at Sierra, demonstrated it live at aiblLIVE London 2026.

In the demo, Ghostwriter worked out why a broadband agent passed half its calls to humans. It then planned, made and tested a fix in under 10 minutes, with a human approving every change.

Sierra’s Ghostwriter in numbers

  • 4 in 5 changes to agents on Sierra’s platform are now made with Ghostwriter.
  • Under 10 minutes to plan, make and test the fix in the live demo.
  • About 90% of the time, Ghostwriter stays quiet while it scans for problems.
  • 5 out of 175 chats were caught handling one type of question wrongly.

All figures as stated on stage by Rebecca Okin.

How does an AI agent improve another AI agent?

It reads the conversations, forms a hypothesis and proposes a tested change. In the demo, Ghostwriter found most transfers came from customers opting out at once. It suggested a short qualifying question before transfer, edited the agent, then ran simulations with AI users. Nothing went live without approval.

This was only through my approval. It did not just go about and do these things without me asking, and there’s always an approvals process that goes with it.

Rebecca Okin, Solutions Engineer, Sierra

What does proactive AI agent monitoring look like?

It looks like a teammate in Slack or Teams who speaks up only when needed. Ghostwriter learns team goals and who owns what. In real customer channels it caught a misconfigured experiment and traced a traffic spike to an unknown fee change. It also reopened a recurring error in its original thread.

How do you stop an AI monitor becoming noise?

Give it judgment about when to interrupt. Ghostwriter scans constantly but stays quiet about 90% of the time. It learns when each team prefers to hear from it and weighs whether an issue can wait until after the weekend. It also takes feedback, such as a request to message later in the day.

Ghostwriter is observing constantly and it’s making judgment calls on when to raise its hand.

Rebecca Okin, Solutions Engineer, Sierra

What controls apply to an agent that edits agents?

The same rules that bind the agent itself. Ghostwriter only sees the customer’s agent and internal team channels. It follows Sierra’s built-in rules, rules agreed with each client and the company’s own policies. Handling of personal data is configurable per client, and every change needs human sign-off.

Key takeaways

  • Ask why before you tune. Test a hypothesis against real conversations first.
  • Respect customers who want a human. Offer quick help and state the wait, without blocking transfer.
  • Rerun key journeys every week. Simulations catch regressions after each change.
  • Keep a human approval step. Let AI draft changes, and require sign-off.

About the speaker

Sierra logo

Rebecca Okin is a Solutions Engineer at Sierra, which builds conversational AI agents for businesses. She spoke on the Customer track at aiblLIVE London on 20 October 2026.

What’s in the full case study

If your team runs customer service agents, this is a working model for keeping them sharp. The full PDF walks through the live demo step by step.

  • The demo, step by step: from a 50% containment rate to a tested fix.
  • Four real problems Ghostwriter caught in customer channels, and how it routed each one.
  • The weekly regression test it offered to run for a financial services customer.
  • Five actions to take back to your own organisation.

For customer service, CX and contact centre leaders, and teams that build or run AI agents. Five pages, about a seven-minute read.

Keep your AI agents improving

AI agent monitoring at Sierra: the agent that improves agents
aibl
Subscribe to aiblBRIEF

Explore

  • For business leaders
  • For AI platforms and vendors
  • About
  • Contact

AI Insights

  • All articles
  • AI case studies
  • Interviews
  • UK AI adoption research

Events

  • aiblLIVE 2026 session library
  • aiblLIVE 2027

More

  • aiblCONNECT
  • AI Enablement Directory
  • aibl Information Pack for AI’s
  • The State of AI Adoption Key Findings
© 2026 AIBL MEDIA LIMITED (DBA: aibl | AI for business leaders), Company registration number: 16634894 - The organisers of aiblLIVE do not, under any circumstances make attendee data available to third parties for resale. Should you receive an email offering data purporting to be from aiblLIVE – or any other event organised by aibl media limited – please delete the message, block the sender, and do not respond or interact in any way.
  • Privacy Policy