The Two Sides of AI: Channeling AI Experimentation at Enterprise Scale
The Two Sides of AI: Channeling AI Experimentation at Enterprise Scale

Scott Lerner shares how Akamai combines employee experimentation with governance to scale AI across global services.
Share
Based on a Keynote at FRONTIER by Planhat, on 10 November 2026.
Give 2,600 employees access to AI, and you’ll quickly have 2,600 ideas about where to use it. Without coordination, many will end up solving the same problems in different ways, creating overlapping tools, inconsistent processes and security risks. Yet restricting experimentation means losing the ideas of the people who understand your operational problems best. How do you give employees the freedom to innovate while maintaining the control an enterprise needs?
That’s the challenge at Akamai, where Global Services supports customers across more than 26 countries, multiple time zones and a complex product portfolio. AI can improve work across consulting, customer success, support and managed services, but deploying it at our scale requires coordinated experimentation, governance and investment.
Guardrails aren't there to be brakes. They're there to allow you to drive fast, safely.”
“
Based on a Keynote at FRONTIER by Planhat, on 10 November 2026.
Give 2,600 employees access to AI, and you’ll quickly have 2,600 ideas about where to use it. Without coordination, many will end up solving the same problems in different ways, creating overlapping tools, inconsistent processes and security risks. Yet restricting experimentation means losing the ideas of the people who understand your operational problems best. How do you give employees the freedom to innovate while maintaining the control an enterprise needs?
That’s the challenge at Akamai, where Global Services supports customers across more than 26 countries, multiple time zones and a complex product portfolio. AI can improve work across consulting, customer success, support and managed services, but deploying it at our scale requires coordinated experimentation, governance and investment.
Guardrails aren't there to be brakes. They're there to allow you to drive fast, safely.”
“
Based on a Keynote at FRONTIER by Planhat, on 10 November 2026.
Give 2,600 employees access to AI, and you’ll quickly have 2,600 ideas about where to use it. Without coordination, many will end up solving the same problems in different ways, creating overlapping tools, inconsistent processes and security risks. Yet restricting experimentation means losing the ideas of the people who understand your operational problems best. How do you give employees the freedom to innovate while maintaining the control an enterprise needs?
That’s the challenge at Akamai, where Global Services supports customers across more than 26 countries, multiple time zones and a complex product portfolio. AI can improve work across consulting, customer success, support and managed services, but deploying it at our scale requires coordinated experimentation, governance and investment.
“Guardrails aren't there to be brakes. They're there to allow you to drive fast, safely.”
Balancing the two sides of enterprise AI
At Akamai, we think about AI as two sides of the same coin. The first is AI embedded in the products we sell, helping customers investigate security events, retrieve information and operate their systems. The second is AI that helps our own employees perform their work, whether through applications developed internally or products adopted from other companies.
For Global Services, internal AI offers substantial opportunities: we handle repetitive support requests, maintain knowledge across systems and coordinate specialists to resolve complex problems. But an agent might give incorrect information, disclose sensitive data or act outside its authority. Governance must protect customers while preserving room for innovation and human judgment.
Once you've identified a worthwhile problem, start small.”
“
Balancing the two sides of enterprise AI
At Akamai, we think about AI as two sides of the same coin. The first is AI embedded in the products we sell, helping customers investigate security events, retrieve information and operate their systems. The second is AI that helps our own employees perform their work, whether through applications developed internally or products adopted from other companies.
For Global Services, internal AI offers substantial opportunities: we handle repetitive support requests, maintain knowledge across systems and coordinate specialists to resolve complex problems. But an agent might give incorrect information, disclose sensitive data or act outside its authority. Governance must protect customers while preserving room for innovation and human judgment.
Once you've identified a worthwhile problem, start small.”
“
Balancing the two sides of enterprise AI
At Akamai, we think about AI as two sides of the same coin. The first is AI embedded in the products we sell, helping customers investigate security events, retrieve information and operate their systems. The second is AI that helps our own employees perform their work, whether through applications developed internally or products adopted from other companies.
For Global Services, internal AI offers substantial opportunities: we handle repetitive support requests, maintain knowledge across systems and coordinate specialists to resolve complex problems. But an agent might give incorrect information, disclose sensitive data or act outside its authority. Governance must protect customers while preserving room for innovation and human judgment.
“Once you've identified a worthwhile problem, start small.”
From the Wild West to structured experimentation
Our AI journey began before we had a formal strategy. As AI tools became widely accessible, employees naturally started experimenting, building applications and automating problems they encountered every day. Much of this happened independently, without the visibility or governance our IT and information security teams would normally expect.
We created MountAIn, an internal program designed to bring structure to that experimentation without suppressing it. Alongside projects identified by leadership, we established a way for employee innovators, whom we called free climbers, to register their ideas. This gave us an inventory of experiments across the organization and allowed us to connect teams working on similar problems. It also created a path for establishing consistent expectations around security, compliance and operational sustainability.
From the Wild West to structured experimentation
Our AI journey began before we had a formal strategy. As AI tools became widely accessible, employees naturally started experimenting, building applications and automating problems they encountered every day. Much of this happened independently, without the visibility or governance our IT and information security teams would normally expect.
We created MountAIn, an internal program designed to bring structure to that experimentation without suppressing it. Alongside projects identified by leadership, we established a way for employee innovators, whom we called free climbers, to register their ideas. This gave us an inventory of experiments across the organization and allowed us to connect teams working on similar problems. It also created a path for establishing consistent expectations around security, compliance and operational sustainability.
From the Wild West to structured experimentation
Our AI journey began before we had a formal strategy. As AI tools became widely accessible, employees naturally started experimenting, building applications and automating problems they encountered every day. Much of this happened independently, without the visibility or governance our IT and information security teams would normally expect.
We created MountAIn, an internal program designed to bring structure to that experimentation without suppressing it. Alongside projects identified by leadership, we established a way for employee innovators, whom we called free climbers, to register their ideas. This gave us an inventory of experiments across the organization and allowed us to connect teams working on similar problems. It also created a path for establishing consistent expectations around security, compliance and operational sustainability.
Turning experiments into scalable solutions
One of our first successful applications was Agent Pulse, a live-event anomaly analyzer developed for major sporting events, including Indian Premier League cricket and international football. During a live broadcast, there may be only minutes to diagnose and correct a technical issue before the event moves on. Agent Pulse reduced diagnosis time for known issues from approximately 10–15 minutes to two–three minutes, giving our teams a better opportunity to intervene while the event is still underway. We’re now planning to extend it across the major internet events Akamai supports.
Case Summarize retrieves information from complex support histories and produces summaries users can investigate through follow-up questions. We built it internally because Salesforce’s capabilities didn’t meet our needs, then migrated it into Agentforce as the vendor’s platform improved. We’re seeing better results and support without maintaining the application ourselves, reinforcing the need to revisit build-versus-buy decisions regularly.
These projects also taught us to evaluate how applications will operate beyond the proof-of-concept stage. Before scaling a solution, we need to understand who will maintain it, what it will cost and whether it can reliably support thousands of employees. A useful experiment becomes an enterprise capability only when those operational requirements are addressed.
Turning experiments into scalable solutions
One of our first successful applications was Agent Pulse, a live-event anomaly analyzer developed for major sporting events, including Indian Premier League cricket and international football. During a live broadcast, there may be only minutes to diagnose and correct a technical issue before the event moves on. Agent Pulse reduced diagnosis time for known issues from approximately 10–15 minutes to two–three minutes, giving our teams a better opportunity to intervene while the event is still underway. We’re now planning to extend it across the major internet events Akamai supports.
Case Summarize retrieves information from complex support histories and produces summaries users can investigate through follow-up questions. We built it internally because Salesforce’s capabilities didn’t meet our needs, then migrated it into Agentforce as the vendor’s platform improved. We’re seeing better results and support without maintaining the application ourselves, reinforcing the need to revisit build-versus-buy decisions regularly.
These projects also taught us to evaluate how applications will operate beyond the proof-of-concept stage. Before scaling a solution, we need to understand who will maintain it, what it will cost and whether it can reliably support thousands of employees. A useful experiment becomes an enterprise capability only when those operational requirements are addressed.
Turning experiments into scalable solutions
One of our first successful applications was Agent Pulse, a live-event anomaly analyzer developed for major sporting events, including Indian Premier League cricket and international football. During a live broadcast, there may be only minutes to diagnose and correct a technical issue before the event moves on. Agent Pulse reduced diagnosis time for known issues from approximately 10–15 minutes to two–three minutes, giving our teams a better opportunity to intervene while the event is still underway. We’re now planning to extend it across the major internet events Akamai supports.
Case Summarize retrieves information from complex support histories and produces summaries users can investigate through follow-up questions. We built it internally because Salesforce’s capabilities didn’t meet our needs, then migrated it into Agentforce as the vendor’s platform improved. We’re seeing better results and support without maintaining the application ourselves, reinforcing the need to revisit build-versus-buy decisions regularly.
These projects also taught us to evaluate how applications will operate beyond the proof-of-concept stage. Before scaling a solution, we need to understand who will maintain it, what it will cost and whether it can reliably support thousands of employees. A useful experiment becomes an enterprise capability only when those operational requirements are addressed.
Directing innovation toward strategic priorities
As MountAIn matured, employees wanted to experiment but didn’t always know which problems were worth solving. We asked each Global Services function to identify three to five priorities for AI investment: our North Stars.
Those priorities help employees focus on problems with meaningful business impact and give leadership a basis for allocating resources. We’re now planning the next stage: redesigning complete processes around AI rather than simply automating individual tasks within existing workflows. That requires identifying where automation can increase revenue, reduce operating costs or improve customer outcomes, while determining where human involvement remains essential.
Directing innovation toward strategic priorities
As MountAIn matured, employees wanted to experiment but didn’t always know which problems were worth solving. We asked each Global Services function to identify three to five priorities for AI investment: our North Stars.
Those priorities help employees focus on problems with meaningful business impact and give leadership a basis for allocating resources. We’re now planning the next stage: redesigning complete processes around AI rather than simply automating individual tasks within existing workflows. That requires identifying where automation can increase revenue, reduce operating costs or improve customer outcomes, while determining where human involvement remains essential.
Directing innovation toward strategic priorities
As MountAIn matured, employees wanted to experiment but didn’t always know which problems were worth solving. We asked each Global Services function to identify three to five priorities for AI investment: our North Stars.
Those priorities help employees focus on problems with meaningful business impact and give leadership a basis for allocating resources. We’re now planning the next stage: redesigning complete processes around AI rather than simply automating individual tasks within existing workflows. That requires identifying where automation can increase revenue, reduce operating costs or improve customer outcomes, while determining where human involvement remains essential.
Guardrails should allow you to move faster
Scaling AI requires close coordination with IT, information security and the wider business. At Akamai, we’ve worked with these teams to ensure experiments comply with corporate requirements and use approved technologies. We’re also extending the lessons from MountAIn through Horizon, a company-wide initiative developing AI governance and encouraging knowledge sharing across departments.
Effective governance establishes what employees can build, which data they can use and how successful experiments progress into production. With those requirements clear, teams can innovate without repeatedly navigating the same compliance questions or introducing avoidable risks. “Guardrails aren’t there to be brakes. They’re there to allow you to drive fast, safely.”
Start by identifying the problems that matter most this quarter, over the next 12 months and across the next three to five years. For each proposed initiative, ask a simple question: “So what?” If you implement this tool or redesign this workflow, how will it improve your customers’ experience or your business performance? If you can’t answer that, you’re probably working on the wrong thing.
Once you’ve identified a worthwhile problem, start small. Run proofs of concept, test different approaches and retain human oversight while establishing reliability. Evaluate security, maintenance requirements and operating costs before scaling.
Guardrails should allow you to move faster
Scaling AI requires close coordination with IT, information security and the wider business. At Akamai, we’ve worked with these teams to ensure experiments comply with corporate requirements and use approved technologies. We’re also extending the lessons from MountAIn through Horizon, a company-wide initiative developing AI governance and encouraging knowledge sharing across departments.
Effective governance establishes what employees can build, which data they can use and how successful experiments progress into production. With those requirements clear, teams can innovate without repeatedly navigating the same compliance questions or introducing avoidable risks. “Guardrails aren’t there to be brakes. They’re there to allow you to drive fast, safely.”
Start by identifying the problems that matter most this quarter, over the next 12 months and across the next three to five years. For each proposed initiative, ask a simple question: “So what?” If you implement this tool or redesign this workflow, how will it improve your customers’ experience or your business performance? If you can’t answer that, you’re probably working on the wrong thing.
Once you’ve identified a worthwhile problem, start small. Run proofs of concept, test different approaches and retain human oversight while establishing reliability. Evaluate security, maintenance requirements and operating costs before scaling.
Guardrails should allow you to move faster
Scaling AI requires close coordination with IT, information security and the wider business. At Akamai, we’ve worked with these teams to ensure experiments comply with corporate requirements and use approved technologies. We’re also extending the lessons from MountAIn through Horizon, a company-wide initiative developing AI governance and encouraging knowledge sharing across departments.
Effective governance establishes what employees can build, which data they can use and how successful experiments progress into production. With those requirements clear, teams can innovate without repeatedly navigating the same compliance questions or introducing avoidable risks. “Guardrails aren’t there to be brakes. They’re there to allow you to drive fast, safely.”
Start by identifying the problems that matter most this quarter, over the next 12 months and across the next three to five years. For each proposed initiative, ask a simple question: “So what?” If you implement this tool or redesign this workflow, how will it improve your customers’ experience or your business performance? If you can’t answer that, you’re probably working on the wrong thing.
Once you’ve identified a worthwhile problem, start small. Run proofs of concept, test different approaches and retain human oversight while establishing reliability. Evaluate security, maintenance requirements and operating costs before scaling.
The Archive
[
Discover All
]
[
Through the Lens
]
[
Talking Planhat
]
[
Expert Advice
]
[
In Perspective
]
[
In Conversation
]
Load more
The Archive
[
Discover All
]
[
Through the Lens
]
[
Talking Planhat
]
[
Expert Advice
]
[
In Perspective
]
[
In Conversation
]
Load more
The Archive
[
Discover All
]
[
Through the Lens
]
[
Talking Planhat
]
[
Expert Advice
]
[
In Perspective
]
[
In Conversation
]
Load more

Don't miss a beat.

Don't miss a beat.








