Key outcomes
By rewiring its order operations with agentic AI, Zones has unlocked:
- 50% reduction in customer order verification times.
- 20,000 annual hours of manual effort identified for automation
- 29 enterprise customers supported by the governed AI workflow.
- A reusable AI platform for future operational workflows.
- 17-month payback period on the investment.
Overview
Zones is a global technology solutions and services company that helps organizations solve complex IT challenges and achieve better business outcomes. Operating in more than 150 countries, Zones empowers the people behind every organization to do their work securely and efficiently from anywhere in the world. That commitment to delivering exceptional technology experiences starts with its own digital workplace.
The company is known for its highly personalized, white-glove services for large enterprise clients. However, behind that premium experience sat a complex web of customer-specific procedures, many of which existed only in standard operating procedures (SOPs) and the expertise of a handful of specialists.
To deliver that experience more consistently and at greater scale, Zones partnered with Thoughtworks to build its first production-grade agentic AI capability. More than an automation initiative, the project marked a strategic step in the company's AI journey, extending AI from individual productivity into the heart of enterprise operations.
The challenge: Automating a highly complex white-glove service
Enterprise orders flowed into Zones' order management system (OMS) carrying unique, complex contractual requirements for shipping, billing and fulfillment.
Because the OMS couldn't interpret these bespoke rules, customer success managers (CSMs) were forced to manually review every order, consulting SOPs to update up to 17 specific touchpoints before release.
This created several critical business hurdles:
High operational costs: Manual back-office work cost roughly $2 million annually.
Process friction: It increased quote-to-cash times and the risk of human error and SLA breaches, without adding visible customer value.
Limited automation: Past attempts to hardcode rules failed; static rule engines quickly became outdated as contracts and freight terms constantly changed.
To solve this, Thoughtworks proposed an agentic platform that dynamically interprets each customer's unique procedures and recommends the right action for every order.
The goal was to create a foundational, reusable architecture for future AI use cases across the business.
The solution: Rewiring order management with Microsoft Foundry
Two core ideas shaped the platform's design. First, instead of hardcoding business rules, the platform transforms customer-authored SOPs into machine-readable knowledge that evolves alongside changing procedures. Second, humans remain firmly in control: the AI generates recommendations, but every change requires CSM approval.
Thoughtworks delivered the work in phases, also helping Zones secure Microsoft Enterprise Cloud Investment Funding (ECIF) to accelerate development.
The new event-driven workflow operates in four stages:
Interpret: A dedicated service converts each SOP into explicit conditions and actions.
Detect: A custom service monitors new orders, with events carried through Azure Service Bus.
Recommend: A custom C# orchestration layer sends the order and its procedures to AI Agents via Microsoft Foundry, which return executable recommendations.
Stage for approval: Recommendations are staged in Azure SQL, not applied directly. When a CSM opens the order, a pop–up shows the field, its current and proposed values, and the SOP evidence; upon approval, the order updates via REST APIs.
By shifting the CSM's role from manual data entry to simple exception review, a process that previously required up to a minute of manual editing now takes just 10 to 15 seconds, freeing up CSMs to focus on proactive customer relationships rather than administrative tasks.
The technology behind the workflow
The platform is Azure-native, which integrates seamlessly with Zones' existing Microsoft Dynamics 365 environment. Microsoft Foundry provides the environment for the AI agents, while Azure OpenAI powers the recommendations. Supporting services run on Azure Kubernetes Service (AKS), with SOPs stored in Azure Cosmos DB and recommendations staged in Azure SQL.
Azure Service Bus connects the workflow through event-driven messaging, while Azure Monitor, Application Insights, and Power BI provide robust system monitoring and business reporting.
Responsible by design
Every AI recommendation is transparent, traceable, and grounded strictly in approved SOPs and structured order data,not conversational memory. CSMs retain full control over every decision, with supporting evidence attached to every recommendation.
When confidence is low, no recommendation is generated, and the order seamlessly follows the standard manual review process.
This approach ensures consistent, auditable outputs while enabling cost-effective scaling and mitigating AI token risks. Enterprise-grade security and governance are enforced through Azure Managed Identity, Key Vault, role-based and least-privilege access, encryption in transit and at rest, private networking, version-controlled prompts, SOP versioning and a complete audit trail of every system action.
The outcome: Institutional knowledge, operationalized
The platform has established a new operating model for enterprise order management at Zones. It already supports 29 enterprise customers and automates more than 16 customer-specific order management touchpoints.
As the system scales, the operational impact will continue to compound, supported by a 17-month payback period that demonstrates a clear business case for operational AI at scale.
20,000 annual manual effort hours identified for automation:
~14,000 hours expected to be realized under the current operating model.
50% reduction in verification times:Reducing order reviews from up to one minute to just 10–15 seconds
|
Nearly 90% AI recommendation accuracy:
Applying customer-specific procedures consistently across enterprise orders.
500,000+ orders and transactions expected to benefit annually:
Extending AI-driven recommendations across enterprise order operations.
Beyond the numbers
While the operational savings are significant, the platform is delivering strategic outcomes that directly elevate the Zones brand:
An improved, more consistent customer experience across every touchpoint.
Reduced SLA exposure and faster order-to-fulfillment cycles.
Greater consistency and auditability across every recommendation and approval.
Institutionalization of operational knowledge that once lived only in SOPs and in experts' heads.
Reduced dependency on tribal knowledge and a small number of subject matter experts.
This initiative has transformed how we operationalize customer-specific knowledge. By combining AI with human oversight, we've streamlined complex order workflows, and created a scalable foundation for enterprise operations. Thoughtworks has been a trusted partner in helping us turn our vision for agentic AI into a production-ready capability that delivers measurable business value.
What’s next: A scalable pattern for order operations
The platform is now fully operational across enterprise order management and has entered the realization phase. As adoption grows and confidence levels continue to rise, Zones expects to progressively reduce human-in-the-loop intervention while maintaining strict governance and oversight.
Built around a reusable procedure-to-action architecture, the platform can extend beyond order management to support future agentic AI workflows across the business, without redesigning the underlying system.
For Zones, this is a foundation for scaling expertise across the business, wherever it's needed next.
Our enterprise customers trust us to get every order exactly right. Partnering with Thoughtworks, we've turned operating knowledge that once lived in SOPs and in our people's heads into a governed, AI-driven workflow that recommends the right action on every order, with our teams always in control. It's our first agentic AI capability in production, and it's already making us faster and more consistent.
