Part 1 of the Systemic Shift to Scaling AI from Pilot to Scale series, examining how organizations can transform AI initiatives into enterprise-wide value
At Ateko we frequently see organizations struggle to translate the excitement of a pilot into organizational wide value. While AI adoption is high, with most organizations using AI in at least one function, studies suggest that a significant majority (up to 62%) remain stuck in the early experimenting or piloting stages.
“If your organization is still in these early stages as we approach the halfway point of 2026, there’s a good chance you’re already behind the curve when it comes to adopting AI at scale.” This blog series is meant to help break down what can feel like an overwhelming challenge and provide a clearer framework for scaling AI across your organization.
To move beyond isolated AI projects and achieve enterprise-wide scaling across your Salesforce, ServiceNow, and Cloud ecosystems, a systems based approach is necessary, abandoning the previous focus on “quick win pilots.” Based on insights from the AI Daily Brief, OpenAI’s research on scaling, and our experience helping clients deploy solutions like Agentforce Data 360 (formerly Data Cloud) and Now Assist, we have identified four fundamental shifts required for this transition.
1. Shift from Thinking About Tools to Thinking About Systems
The Shift: Historically, we validated software by asking if a specific tool was fit for purpose. We started small, tested a use case, and scaled. But AI moves differently. Its capabilities evolve in weeks, not quarters/years, and its impact reaches every part of the organization.
The Application: To achieve organizational success with AI, we must shift our perspective from a collection of “tools” to an integrated system. True enterprise value is derived from the comprehensive systems built around AI, rather than being solely dependent on the specific model or tool chosen.
For example, don’t worry too much about perfecting a generative prompt’s performance in a tool like Agentforce or Now Assist, right away. It’s better to fine-tune that stuff later, once real users have played around with the tool and given you some feedback.
For a true enterprise wide solution, a strong, connected ecosystem is vital, centered around a flexible data platform such as Salesforce Data360 (formerly Data Cloud) or Databricks to centralize and clean critical data. This requires robust data pipelines across your wider cloud setup (GCP, AWS, Azure, Snowflake, Databricks,). Without easy access to well-managed data, scaling your AI projects will be a constant, uphill battle.
Questions to ask yourself:
- Do you have executive alignment, governance and data access sorted out?
- Are you evaluating AI models in isolation, or as part of a unified systems based strategy that incorporates important feedback loops?
- Does your strategy account for the unstructured data stored in your cloud environment (AWS/GCP/Azure) and how that feeds into your solution?
- Is your “system” designed so that Agentforce or Now Assist has access to your critical internal knowledge bases, understanding not just who the customer is, but how your company ticks? How will your AI solution know how to interpret those obscure acronyms specific to your business or industry that are used regularly and how can you solve this problem at scale?
2. Thinking at a New Velocity
The Shift: Tools no longer stay static. This demands a new operating rhythm that balances speed with structure and evolves as fast as the technology itself.
The Application: For our solutions, this means accepting that the AI models and agents powering your applications are not “set and forget.”
This approach will feel familiar to anyone in Business Intelligence. In BI, we rarely build a dashboard, launch it, and never touch it again. Instead, we work iteratively: we release a version, observe how the business uses the data, and refine the metrics and views based on real-world feedback.
It is crucial to apply this exact same mindset when thinking about AI at scale. You generally cannot treat an AI agent like a fixed, one-time launch. You launch it, monitor its conversations, identify where it needs better grounding data from tools like Data360 (formerly Data Cloud), or access to knowledge bases stored elsewhere, and refine its instructions. The difference between a good agent and a great one is often the speed of this iteration loop, tuning the agent as if you were refining a high-value dashboard.
Questions to ask yourself:
- Is your governance model flexible enough to handle weekly iterations, or is it stuck in quarterly release cycles?
- Do you treat your AI Agents like static applications, or like agile dashboards that require constant tuning?
- Do you have a feedback loop in place to monitor “drift” in your agents? Drift is when your AI agent’s answers slowly become less accurate or helpful because your business information and conditions are constantly changing.
3. Realizing Solutions Can Come from Anywhere
The Shift: Innovation is no longer the sole domain of your technical teams. Innovation can, and should, come from any team and any seniority level. There is currently no prerequisite for figuring out how to use AI better.
The Application: In the context of Business Intelligence, a Sales Representative using Agentforce, might discover a prompt that drastically improves their ability to prioritise their leads and consolidate key information for them on these people.
Because Agentforce is built on low-code principles, your Salesforce Admins are now your most valuable AI architects. They understand the data model better than anyone. These internal champions can translate general capabilities into specific organizational context. It is imperative that we create open channels for idea intake to harness this collective intelligence.
Questions to ask yourself:
- Do you have a mechanism for non-technical staff to submit AI use cases?
- Are you empowering your platform Admins to act as AI product owners using tools like Agent Builder in Salesforce, or AI Agent Studio in ServiceNow?
- How are you capturing AI wins and ensuring that these wins are shared across the key business units that could use them?
- Is your company actively encouraging their teammates to share and discuss ideas and win with AI?
4. Focusing on Compounding ROI
The Shift: It is easy to view AI impact as disconnected metrics. We must instead view these as cumulative and linked, resulting in compounding ROI on multiple fronts, from increased efficiency to measurable impact on time-to-resolution.
The Application: If an AI agent within a workflow saves 10 minutes on a ticket, that is an isolated save. Compounding ROI is achieved when that time saved frees up the agent to handle a complex “Priority 1” issue recorded in ServiceNow or Salesforce Service Cloud, which results in a contract expansion in Sales Cloud.
When we connect these systems, we ensure our high-effort projects translate into the ROI that matters most to your business, whether that’s net new revenue for Sales or faster resolution times for Service.
Questions to ask yourself:
- Are you measuring time saved, or how that saved time is reinvested?
- Do your KPIs link efficiency in Service (ServiceNow/Service Cloud) to revenue in Sales (Sales Cloud)?
- Are you optimizing for individual task speed or total workflow value?
Next Steps
These four mental shifts, from tools to systems, increasing velocity, democratizing solutions, and compounding ROI, prepare us for the framework ahead. In the coming posts, we will dive deeper into setting foundations with Data Cloud, creating AI fluency, and building scalable products with tools like Agentforce.
If you are ready to move your AI initiatives from the “lab” to the “line of business,” connect with us. Our practice leaders can help you engineer the systems required for scale.
