Contact centres frequently announce new AI initiatives, but many of those initiatives quietly stall, not because the technology doesn’t work but because the programme behind it was never designed to succeed.
To find out more, we spoke to Chris Mounce, Product Training & Enablement Specialist at evaluagent, about how the difference between AI pilots that scale and those that fade away isn’t technical capability – it’s strategy.
Video: AI Pilot Programs: Augment Before You Automate (to Keep People in Control)
Watch the video below to hear Chris explain what contact centres can do the get their AI pilots off to the best start and why they need to augment before automation.
With thanks to Chris Mounce, Product Training & Enablement Specialist at evaluagent, for contributing to this video.
This video was originally published in our article ‘Get Your AI Pilot Off to the Best Possible Start’
How to Get AI Pilots Right in the Contact Centre
The most successful contact centres don’t try to do everything at once; instead, they start small, stay focused, and build momentum through measurable impact.
“Every week another contact centre announces an AI initiative, and every week a good few of those stall. Not because the technology failed, but because the programme was never set up to succeed.
Starting small with AI isn’t a compromise. It’s a strategy. And there are three things that separate the pilots that build momentum from the ones that generate reports that no one acts on.”
To get you started, here are three principles that help AI pilots deliver:
1. Start With the Problem, Not the Tool
One of the most common mistakes in AI adoption is choosing the technology first and then searching for a use case to justify it.
This approach almost always leads to weak outcomes, as Chris explains:
“The first is how you define the problem. Most AI pilots fail before they start because teams choose a tool first and then look for a use case to justify it. And that’s the wrong way around.
Contact centres have a genuine advantage here – – you’ve got transcripts, QA data, handling time, outcome records – so before any configuration starts, use that data to define one specific outcome that you want to change.
And what does that problem actually cost today in time, and risk, and in quality? A clear hypothesis is harder to write than it sounds, but it’s what separates a real pilot from a proof of concept that goes nowhere.”
Contact centres already have a powerful advantage: rich operational data, including:
- Interaction transcripts
- Quality assurance (QA) data
- Handling times
- Outcome and resolution records
The key is to use this data to define a single, specific problem before introducing any AI solution, by asking:
- What outcome do we want to improve?
- What is this problem costing today (in time, quality, or risk)?
A clearly defined hypothesis is what separates a meaningful pilot from a proof of concept that never progresses.
2. Augment Before You Automate
The most successful AI pilots don’t start with customer-facing automation; instead, they begin behind the scenes supporting agents and internal teams. For example, this can include:
- Automated interaction scoring
- Post-call summarization
- Real-time knowledge retrieval
These use cases share a critical advantage: a human remains in the loop, and outputs can be reviewed, validated, and corrected before they impact the customer.
“The second principle is augment before you automate. The pilots that scale don’t start with customer-facing AI. Start behind the scenes with agents and quality teams, where the stakes of a wrong answer are lower and the feedback loop is much faster.
Automated interaction scoring, post-call summarization, real-time knowledge retrieval – what these things have in common is simple. A human can review the output before it matters.
And that’s not a limitation, that’s good design. It keeps people in control, makes errors visible and correctable and builds the internal confidence that you need before you go further.”
But this isn’t a limitation, it’s simply good design, as by keeping people in control:
- Errors are visible and fixable
- Trust in the system builds over time
- Feedback loops are faster and more effective
This foundation is essential before moving to higher-risk, customer-facing applications.
3. Measure Quality, Not Just Speed
Many AI pilots focus heavily on efficiency metrics such as reduced handling time, faster wrap-up, and increased volume.
While these are important, they don’t tell the full story, especially as speed improvements can mask deeper issues, such as increased errors, more repeat contacts, and compliance risks.
“Measure both speed and quality and then make a real decision. Too many pilots measure only efficiency, handling time, wrap time, volume.
But speed gains can utterly mask a rise in errors, repeat contacts, or compliance failures that only surface later. So set guardrails before you go live, and define what would make you stop, and then commit to a genuine go/no-go decision at the end of a fixed window.”
That’s why it’s critical to measure both speed and quality from the outset, and before launching a pilot:
- Define clear guardrails for acceptable performance
- Identify what would trigger a stop or rollback
- Commit to making a genuine go/no-go decision at the end of a fixed evaluation period
This ensures that success is measured holistically, not just by how fast things move, but by how well they work.
From Pilot to Progress
AI success in the contact centre isn’t about ambition, it’s about execution.
Starting small isn’t a compromise; it’s a strategy for building confidence, capability, and measurable results.
By focusing on the right problem, keeping humans in the loop, and measuring what truly matters, contact centres can turn AI pilots into scalable, impactful programmes, rather than experiments that never move beyond the testing phase.
Author: Robyn Coppell
Reviewed by: Jo Robinson
Published On: 19th Jun 2026
Read more about - Video, Artificial Intelligence (AI), Chris Mounce, Employee Engagement, EvaluAgent, Leadership, Technology Enablement Strategy, Technology Roadmap, Videos
