Script

So for people who don’t know you, we connected in the UK ServiceNow Community comms. You’ve been amazing in connecting and partnering with myself and Eclipse AI to drive the AI ship forward for our customers.

You lead the team at ServiceNow responsible for AI Success at Scale,

which really is aiming to help organisations harness the power of AI, automation, and ITSM best practices to drive efficiency and 10x customer experiences.

You have a significant experience in the SaaS space.

Led strategic transformation initiatives across multiple industries, including defence, construction, and retail.

Your passion and expertise spans AI adoption, IT operations, enterprise automation, and service improvement frameworks.

In short, it is brilliant to have you here today, and hopefully for many more discussions to help our community unlock the value of AI.

Scott: It is an exciting era to be in and it is exciting to work with our partners in the pursuit of AI value and success at scale for our customers, so it is great to be working a partner like Eclipse AI who is so passionate about AI as we are.

Chris: Absolutely, and with the dropping of ServiceNow’s AI Agent Studio and Orchestrator, things are exciting in the ServiceNow eco-system. Which leads me to the first thing I would love for us to start of with.

After what seems like forever, AI Agents are now accessible to build and improve your company’s efficiency and value of output overall. But I think there’s quite a bit of confusion in the market about what exactly an AI agent is versus an AI assistant. Could you help us clarify that distinction?

Scott: That’s a great place to start, Chris. There’s definitely confusion about the terminology, and it’s important we get clear on it because it fundamentally changes how organisations approach implementation.

An AI assistant, which most people are now familiar with, is essentially a tool that responds to your queries and requests, but it waits for your input. Think of ChatGPT, Claude, or ServiceNow’s Now Assist - they’re incredibly powerful but fundamentally reactive. You ask, they answer.

AI agents, on the other hand, take a significant step forward. They don’t just respond - they act autonomously on your behalf. They can make decisions, execute tasks, and coordinate work across systems without constant human guidance. They’re proactive rather than reactive.

The difference might sound subtle, but it’s transformative in practice.

Chris: That’s a really helpful distinction. I often explain it to clients by comparing an assistant to a junior colleague who’s clever but needs constant direction, versus an agent who’s more like an experienced team member who you can give an objective to and trust them to handle the details.

Would you say that’s a fair analogy?

Scott: Agreed, Chris. An assistant is there to help you with the task you’re already doing, but an agent can take the entire task off your plate. If we use your analogy, an assistant is like having someone in the room who can answer questions while you’re working, whereas an agent is like delegating the entire piece of work.

And ServiceNow’s AI Agent Orchestrator takes this even further - it’s like having a coordinator who knows which agents to assign different parts of a complex task to, ensuring they all work together seamlessly.

Chris: I’ve noticed a lot of confusion in the market about this. I was recently working with a client who was using Microsoft Copilot, and they were struggling to understand the difference between that and the new Copilot Agent Studio capabilities. They’d built workflows that used Copilot to summarise meeting notes and post them to SharePoint, but they couldn’t see how agents were different from these automated workflows.

Scott: That’s such a common confusion point that we see from ServiceNow. Many organisations have already built workflows around AI assistants, so they naturally ask, “What’s the difference between my current automation and these new agents?”

The key difference is in decision-making autonomy. Traditional workflows, even those incorporating AI assistants, follow predetermined paths that a human designed. Every branch in the decision tree had to be anticipated and programmed in advance.

Chris: Yes and then AI agents, by contrast, can assess situations and determine appropriate actions dynamically. They can decide which tools to use based on context rather than following a rigid script. That’s why I often say we’re moving from a world where technical people build workflows to emulate human decisions, to a world where we can just have a conversation with an agent to build that decision-making capability.

After what seems like for ever, AI Agents are now accessible to build and improve your company’s value of output. Still there seems to be a hesitation from enterprise to adopt AI to greater extent then just the AI assistant era. What are you finding?

Scott:

You know, Chris, what I’m finding is that there’s a fundamental disconnect in how organizations are approaching AI agents. Most companies are still treating this as a purely technical implementation—something that requires specialised IT skills and complex configurations.

The revolution that’s being missed is that AI agents represent a democratization of automation creation. Instead of business teams having to submit tickets and wait for IT to build workflows, they can now have a conversation with an agent to build another agent. It’s like teaching a new employee how to do a job, rather than programming a machine.

This shift hasn’t quite sunk in yet. Many organizations—especially in the public sector—are still keeping agent creation locked down in technical silos. They’re approaching it with the same mindset they used for previous technologies, where IT had to control everything.

The hesitation we’re seeing isn’t primarily about the technology itself. It’s about organizations not yet making the cognitive shift to viewing agents as something that regular employees can create and manage. They haven’t realised that they already have all the use cases they need—they’ve been hiring junior staff to do repetitive tasks for years, and those are exactly the tasks that agents can now handle.

What’s fascinating is that when this clicks for people—when they understand they can build agents through conversation rather than code—their entire perspective changes. Suddenly, they start seeing opportunities everywhere. But that requires a willingness to truly democratise AI capability across the organization, and that’s the challenging cultural shift many enterprises aren’t quite ready to make yet.

Chris: Ye this resonates massively with what we are seeing in the consultancy business. In a customer of ours, we have seen the CMO and the CIO argue over who should lead their business into the use of AI Agents. And quite frankly just demonstrates just how democratised this has truly become that the marketing department feels they can implement AI Agents over the IT department. But that leads me onto another discussion point I want to explore with you. Data and the fusion between traditional system design and AI agents.

The stakes couldn’t be higher. According to recent data, while 87% of business leaders believe their data ecosystem is ready for AI, 70% of technical practitioners spend hours daily addressing fundamental data issues.

This reality gap isn’t just a technical problem—it’s an existential threat to our customer’s organization’s ability to compete in an AI-powered world.

I’ve spent years watching organizations build increasingly complex workflows that inevitably hit the ceiling of what traditional automation can achieve. And drags the human resources into the pit of the mundane and the tedious.

The revelation is simple but profound: We should not be using system design to channel the agent; we should let the agent channel the system design.

**In other words, we need to stop thinking about AI agents as machines and start thinking about them as apprentices. Otherwise we risk overusing traditional workflow over agentic workflow and vice versa. What are you seeing here?

Scott:** Traditional workflow automation has always suffered from a fundamental constraint:

it requires technical expertise to implement.

This creates inevitable bottlenecks where business needs outpace IT capacity to build solutions. ServiceNow’s existing workflow capabilities are powerful, but they still follow this pattern—a process owner identifies a need, submits a request, and waits for technical teams to implement.

The data tells a stark story: Between 2023 and 2024, we saw a profound change in enterprise AI strategy from 80% buy/20% build to 53% buy/47% build. This reflects a growing recognition that off-the-shelf solutions, from agencies, often can’t address the unique needs of each organization.

The problem isn’t the technology—it’s the paradigm. When we construct rigid workflows, we’re essentially trying to predict every possible scenario in advance. This approach breaks down precisely when we need it most: in complex, cross-departmental processes where edge cases are the norm rather than the exception.

Chris: Yes that is what I love about ServiceNow’s AI Agent architecture.

It flips this model on its head.

Instead of building linear processes with predetermined endpoints, AI agents operate in what I call a “radial model”—the agent sits at the hub of a wheel with spokes extending outward to different resources, tools, knowledge bases, and even other specialised agents.

Linear Approach vs **Radial Approach** 
Agent as executor vs **Agent as orchestrator**
Fixed workflows vs **Adaptive pathways**
Task-centered vs **Purpose-centered** 
Limited by design vs **Limited by capabilities**

The Agent Orchestrator serves as the central intelligence that understands your company policies, people, and processes—what ServiceNow describes as “the most knowledgeable employee in the company.” It’s the component that plans work and ensures specialized AI agents collaborate effectively.

Meanwhile, the Agent Studio provides the environment where non-technical users can create specialized agents through natural language instructions rather than code or configuration. This is where the democratisation truly happens.

I’ve watched clients struggle with other AI Agent platforms, unable to distinguish between simple LLM capabilities and true agent functionality.

The key difference is autonomy and purpose—agents don’t just respond; they decide which tools to use based on context and objectives. They’re not merely tools; they’re collaborators.

Scott:

Complete agree, and that is the most profound implication of this shift, is who can create automation. Consider this fundamental change:

Old world: “I need automation for this process. Let me submit a request to IT.”

New world: “I need automation for this process. Let me build an agent to handle it.”

This represents nothing less than the democratisation of what was previously a specialised technical discipline. Just as spreadsheets liberated financial modeling from the priesthood of the IT department, AI agents are poised to democratise process automation.

Chris: I was talking to a customer executive for a large UK enterprise, they spoke of their frustration with their demand management process. Templates in SharePoint, transcripts of meetings, approvals in ServiceNow—all requiring manual coordination.

When I explained how a non-technical team could build agents to handle this end-to-end, his reaction was revelatory:

“You mean we don’t need to wait for IT to build this for us?”

And of course that leads us to the other issue we are seeing of course, and it is why Eclipse AI can pay the bills, is that customers are struggling to understand the use cases for AI agents. The cognitive shift is so massive that the obvious cannot be seen.

Scott: Yep, the counterintuitive truth that most organisations miss: you already have the use cases. You don’t need to invent them—they’ve been developing naturally throughout your organisation for years. The biggest misconception in AI agent implementation is believing we need to start with a blank slate.

Chris: Yes! The message I’ve been sharing with clients is disarmingly simple:

“You’ve been doing this for years, but what you’ve been doing is bringing in graduate and juniors, early in their careers to handle repeatable tasks.”

These are precisely the tasks that AI agents excel at handling.

When I say this to executives, I see the lightbulb moment—suddenly, they’re not trying to imagine hypothetical “AI use cases”; they’re looking at existing processes through a new lens. One ServiceNow customer I worked with identified 27 distinct tasks in their employee onboarding process that were being handled by junior staff. Each represented an opportunity for AI agents to handle.

Scott: The best starting points almost always involve processes where:

• Information needs to be collected from multiple systems
• Decisions follow consistent but nuanced patterns
• Tasks frequently cause bottlenecks or delays
• Employees regularly express frustration about repetitive work

While chatbots and simple customer service automation often come to mind first, they barely scratch the surface of what’s possible. The most transformative applications typically cross departmental boundaries—precisely where traditional workflow automation struggles.

Chris: I agree, traditional system design is overall cheaper to run as AI Agents use LLMs which is a high cost of compute. But they cannot produce the value at more complicated or nuanced decisioning that fixed or even dynamic workflows just struggle with. In our work with customers, I’ve seen remarkable results when agents are applied to processes like:

IT Change Management: Agents can coordinate the entire lifecycle—gathering requirements, assessing risk, scheduling implementations, and documenting outcomes—adapting dynamically to different types of changes rather than following rigid paths.

Employee Service Delivery: Rather than having separate processes for IT, HR, facilities, and finance requests, agents can provide a unified experience that routes, tracks, and resolves issues regardless of which departments need to be involved.

Customer Onboarding: Perhaps the most powerful example involves coordinating all the steps needed to bring a new customer online—from contract processing to account setup, service provisioning, and initial support—orchestrating activities across sales, legal, operations, and customer success teams.

BUT for me, what I need to impress strongly here, what makes these examples powerful isn’t just the efficiency gained; it’s that they reimagine processes that previously required constant human coordination and handoffs—the organisational equivalent of passing notes in class—it elevates the humans from the mundane and propels them into a space and time where they can produce more value.

Now every org is different right and their needs at anyone point in time will be different. So how do you see customers finding what would be best for AI agents and how to see the most value of their use realised?

Scott:

To identify the most promising opportunities in your organisation, look for these telltale signs:

1. Process Fragmentation: Tasks that require people to constantly switch between systems or teams.
2. Invisible Work: Coordination activities that don’t appear in formal process maps but consume significant time.
3. Information Hunting: Scenarios where people spend more time gathering information than acting on it.
4. Decision Fatigue: Repetitive judgments that follow consistent patterns but still require human attention.
5. Scalability Constraints: Processes that work fine at current volumes but would break under growth.

Chris: This is an excellent Pain Point framework. We recently helped a financial services client map these kind of pain points against their operations. They discovered their most promising opportunity wasn’t in their customer-facing processes (where they’d been focusing) but in their compliance documentation workflow—a complex, multi-department process riddled with precisely these challenges.

Now there is another subject I am quite passionate about that I would like to get your take on, especially from a ServiceNow perspective.

Let’s be brutally honest about what’s on everyone’s mind: will AI agents replace jobs? For me, the answer is both simpler and more nuanced than most commentary suggests. What are your thoughts?

Scott: In my conversations with executives and employees alike, I’ve found that organisations generally fall into two mindsets:

The Replacement Mindset: “How can we use AI to reduce headcount and cut costs?”

The Elevation Mindset: “How can we use AI to free our people from mundane tasks and elevate them to higher-value work?”

The data is clear on which approach yields better outcomes. McKinsey’s research shows that organisations focusing primarily on cost reduction through AI achieve 75% lower ROI than those prioritising growth and innovation.

Chris: That is so poignant. But this isn’t just about business outcomes—it’s about human potential. I’ve always adhered to a principle I learned early in my career: the best way to grow is to make yourself dispensable. It seems counterintuitive, but when you make yourself dispensable in your current role, you create the opportunity to grow into something more valuable.

This same principle applies perfectly to how organisations should approach AI agents. They’re not a replacement strategy; they’re an elevation strategy, they’re as you’ve said, a collaborator.

Scott: Yes absolutely. The organisations that will thrive aren’t those that simply replace tasks with automation; they’re those that reimagine roles entirely. This requires a profound cognitive shift for both leaders and employees.

For employees, the shift is from being task executors to becoming:

• Automation Architects: Defining what agents should do and how they should behave
• Exception Handlers: Managing the complex cases that fall outside agent capabilities
• Strategic Advisors: Focusing on improvement and innovation rather than execution
• Agent Managers: Overseeing and optimizing agent performance

Chris: That resonates again. I’ve seen this transformation up close. At one healthcare organisation, a team of eight administrators who previously spent 70% of their time on routine documentation now manage a fleet of agents that handle that work. Their new focus? Patient experience improvement initiatives that had always been deprioritised due to operational demands.

Scott: Yes and for executives, the cognitive shift is equally profound—from viewing AI as a cost-saving tool to seeing it as a force multiplier for human talent. Marc Benioff’s claim, that he quickly soon retracted, that Salesforce will “stop hiring software engineers” represents precisely the wrong mindset. ServiceNow’s approach, focusing on how to “10x our customers’ work experience for all employees,” points in the right direction.

Chris: It is about “Building a Culture of Continuous Elevation”

The transition doesn’t happen automatically. It requires intentional culture-building around three key principles:

1. Redefine productivity: Move from measuring output to measuring impact or value
2. Reward innovation: Celebrate those who find new ways to leverage agent capabilities
3. Reinvest gains: Visibly channel efficiency improvements into growth initiatives

We challenge clients to created what we call an “Elevation Pledge”—a formal commitment that any time saved through agent automation would be reinvested in either professional development or innovation projects. This simple policy transforms the narrative from “will I be replaced?” to “what could I accomplish if I had 40% more time?”.

But to get here, does require some amount of preparing for success, the ground work, whilst ServiceNow’s AI Agent capabilities are powerful, they don’t exist in a vacuum.

Scott: yes they require a foundation to build upon—what I colourfully described as “the combi boiler problem.” You’ve got all these people excited about hot water, but if the boiler isn’t powerful enough, everyone ends up with cold showers.

From a technical perspective, you need to start with your enterprise data.

Data Readiness: Agents can only work with the information they can access. This doesn’t mean you need perfect data (a myth that paralyses many organisations), but you do need:

• Clear data ownership and governance
• Accessible, properly structured information
• Consistent naming conventions and taxonomies

Integration Framework: The most powerful agent use cases typically span multiple systems. ServiceNow’s platform provides robust integration capabilities, but you’ll need to ensure:

• Appropriate API access and permissions
• Clear security boundaries and controls
• Performance considerations for real-time operations

Chris: The data and architecture are indeed crucial to success but also my advice would be to not bite off the entire meal in one go as you might choke!

Phased Implementation: Rather than viewing agent adoption as a binary switch, think of it as a maturity journey:

1. Single-domain agents with limited autonomy (read-only)
2. Cross-domain agents with supervised write capabilities
3. Orchestrated agent teams with progressive autonomy

When working with clients, I often recommend starting with what I call “observable agents”—those that can demonstrate their reasoning and planned actions before implementing them. This builds trust while allowing the organisation to verify performance in real-world scenarios.

Scott: The other concern in helping enterprise adopt AI is the Organisational Readiness

The technical foundation is only half the equation. Equally important is preparing your organisation for this new way of working:

Governance Without Strangulation: Create frameworks that enable rather than restrict. The worst outcome is governance so cumbersome that it recreates the very bottlenecks agents are meant to eliminate.

ServiceNow’s Yokohama release introduced AI Acceptable Use Policies and comprehensive logging specifically to address these concerns. These capabilities allow security teams to set clear boundaries while still enabling business users to create and deploy agents within those guardrails.

Skills Development: While agent creation is becoming democratised, it still requires specific skills:

• Precise problem definition
• Clear instruction writing
• Testing and feedback interpretation
• Understanding agent limitations

Chris: Yes and then there is Change Management: Never underestimate the human side of technological change. Effective adoption requires:

• Executive sponsorship and visible use
• Early wins with measurable impact
• Community building around agent creators
• Recognition of innovative implementations

What is the ServiceNow position on a successful agent implementation they would recommend our customers?

Scott: Successful agent implementation typically follows this pattern:

1. Pilot Phase
   • Select 2-3 high-value, low-risk use cases
   • Establish clear success criteria
   • Build initial agents with close support and training
   • Document lessons learned and adjust approach
2. Expansion Phase
   • Develop internal enablement materials
   • Create reusable templates and patterns
   • Establish a community of practice
   • Expand to 5-10 additional use cases
3. Transformation Phase
   • Develop cross-functional agent orchestration
   • Integrate with strategic business initiatives
   • Establish centres of excellence
   • Implement continuous improvement cycles

This phased approach allows organisations to build capability and confidence while progressively tackling more complex scenarios. It’s remarkably similar to how successful ServiceNow implementations have always worked—starting with clear foundations and expanding methodically.

Chris: Yes I can attest to some of these frameworks for implementing ServiceNow and they have solid foundations for implemetation of any product. This has been super interesting conversatoin.

For me the organisations that will thrive in the coming decade aren’t necessarily those with the most advanced AI capabilities—they’re those that most effectively democratise those capabilities across their workforce. This isn’t just about technology; it’s about unleashing the collective intelligence and creativity of your entire organisation.

What makes this moment unique is the convergence of three factors for me:

1. Technical Feasibility: ServiceNow’s AI Agent Orchestrator and Studio make previously complex automation accessible to non-technical users.
2. Business Necessity: The pace of change demands greater agility and efficiency than traditional approaches can deliver.
3. Human Potential: Employees increasingly expect meaningful work that leverages their unique human capabilities.

Scott: We’re entering an era where the distinction between “technical” and “non-technical” roles is blurring. Just as the spreadsheet democratised financial modelling and the dashboard democratised data analysis, AI agents are democratising process automation.

Chris: The question is no longer whether enterprise will adopt these capabilities—it’s whether it will have lead or follow in this transformation. Those who lead stand to gain not just operational efficiencies but strategic advantages in adaptability, innovation, and talent development