Artificial intelligence is changing procurement at two speeds. It can compress work that once took days into minutes, helping teams analyze spend, review contracts, identify supplier concerns, and guide employees toward compliant purchases. At the same time, every new AI-enabled action creates another vulnerability where poor data or weak oversight affects a business decision.

That is the leadership challenge procurement now faces: accelerating the pace and improving the quality of work while maintaining accountability across every supplier interaction and enterprise transaction. That challenge is especially significant because procurement teams are also being asked to control costs, manage risk, strengthen resilience, and support broader digital transformation efforts.

The urgency is real. In research from the 2026 Economist Enterprise report titled “Procurement at a crossroads: from optimism to realism,” sponsored by SAP, 56% of executives identified AI strategy as the main catalyst for procurement’s digital agenda. The study, which covered 2,648 C-suite leaders, also recorded lower confidence in the function’s ability to translate technology investments into consistently better outcomes.

What those findings suggest is that the next phase of AI adoption is not about access to technology. It is about building the governance, accountability, and data foundations needed to turn potential into measurable business value. The practical question is where to start.

Start with the outcome, not the technology

AI programs often begin with finding the best possible tool for a problem. Procurement leaders should reverse that sequence and focus on the desired outcome.

The first questions they should ask are, “What outcomes would impact the broader business, and what decision or workflow needs to improve?” A sourcing team may need to shorten event preparation. A category manager may need earlier warning of price or supply changes. A purchasing organization may want to reduce off-contract buying. Each objective carries different data requirements, risk levels, and measures of success.

Defining the outcome first forces the question early, before deployment choices narrow your options. Leaders can specify which actions AI may complete, which recommendations require review, and which decisions must remain under human control. They can also set escalation rules for exceptions involving sensitive data, high-value commitments, supplier concentration, or regulatory obligations.

This turns governance from a final approval step into part of the operating design.

Build a connected data foundation

AI cannot provide dependable guidance when supplier records, contract terms, spend information, and risk signals are fragmented across systems. More importantly, an agent cannot safely execute work without the context that accompanies this information. Procurement needs unified data governance that covers common definitions, data ownership, access controls, and traceable sources. Without it, AI operates on assumptions rather than facts.

Perfection is not a realistic prerequisite, and waiting for it will stall progress. But organizations should be explicit about uncertainty. When information is incomplete, the system should surface that limitation or route the matter to a person rather than present an assumption as fact. As the Economist Enterprise research highlights, fragmented data remains one of the most significant barriers to realizing AI’s potential in procurement, and it is a barrier that governance can address.

Apply human oversight where it matters most

The right balance between human and AI involvement varies depending on the procurement activity. Routine, rules-based work can support greater automation, while strategic supplier decisions require a different standard.

The Economist Enterprise research shows that fewer than one in 10 respondents would give AI the lead across most procurement choices within three years. By contrast, 46% expect the technology to assist with tactical work, while people retain authority over strategic matters.

That balance reflects something procurement practitioners understand from experience. Data can indicate that a supplier offers favorable terms or strong performance. It cannot tell you whether that supplier will collaborate during a disruption, bring you new ideas before they to a competitor, or treat your business as a priority when capacity is tight. Those judgments depend on relationships, commercial context, and years of accumulated experience that no system fully captures.

As organizations adopt similar tools and draw from increasingly comparable data, the real source of differentiation will be how procurement leaders interpret those outputs and apply judgment.

Human review should therefore be concentrated where the consequences are greatest, not added indiscriminately to every automated step. Clear thresholds can protect control without recreating the delays AI is intended to remove.

Measure value and risk together

Responsible adoption will not scale through policy alone. Employees need to understand how AI changes their work, when to challenge an output, and who is accountable for the final decision. Procurement, finance, IT, legal, and operations also need a shared view of ownership before something goes wrong rather than after.

Metrics to determine success and failure should be defined before deployment. Cycle-time reduction, contract compliance, spend under management, user adoption, supplier performance, and risk response can all show whether a use case is working. These measures should be paired with indicators such as exception rates, human overrides, data-quality failures, and control breaches.

This matters because AI can produce visible efficiency without improving the decisions that matter most. In the same research, many executives reported that AI has yet to meaningfully improve procurement decision-making quality despite growing investment in the technology. That gap is the real opportunity.

Procurement leaders who link AI to specific outcomes, build connected data, assign clear decision rights, and prepare their teams to work alongside the technology will move beyond isolated automation toward something more durable. The organizations that get this right will not simply be the ones that automate the fastest. They will be the ones where AI amplifies judgement, relationships, and experience that procurement professionals have always brought to the table, and where accountability for the decisions that matter most remains firmly in human hands. 

Gordon Donovan is vice president of Research for Procurement and External Workforce at SAP.

For audiences, Cirque du Soleil is about wonder: gravity-defying performers, breathtaking costumes, immersive music, and moments that feel almost impossible. But behind every performance is something just as remarkable: a highly complex global business operation.

It reinvented circus arts and became a world leader in live entertainment, performing for more than 365 million spectators in 90 countries. Today, the organization operates 38 shows in cities around the world, supported by over 4,000 touring artists and staff from more than 80 countries.

Each touring show functions like a moving profit center. A production may operate in Mexico City, then move to London, then Paris—bringing with it new companies, employees, assets, tax requirements, local regulations, merchandising operations, and artists from dozens of nationalities.

As Philippe Lalumière, vice president of Information Technology at Cirque du Soleil, explains, “People underestimate the complexity of running a circus.”

Orchestrating an autonomous accounts payable process

This is especially true for the company’s accounts payable (AP) department. Every year, more than 70,000 invoices flow through Cirque du Soleil, non-stop, 24/7. Around 40% of vendor inquiries are standard requests for invoice status. But answering those questions was far from simple. AP specialists had to search across systems, review invoice histories, understand payment status, determine the cause of delays, and manually respond.

The work was repetitive, time-consuming, and emotionally draining. “From a morale point of view, receiving e-mails from suppliers, some of them a bit impolite because they’re asking, ‘When are we going to get paid?’—it’s not a fun job,” Lalumière says.

Finding the right problem to solve—putting AI into action

As an SAP customer for more than 25 years, Cirque du Soleil knows firsthand how to leverage SAP technology not only to run its operations, but also to reinvent them.

Cirque du Soleil, known as an early adopter and leader in digital transformation, was approached by SAP AppHaus with a question: how could SAP generative AI technology be used to improve your business processes?

The answer emerged through collaborative workshops involving SAP, Cirque du Soleil’s IT team, and business users across its departments. Accounts payable quickly rose to the top due to its lean structure, high volume of interactions, and clear automation potential. And the choice aligned with the broader business evolution—AP automation is one of the most widely adopted AI use cases across industries.

With a clear opportunity identified, Cirque du Soleil and SAP moved from ideation to execution, developing an AI-powered solution that automated AP processes and improved responsiveness.

AI enters stage right—from vision to reality

The workshops led to something more than an automation project—they led to Genato, a multilingual AI agent that now works alongside the AP team.

The polyglot agent scans the AP inbox, analyzes sentiment, identifies urgency, extracts invoice numbers from messages and attachments, connects to SAP data to retrieve invoice status, and drafts a response for human review. If it cannot find an invoice or resolve an inquiry, it flags the issue.

“Trust in the quality of the answers coming from the agent was an initial concern,” Lalumière says. “But, the team quickly realized that Genato’s information was spot on.”

Together, Genato and the AP team now serve Cirque du Soleil’s diverse global supplier network more efficiently.

Lalumière, however, is clear about one thing: “Yes, AI is a very powerful tool, but it’s not pixie dust. Sprinkling AI into processes is not going to solve everything. There is work involved.”

That work included designing appropriate connectors and integrating with the company’s SAP and AP systems using SAP Business AI Platform. This enterprise AI foundation can bring together capabilities and technologies (think SAP Business Data Cloud, Business Transformation Management solutions, and SAP Business AI), unifying AI, data, process context, and governance, so customers can build, integrate, scale, and run AI that delivers business impact while working to ensure the solution is sustainable and cost-efficient.

The impact was evident almost immediately. Generative AI began prioritizing urgent supplier requests, retrieving invoice information, drafting responses, and translating communications automatically, delivering measurable improvements across the AP organization:

• 97.92% improvement in handling priority requests
• 25% reduction in AP backlog
• 25% faster response times for non-urgent inquiries
• Improved vendor satisfaction and employee morale

The impact on employees cannot be overstated. By removing repetitive research and emotionally charged supplier follow-up from the team’s workload, Genato has become more than a tool. “We now include Genato as a virtual team member,” Lalumière explains, “It’s a paradigm shift.”

That adoption is the clearest sign of success. Lalumière recalls that after the prototype moved into production, one of the end users made her feelings clear: “No, no, no, you cannot take it away from me! There is no way I’m going to live without it now.” Today, she is one of the solution’s biggest ambassadors.

Collectively, the solution demonstrates how AI can be introduced into a focused business process, quickly earn user trust, and create a blueprint for broader enterprise innovation.

As Lalumière puts it, “You could say we’re an SAP shop.”

The next act of AI innovation

For Cirque du Soleil, accounts payable is only the beginning.

Lalumière believes that “AI is a huge and beautiful tool, but you still need human judgment to prioritize use cases and move forward.” He also sees AI evolving across three layers: personal productivity with solutions such as Joule; AI applied to business processes, such as Genato; and, eventually, AI embedded in the audience experience itself. “I think we’re at the dawn of a new era of circus arts,” he says.

His advice to others beginning their AI journey is simple: start small, involve users early, and be willing to experiment. “Try a small proof of concept and be ready to throw them away,” Lalumière says. “The AI train is moving, and you have to hop on.”

For a company built on reinvention, transforming accounts payable may seem far removed from the spotlight, but for Lalumière, the principle is the same: innovation happens when people are willing to rethink what’s possible. Today, that mindset is improving back-office operations. Tomorrow, it may help redefine the audience’s experience.

Sid Misra is CMO of SAP Business AI Platform Technology Foundation.
Top image courtesy of Cirque du Soleil.

Many organizations are currently undergoing an SAP S/4HANA transformation, and some are already investing in AI technology. But how do new technologies actually work in day-to-day operations?

Between technical deployment and actual adoption, there is often a gap. What is needed goes beyond tools—it requires enablement, communication, and organizational change management that puts people first. Experiences from SAP customers show how this can succeed.

Global transformation, local anchoring as a guiding principle

Dulaan Punsag-Odefey cites a number that immediately makes the challenge tangible: 265. That is how many key users serve as change ambassadors at Hapag-Lloyd, distributed across six regions worldwide. The shipping company is one of the five largest in the world and is in the midst of an SAP S/4HANA Finance transformation. The “Fast Forward” program affects 4,000 SAP users in more than 140 countries.

Punsag-Odefey, organizational change management (OCM) lead at Hapag-Lloyd, explains: “Change management is anchored in our program as a strategic enabler. Not as an add-on, but as a core element.”

In practice, this means the 265 ambassadors translate global standards into local language and local practice. Controllers are expected to evolve into business partners who no longer just produce Excel spreadsheets but deliver decision-ready recommendations for sales and operations. The message to the workforce is “It will be different, but better.”

How a federal agency demonstrates effective change management

That transformation can succeed without following the textbook is demonstrated by the Bundesanstalt für Post und Telekommunikation (BAnst PT), Germany’s Federal Agency for Post and Telecommunications. The agency implemented SAP S/4HANA in just one year: greenfield, public cloud, 1,000 affected users, go-live on January 1, 2026. In parallel, the organization moved to a new administration building with a new-work concept.

Simone Kunze, specialist in the SAP Service Center at BAnst PT, knows the phrases that come up in every organization: “Is there an official directive for this?” or “Standard won’t work for us.” What helped was honest communication and direct moderation within business units instead of token feedback sessions.

A fit-to-standard approach replaced legacy custom solutions. Key users were developed from project experts who had already built depth through workshops, user stories, and test cases. The investment in support paid off: after go-live, dozens of thank you e-mails came in regarding the new SAP travel expense management and Fiori apps, and survey response rates exceeded 70%.

Thilo Menges from the Medical University of Lusatia (MuL) takes this one step further. His project has a unique starting point: with 3.6 billion euros in funding, an entirely new organization is being built from scratch, including SAP technologies. Menges makes a point that many organizations do not state this clearly: “For me, change management is an investment protection measure. This is an organizational project, and people need support in change processes.”

Change management in his project accounts for 4.3% of the total budget—the largest single line item in the SAP contract.

SAP’s change management framework with its six dimensions, which MuL follows, is integrated into the SAP Activate methodology. Particularly important are early assessments, target-group-specific communication, and the identification of trusted multipliers rather than a blanket approach.

What was also highlighted: learning does not end at go-live. Tools like WalkMe enable context-sensitive support in the flow of work, especially for infrequent processes. Enabling the organization to independently maintain and evolve these systems is critical for sustained success.

The organizations mentioned above are supported by change management consultants from SAP. More information is available here.

From shadow AI to structured integration at KIT

While BAnst PT and Hapag-Lloyd are primarily transforming SAP system landscapes, the Karlsruhe Institute of Technology (KIT) faces a different question: how do you get 25,000 students and 10,000 employees to use AI responsibly, instead of each person experimenting on their own in the shadows?

In 18 months, KIT made the journey from uncontrolled AI usage to an AI toolbox with governance rules. Rather than issuing bans, KIT focuses on enablement.

Andreas Sexauer from the Center for Technology-Enhanced Learning at KIT describes the approach as follows: a mandatory qualification module covers foundational knowledge and legal aspects before students and faculty gain access to the AI toolbox. In parallel, use case workshops run across departments, from leadership teams to the legal department. After the teaching rollout in April 2025, 31 didactic chatbots were created in the first seven days. Faculty configure them directly in the learning management system for their respective courses.

KIT also takes a pragmatic approach to costs: after initially providing free access, a budget cap per person per month was introduced.

What research confirms

Whether shipping company, federal agency, or university, all customers report similar patterns. Prof. Dr. Renate Osterchrist from the Munich University of Applied Sciences provides the scientific foundation: she analyzed 119 studies on the effectiveness of change interventions, examining six intervention fields: communication, support, involvement, reinforcement, social influence, and coercion. Key effectiveness factors in change include:

• Dialogue formats are more effective than one-way communication.
• Coaching for managers improves not only their leadership capability but also measurably enhances implementation competence. Coaching and peer exchange are also very beneficial for employees.
• The dimension of coercion had been under-researched until now: clarity in messaging about what behavior is expected proves helpful. Manipulation and political maneuvering, on the other hand, reduce commitment.
• The frequently cited claim that 70% of all change projects fail is not supported by current data.
• The statement “Honestly, I have never experienced a change where there was too much communication” further underscores the important role of communication.

When AI agents enter the picture, change becomes even more critical

What happens when not only new systems are introduced but AI agents take over parts of the work? This is precisely the question that arose when Joule Studio 2.0 was demonstrated live at the forum. With this solution, SAP customers can create AI agents that access their business context: SAP Knowledge Graph, process models, and domain knowledge. The agents are code-based and transparent. Developers describe the desired outcome in natural language; Joule Studio can generate the specification and executable code. Both no-code and pro-code approaches are possible.

The discussions that followed amongst SAP customers and partners made clear that the change management described above will be essential here. When agents take over tasks, roles change, responsibilities shift, and the way humans and machines collaborate is transformed.

A full documentation of the SAP Learning and Adoption Forum 2026, including videos, slides, and a chatbot, is available on SAP Community.

Thomas Jenewein is business development manager for AI, Transformation, & Adoption Services at SAP.