CASE STUDY: Sopra Steria

Sopra Steria’s Digital Platform Services business line is a managed service provider. Across France, Poland, and India, 1200 experts operate the cloud, infrastructure, and service-desk environments behind dozens of European enterprises in Telecom & Energy, Aerospace, Transport, Banking, Industry, Public Sector, and Defense. The expertise from engineers is the value Sopra Steria sells to its clients. When Anthony Pinto, Head of AI for DPS, set out to deploy AI across the service desk, the discipline was clear: every KPI the deployment needed to move would be baselined before any feature went live. NiCE Copilot for Agents went live inside the existing NiCE CXone footprint, connected to client knowledge environments through NiCE Knowledge Management, and rolled out under a federated model with reusable assets. Average handle time on standard tickets dropped 9%. First call resolution improved four points. And on a separate Sopra Steria contact center platform that does not yet include the tool, agents on the legacy client are now asking when they will be migrated to the one that does.

There are agents on a legacy client asking, when can we migrate to the new contact center? I want this tool in my day to day. To create that kind of feeling in the team is really cool.

Anthony Pinto

Head of AI,
Digital Platform Services,
Sopra Steria

01 Before

Actually, we provide expertise through a service. The next phase was making the workflow match.

Sopra Steria’s Digital Platform Services business line is a managed service provider. Cloud, infrastructure, modern workspace, service desk: +1200 engineers across France, Poland, and India operate the systems behind dozens of European enterprises in Telecom & Energy, Aerospace, Transport, Banking, Industry, Public Sector, and Defense. In France alone, DPS serves more than 50 enterprise clients. When a client signs a contract with Sopra Steria, what they are buying is a workforce. The engineer who picks up the call is the moment of truth on every commercial relationship the business holds.

The work itself was substantive. The shape of the workflow was where the next opportunity sat. A single client interaction asked an engineer to move across a chain of separate systems. Receive the call or chat in the contact center. Open the client’s ITSM, which might be ServiceNow or Zendesk. Search the knowledge base, which might be Confluence or SharePoint. Produce an answer. Return to the contact center and respond. Each system was its own URL, its own login surface, and each pivot was time the engineer was no longer giving to the client.

DPS had also been thinking carefully about how to invest in AI. A first pilot had delivered something the team couldn’t prove. There was no baseline captured before deployment, so when the question came — did this actually work? — there was no honest answer. That gap taught a discipline that shaped every decision after it: the value of any AI investment has to be measurable against a baseline that exists before the deployment.

“It will be the beginning of AI, so everyone is trying to push and don’t do it well. That’s why we changed the way we do,” Anthony Pinto said. “We want to be sure we have the numbers before to start.”

That principle would govern the next AI investment from the first conversation onward.

02 Opportunity to change

Invest in the engineers. Then prove the investment.

DPS framed the AI investment commercially from the start. The engineers were what clients were paying for. Anything that made the engineers faster, better-equipped, and more consistent on a client interaction was an investment in the offer itself. AI in this context was not a deflection tool to keep customers away from humans. It was a capability layer designed to make the humans more valuable.

That commercial framing came with a discipline requirement. Before any AI feature went into production, the team would define the KPIs the deployment needed to move and capture the baseline against those KPIs in the existing operation. Average handle time. First call resolution. Knowledge article reuse. Agent eNPS. Customer CSAT and NPS delta pre and post deployment. Each metric was tied to a specific service-desk activity. Each baseline was owned by the delivery team, not the AI team alone.

The federated structure followed from the same logic. DPS established a central AI program at the group level to define common practices, with country-level champions adapting the implementation to local market requirements. Every asset the program produced (integration architecture, prompt engineering guidelines, quality monitoring playbooks, ROI measurement templates) was systematically captured in the team’s internal AI knowledge base. A new client deployment in any DPS market would not start from scratch. It would start from a validated baseline that the next client could see was already working.

The choice of platform followed directly from the measurement discipline. DPS was already standardized on NiCE CXone for service-desk operations, which meant the AI layer could be built on the live operation itself — the same environment where the baselines were captured and where the engineers actually work. Layering AI onto a separate platform would have severed that link: the deployment could no longer be measured against the operation it was meant to improve. Building on the contact center already in production kept the AI capability, the engineers, and the KPIs inside one operational reality. Time to value and the sovereignty posture European clients increasingly require both confirmed the same direction.

03 NiCE solution

What the engineers got. What the clients saw.

We have all in this same interface. I am with the client in voice or chat, and directly it’s connected to my knowledge base and gives me the right insight about how to solve the issue. The unified experience is really, really what they like.

Anthony Pinto

Head of AI,
Digital Platform Services,
Sopra Steria

NiCE Copilot for Agents went live across the DPS service desk integrated into the existing NiCE CXone footprint. The capability that did the heaviest lifting on day one was simple to describe and consequential in practice: every input the engineer needed during a client interaction was now accessible from inside a single panel adjacent to the conversation.

The unified workspace removed the pivots that had been costing time. Engineers stayed on the call, surfaced the right knowledge article inside the copilot pane, applied the procedure, and closed the interaction. The cognitive load of the prior workflow became something the engineer no longer had to think about.

Knowledge base integration: the capability behind the capability

NiCE Copilot for Agents only delivers its full value when it is wired into a real knowledge base, and the DPS team treated that as a precondition rather than an afterthought. The Copilot subscription was extended to NiCE Knowledge Management almost immediately. Its built-in connectors for ServiceNow, SharePoint, Confluence, and other ecosystems let DPS connect a new client’s existing knowledge environment to Copilot without rebuilding from scratch. When DPS signs a new client onto Copilot, the team can validate the connector against that client’s repository and start surfacing live, account-specific procedures inside the engineer’s panel within days rather than months. Every accelerated onboarding builds the case for the next one.

Automated Summary, customized per client

NiCE Copilot for Agents includes an Automated Summary capability that generates a structured wrap-up of every interaction. DPS extended that capability to a structural reality of MSP work: every client wants its case notes formatted differently, with different fields, different narrative structures, and different downstream automation requirements. Copilot lets DPS configure the summary structure per client, so an engineer working on a banking account in France produces a summary in that account’s format while an engineer on an industrial account in Poland produces a different one. Both happen automatically. Both happen inside the same workflow.

Co-creation, not hand-off

The way DPS positioned the partnership with NiCE was the framing that mattered most for the client conversation. When a client asked DPS what its AI capability was, the answer was not “we use NiCE CXone.” The answer was that NiCE and Sopra Steria together brought the value: NiCE supplying the technology and the configurable AI surface, DPS supplying the deep client context, the process knowledge, the integrations with each client’s existing ITSM and knowledge ecosystem, and the multilingual operational depth across three countries.

You bring the tech part. We bring the context, because we know our clients, we know their processes, we know how they want to work. All together, we bring the value.

Anthony Pinto

Head of AI,
Digital Platform Services,
Sopra Steria

Federated deployment across shared delivery centers France, Poland, and India.

DPS operates across three primary geographies with markedly different client profiles. France weighted toward financial services and public sector with strict sovereignty requirements. Poland with a heavier industrial mix. India delivering a meaningful share of the global service-desk capacity. The federated model held all of it together. Reusable assets, including chat scripts, voice scripts, copilot profiles, and Automated Summary configurations, were maintained in a shared repository and pulled forward into each new client deployment. The architectural posture is that any client requirement should have two answers ready: a state-of-the-art SaaS option, and an on-premise alternative for clients whose sovereignty posture requires it. The decision in any account is shaped by what the client needs.

04 Results

Faster engineers. A frontline that pulls. Clients asking to be next.

DPS now has the measurement framework the team built the discipline to capture. On standard service-desk tickets, average handle time has been reduced by 9%. First call resolution has improved by four percentage points. Both metrics were captured against baselines DPS recorded before any Copilot feature went live, and both were validated three months after deployment against the assumptions in the original business case. For a managed service provider whose engineers carry every commercial relationship, those numbers translate directly into a stronger offer to the next client.

“For handle time, we have reduced 9% on all our standard tickets. For first call resolution, we have won four points more than before. The requirement for FCR is really about the knowledge base, and how that knowledge base is structured,” Anthony Pinto said.

The qualitative result may be the more durable one. The Copilot rollout could have been the kind of program that triggered scrutiny inside DPS’s largest French service-desk delivery center, where any program touching frontline workflow is examined carefully for its implications on engineer roles. It did the opposite. Engineers at the site experienced the unified workspace as relief from a long-standing source of friction, and the response from the floor moved from neutral to active support.

Before, it was so difficult for them to operate different screens, switch between different solutions. They see what the advantage is for them in their day to day. We are happy. It’s well received from our Teams.

Anthony Pinto

Head of AI,
Digital Platform Services,
Sopra Steria

The most direct measure of frontline adoption arrived through a population DPS had not engineered for. A client still operating on Sopra Steria’s legacy contact center platform, one that had not yet been migrated to NiCE CXone and therefore did not yet include Copilot, surfaced agents asking openly when they would be moved over. The pull was on the tool, not against it. For an AI deployment, in a market where AI deployments routinely fail on adoption, that signal carries weight a CSAT delta cannot.

05 Future

Co-pilot today. Cognigy next. Autopilot after.

DPS sees the roadmap as a three-stage progression, and the team is already laying the foundation for the next stage. The current state is supervised co-pilot: AI surfacing information for the engineer, the engineer handling the interaction. The next stage is co-pilot agentic, where the copilot pane can directly trigger actions in client systems rather than only displaying procedures. DPS is committed in evaluation with NiCE on extending the Copilot surface with NiCE AI Agents (Cognigy) to make those actions executable inline. The stage after that is autopilot: AI agents handling the high-volume, low-complexity contact types, including ticket status checks, comment additions, and routine VPN access requests, autonomously, freeing the engineer population to focus on the issues that genuinely require human judgment.

The same KPI discipline that governed the current deployment will govern the next two. Each new capability will be defined against measurable outcomes, baselined before launch, and validated against the assumptions in its business case three months after go-live. The discipline that proved itself on Copilot is now part of how DPS evaluates every AI investment that comes after, including the ambition to scale Copilot at the group level so any country team picking it up inherits the validated blueprint from day one.

The technology is here. It works well. To be successful, you have to invest in change management on the people. That is where the project will struggle. If you bring a solution that is well integrated and user friendly, your teams will adopt it. If you build it yourself, you have to invest a lot more in change management. Either way, the people are the project.

Anthony Pinto

Head of AI,
Digital Platform Services,
Sopra Steria