Genesys Cloud - Main

Hi Sanderley,

We've been using Agent Copilot primarily with custom conversation summaries, and one of the biggest lessons for us has been that the prompt makes a huge difference.

Initially, using the standard summary prompt, we didn't see a lot of buy-in from our agents. The summaries weren't consistently capturing the information they found most useful, so adoption was fairly limited.

Once we invested time in developing a custom summary prompt tailored to our environment and refined it based on agent feedback, the response changed significantly. Agents consistently told us the summaries were much more useful than the standard prompt, and we've seen much stronger day-to-day adoption as a result.

My biggest recommendation would be not to underestimate the value of prompt engineering. Spend time refining the summary based on agent feedback rather than relying solely on the default prompt,it made the biggest difference for us.

I also shared some of our approach to building a custom summary prompt in this Community discussion if it's helpful:
https://community.genesys.com/discussion/anyone-using-co-pilot

What a valuable feedback and contribution @Phaneendra Avatapalli! I[m sure this will be on top of mind for Sanderley´s team. As CSM of the account I´ll also be looking forward to work with him and extended teams on that approach. 

Hi Sanderley,

We've been using Agent Copilot primarily with custom conversation summaries, and one of the biggest lessons for us has been that the prompt makes a huge difference.

Initially, using the standard summary prompt, we didn't see a lot of buy-in from our agents. The summaries weren't consistently capturing the information they found most useful, so adoption was fairly limited.

Once we invested time in developing a custom summary prompt tailored to our environment and refined it based on agent feedback, the response changed significantly. Agents consistently told us the summaries were much more useful than the standard prompt, and we've seen much stronger day-to-day adoption as a result.

My biggest recommendation would be not to underestimate the value of prompt engineering. Spend time refining the summary based on agent feedback rather than relying solely on the default prompt,it made the biggest difference for us.

I also shared some of our approach to building a custom summary prompt in this Community discussion if it's helpful:
https://community.genesys.com/discussion/anyone-using-co-pilot

Thank you, Alexandre! I really appreciate your kind words. We're rolling out Agent Copilot gradually across our queues, refining the custom summary prompt based on the context and requirements of each queue. So far we've completed four queues, with a few more still to go. The agent feedback has been invaluable throughout the process, and we're looking forward to continuing to refine and improve the experience.

What a valuable feedback and contribution @Phaneendra Avatapalli! I[m sure this will be on top of mind for Sanderley´s team. As CSM of the account I´ll also be looking forward to work with him and extended teams on that approach. 

For us:

• What business outcomes have you observed? -- The most important outcome so far has been how it necessitated a review, overhaul, and commitment to our knowledge base.  Our CoPilot pilots forced us to start working on the foundation that will support all of our future AI projects.  We've also seen a marked increase in engagement from our front-line teammates along the way -- anxious to be part of the solution (1. for their own career development and 2. not to feel like they're going to be replaced by it)
• Which KPIs improved the most? - AHT for newhires, specifically, our onboarding and training has fundamentally changed as well, and participation in the pilot groups gave us an avenue to promote career development --- which I believe has helped frontline attrition, but it's too soon to draw conclusions there
• Has it helped improve CSAT, FCR, or Quality Scores? - no attributable impact, though logically it should contribute to FCR which raises everything else...just don't have anything concrete
• Which Copilot capabilities (call summaries, suggested responses, knowledge surfacing, wrap-up assistance, etc.) have delivered the most value? -- the auto-summarization has worked really well almost immediately once implemented and continues to help free up resources.  Being able to customize/tailor the summaries is a key feature.  Knowledge surfacing has been a struggle for us, especially at the beginning when it became obvious that we hadn't done a good job of maintaining/structuring our knowledge base over the past several years.  
• Were there any challenges during implementation or user adoption? -- Early trepidation re: the usual fears of AI replacing their jobs -- once we established trust and outlined how we were planning on using the production/AHT gains.  The other obstacle we faced early on was that we were really adding on to their work/AHTs each call with our pilot groups because we put an enhanced feedback loop in place to supplement the feedback system within GC (which is too limited).  That had people concerned about if it would "count against them" and whatnot.  You may also have to handle people who are fundamentally opposed to anything AI because of data centers being built in their area.  That's a trickier issue to handle and I think we're all struggling with a proper response.  
• If you could start over, what would you do differently? - I'd start sooner.  The sooner you begin projects like these with a pilot group the sooner you'll start to figure out what supporting system/services you'll need and which of those you'll need to improve/address

------------------------------
Chad Blanton
Director
------------------------------

For us:

• What business outcomes have you observed? -- The most important outcome so far has been how it necessitated a review, overhaul, and commitment to our knowledge base.  Our CoPilot pilots forced us to start working on the foundation that will support all of our future AI projects.  We've also seen a marked increase in engagement from our front-line teammates along the way -- anxious to be part of the solution (1. for their own career development and 2. not to feel like they're going to be replaced by it)
• Which KPIs improved the most? - AHT for newhires, specifically, our onboarding and training has fundamentally changed as well, and participation in the pilot groups gave us an avenue to promote career development --- which I believe has helped frontline attrition, but it's too soon to draw conclusions there
• Has it helped improve CSAT, FCR, or Quality Scores? - no attributable impact, though logically it should contribute to FCR which raises everything else...just don't have anything concrete
• Which Copilot capabilities (call summaries, suggested responses, knowledge surfacing, wrap-up assistance, etc.) have delivered the most value? -- the auto-summarization has worked really well almost immediately once implemented and continues to help free up resources.  Being able to customize/tailor the summaries is a key feature.  Knowledge surfacing has been a struggle for us, especially at the beginning when it became obvious that we hadn't done a good job of maintaining/structuring our knowledge base over the past several years.  
• Were there any challenges during implementation or user adoption? -- Early trepidation re: the usual fears of AI replacing their jobs -- once we established trust and outlined how we were planning on using the production/AHT gains.  The other obstacle we faced early on was that we were really adding on to their work/AHTs each call with our pilot groups because we put an enhanced feedback loop in place to supplement the feedback system within GC (which is too limited).  That had people concerned about if it would "count against them" and whatnot.  You may also have to handle people who are fundamentally opposed to anything AI because of data centers being built in their area.  That's a trickier issue to handle and I think we're all struggling with a proper response.  
• If you could start over, what would you do differently? - I'd start sooner.  The sooner you begin projects like these with a pilot group the sooner you'll start to figure out what supporting system/services you'll need and which of those you'll need to improve/address

------------------------------
Chad Blanton
Director
------------------------------

Hello, Sanderley!

From the customer conversations I've had, AHT tends to be the easiest benefit to quantify, but rarely the most valuable one.

The feedback I hear most often is that Agent Copilot helps agents feel more confident, especially newer agents, by surfacing the right knowledge at the right time and reducing the need to switch between applications.

Beyond AHT, the KPIs customers are typically watching are ACW, FCR, QA scores, CSAT, onboarding time, and knowledge utilization. In several cases, the reduction in after-call work through AI-generated summaries has been one of the quickest wins.

Its success depends on having a solid knowledge strategy, clear success metrics, and strong user adoption. The customers seeing the best results are the ones that continuously improve both their knowledge content and their prompts based on agent feedback.

I agree with you @Lidyane Palmeira and wanted to add one more component related to AI-generated summaries. The value of the information doesn´t stays only on how quick the agents populate the data, but also for BI analysts, Admins and Managers, they can take advantage of the great summaries to run further analyses on their BI for extra business intelligence information alongside having the data with great level of information uploaded into their CRMs etc. Great job thanks for contributing to Sanderley´s question. This help Equatorial overall view about the benefits of this solution

Hello, Sanderley!

From the customer conversations I've had, AHT tends to be the easiest benefit to quantify, but rarely the most valuable one.

The feedback I hear most often is that Agent Copilot helps agents feel more confident, especially newer agents, by surfacing the right knowledge at the right time and reducing the need to switch between applications.

Beyond AHT, the KPIs customers are typically watching are ACW, FCR, QA scores, CSAT, onboarding time, and knowledge utilization. In several cases, the reduction in after-call work through AI-generated summaries has been one of the quickest wins.

Its success depends on having a solid knowledge strategy, clear success metrics, and strong user adoption. The customers seeing the best results are the ones that continuously improve both their knowledge content and their prompts based on agent feedback.

Great topic, San!

I've been tracking global AI benchmarks, and bridging the ROI gap is a real challenge. Recent MIT research shows that up to 95% of enterprise AI initiatives struggle to show direct P&L or EBIT impact.

Quantifying pure financial return and core operational metrics like AHT, Talk Time, or overall productivity is notoriously tough, and early rollouts can even bump Talk Time up temporarily. In practice, the shift is about looking at Customer Experience (CX) and Agent Experience (EX) wins first:

• Agent Experience (EX): Lower cognitive load, smoother onboarding, and cleaner notes through automated summaries.

• Customer Experience & Quality (CX): Better First Contact Resolution (FCR) and more consistent compliance scores driven by real-time knowledge surfacing.

I believe the real key is connecting these early CX and EX gains to long-term operational savings, taking a pragmatic approach to analyze what actually moves the needle for specific business reality.

Looking forward to hearing how others in production are quantifying these AHT gains!

------------------------------
Luciano Gabriel De Sousa Lima
Senior Projects and Performance Analyst
------------------------------

Great topic, San!

I've been tracking global AI benchmarks, and bridging the ROI gap is a real challenge. Recent MIT research shows that up to 95% of enterprise AI initiatives struggle to show direct P&L or EBIT impact.

Quantifying pure financial return and core operational metrics like AHT, Talk Time, or overall productivity is notoriously tough, and early rollouts can even bump Talk Time up temporarily. In practice, the shift is about looking at Customer Experience (CX) and Agent Experience (EX) wins first:

• Agent Experience (EX): Lower cognitive load, smoother onboarding, and cleaner notes through automated summaries.

• Customer Experience & Quality (CX): Better First Contact Resolution (FCR) and more consistent compliance scores driven by real-time knowledge surfacing.

I believe the real key is connecting these early CX and EX gains to long-term operational savings, taking a pragmatic approach to analyze what actually moves the needle for specific business reality.

Looking forward to hearing how others in production are quantifying these AHT gains!

------------------------------
Luciano Gabriel De Sousa Lima
Senior Projects and Performance Analyst
------------------------------

I couldn´t agree more with you @Luciano Gabriel De Sousa Lima and congratulations for always being on top of these discussions and going deepdive studiyng the AI trends. It´s a privilege to work with you and the team

For me the GC Agent Copilot capability of call summarization and suggesting wrap-up codes are the one what stood out. And also I completely agree with our community members regarding the necessity of prompt engineering in today's AI journey, the custom prompt is too good to help much more with this feature.