AI pilots are old news. Now, the focus is on processes and pricing
Artificial intelligence has already proven it works. That is no longer news to anyone following the enterprise tech market. The pilots were run, the results showed up, and excitement took over boardrooms around the world. But then came the hangover: okay, so what do we actually do with all of this?
Since generative AI arrived in late 2022, organizations of every size and sector dove into experiments with the technology. Customer service, software development, procurement, marketing, internal operations — practically every business function got some kind of pilot project involving language models or intelligent automation. Early efforts focused on proving the technology could handle specific tasks. Could it answer customer questions? Assist support agents? Generate content? Help employees work more efficiently?
The answer to all of those questions was yes. And that is exactly where the real problem started showing up. 🤔
The question nobody planned to answer
Last week, two companies from very different corners of the tech industry arrived at the same conclusion, each through its own path. Levelpath, a procurement platform that helps organizations manage purchasing, contracts, and vendor spend, and NiCE, which was gearing up for the annual Customer Contact Week, both asked the same question: what happens after an AI pilot succeeds?
It sounds like a simple question, but it is stalling entire organizations around the world. Scaling an AI project is not about flipping a switch and watching the results multiply on their own. It is a structural shift that touches people, workflows, responsibilities, budgets, and increasingly, the way contracts are negotiated and priced.
What emerged from both conversations is that a new set of questions is dominating corporate agendas. Who is accountable for the outcomes AI delivers? How should success be measured? How much visibility do organizations actually need into usage, costs, and performance? The answers to these questions could determine the future of AI adoption at scale.
The numbers that are scaring procurement leaders
Levelpath recently conducted a survey of enterprise software buyers, and the results are eye-opening. A full 57% of respondents reported encountering at least one AI-related spending issue in the past six months. The most common problem? Invoices coming in higher than what had been budgeted. Right behind that were teams hitting usage limits and organizations having to redirect budget from other priorities to cover AI costs.
These numbers put AI pricing at the center of a conversation that used to be predominantly technical. It does not matter if the model works flawlessly if the CFO cannot predict how much the whole thing is going to cost next month.
The most consistent thing we are hearing from procurement leaders, across industries and company sizes, is that the governance gap catches them off guard, said Stan Garber, president and co-founder of Levelpath.
That statement sums up the moment pretty well. Organizations were not prepared to manage AI as an ongoing, variable budget item. They treated the pilot as an isolated experiment, and when it came time to scale, they discovered they did not have the mechanisms in place to control costs, monitor usage, and ensure the investment was generating real returns.
Transparency became the priority — and spending caps took a back seat
One particularly interesting data point from the Levelpath survey shows how buyers are reacting to this new reality. Respondents were twice as likely to negotiate for greater transparency around AI usage and spending than to impose hard spending limits. In practice, 32% pushed vendors for more detailed usage reports, while only 16% implemented spending caps.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
This is significant because it signals a shift in approach. Instead of simply cutting, companies want to understand. They want to know exactly where the money is going, which features are consuming the most resources, which departments are using the most, and most importantly, whether that usage is delivering the value that was promised during the pilot phase.
Beyond transparency, buyers are also building flexibility into their contracts. The survey found that 39% added exit or transition clauses, while 36% shortened contract terms. These moves suggest organizations are reluctant to make long-term commitments while AI pricing models, vendors, and technologies keep evolving rapidly.
It makes sense. The landscape shifts every quarter. New models emerge, inference costs drop, open source alternatives gain traction. Signing a three-year contract in this environment is a risk few are willing to take without safeguards in place. 📊
The pilot that tested more than just the technology
Stan Garber from Levelpath made an observation worth highlighting. According to him, the organizations making the most progress are the ones that tested more than the technology during the pilot phase.
The leaders who are getting it right treated the pilot as a proof of concept for the operating model, not just the technology, Garber said.
This is a fundamental distinction. Most companies used the pilot to answer a simple question: does the AI work? The more prepared ones used the pilot to answer much more complex questions: how are we going to manage this day to day? Who will oversee it? How will we measure results? Which processes need to change? What will it cost at real scale?
When a pilot is treated as a proof of concept for the operating model, the transition to production happens much more smoothly. The governance, accountability, and cost challenges have already been identified — and in many cases, partially resolved — before the company commits budget and resources at scale.
When AI leaves the lab and enters customer operations
The governance challenges procurement leaders describe do not disappear once a contract is signed. They resurface — often amplified — when organizations start integrating AI into business processes and customer-facing operations. This is exactly where NiCE comes into the picture.
The company, which recently wrapped up its annual Interactions conference and was preparing for Customer Contact Week, reports that conversations with clients have shifted significantly over the past eighteen months.
Eighteen months ago, most customer conversations were anchored in pilots focused on point use cases like chatbots or agent assist, said Dan Belanger, president of NiCE Americas.
Belanger says the discussions now focus on how AI fits into customer service operations and broader business workflows.
Customers who are leading in this space have moved away from isolated deployments toward end-to-end orchestration. That means connecting customer intent, workflows, and resolution into a single system, Belanger added.
This shift runs deep. It is no longer about having a chatbot that answers frequently asked questions. Organizations now want to know if AI can authenticate a customer, update an order, process a refund, schedule a service visit, or resolve a complete issue — without creating additional work somewhere else in the system. AI has stopped being a side experiment and has become part of the actual operational flow.
The Fabletics case: from conversation to resolution
NiCE points to clients like Fabletics to illustrate this evolution. The activewear company moved beyond traditional rule-based bots that could answer questions but not complete tasks. Fabletics’ AI deployment now supports customer authentication, order management, and other workflows that allow interactions to move from conversation all the way through to full resolution.
We did not start this project to add AI to the contact center. We started it to give our customers faster, more flexible service, and to see what happens when AI can actually make decisions in real customer interactions, said Jack Roberts, senior global director of technology and GMS applications at Fabletics.
That statement is telling because it shows an inversion of priorities. Fabletics did not start with the technology — it started with the customer problem. AI was the means, not the end. And when a company adopts that mindset, the implementation tends to produce much more consistent results because it is grounded in real needs, not technological fascination.
The case also illustrates the kind of end-to-end orchestration Belanger mentioned. Instead of having AI function as a superficial layer answering generic questions, it is woven into the system in a way that lets it take concrete actions — authenticate, look up, modify, resolve. That is a massive difference in terms of value delivered and implementation complexity. ✨
The same challenge seen from different angles
What makes this story interesting is that procurement leaders and customer service leaders are approaching AI from completely different directions, yet both face the same reality: proving the technology works is easier than building the processes needed to manage it over time.
On the procurement side, the challenges are cost visibility, budget predictability, contract flexibility, and spend governance. On the customer service side, the challenges are integration with existing workflows, measuring outcomes, accountability when automation fails, and end-customer experience.
But at the core, all of these challenges are variations on the same theme: the organizational infrastructure that needs to exist for AI to truly work — not as an isolated innovation project, but as an integral part of business operations.
Pricing: the knot nobody wanted to untangle
If processes and governance are already complex topics, pricing is what is generating the most friction between AI vendors and their enterprise customers. During the pilot phase, billing models tend to be simple or even subsidized — after all, everyone wants the proof of concept to succeed. The problem shows up when it is time to sign a long-term contract for use at scale.
How many API calls are included in the plan? What happens when volume exceeds projections? How can the customer predict monthly costs if usage is variable? These questions sound basic, but they have been stalling entire negotiations. Both Levelpath and NiCE identified that the lack of predictability in billing models is one of the top brakes on adoption at scale.
The trend gaining momentum in the market is a shift toward value-based or outcome-based pricing models, instead of charges based on usage volume or per seat. Under this format, the company pays in proportion to the benefit AI delivers — whether that is reduced handling time, contracts processed, or cost savings generated. This aligns vendor and customer incentives in a much healthier way and makes it easier to justify the investment internally. It is not yet a universal model, but it is becoming increasingly common in renewal and expansion conversations. 💰
Governance is not bureaucracy — it is infrastructure
One point that deserves special attention is how governance is being perceived in this new chapter. For a long time, governance was treated as a synonym for bureaucracy — something that slows projects down and stifles innovation. But what the Levelpath data and the experiences shared by NiCE show is exactly the opposite.
Companies that invested in governance from the start of the pilot are scaling faster and with less pain. They know who is accountable for every automated decision, they have clear performance metrics, they can audit system behavior, and they have escalation mechanisms in place when something goes off track.
Meanwhile, the ones that treated governance as a detail to sort out later are finding out that later comes at a steep price. Decisions made by automated systems with no review mechanism, missing audit trails, unclear lines of responsibility — these problems do not show up in a pilot, but they blow up when scale arrives. And when they blow up, they tend to cost more than the project ever saved. 😬
Transparency also plays an important role within organizations themselves. Teams working with AI tools need to understand what the system is doing, what its limitations are, and when it might get things wrong. That does not mean every employee needs to understand the technical details of a large language model, but it does mean there needs to be an accessible layer of explainability, clear documentation, and open channels for reporting unexpected behavior.
What comes after the pilots
What Levelpath and NiCE are signaling, each from its own context and specific set of challenges, is that the enterprise artificial intelligence market is entering a new phase of maturity. The technology itself is no longer the main competitive differentiator. What will separate the companies that successfully scale AI from those stuck in eternal pilots is the ability to build solid process structures, invest in real governance, and have honest conversations about pricing and transparency.
This requires a mindset shift that goes beyond the technology team. Executive leadership, legal, finance, operations, and human resources all need a seat at the table when discussing how AI will function in day-to-day business. Not as a bureaucratic layer that blocks innovation, but as a framework that ensures innovation happens sustainably, responsibly, and with results that actually show up on the balance sheet.
Companies figuring this out now still have time to get organized. The market signals are clear, the data is on the table, and real-world examples — like Fabletics — show it is possible to move beyond the pilot successfully. But it takes planning, cross-functional dialogue, and a willingness to treat AI not as a tech project, but as a business transformation.
The pilot was just the beginning. The really interesting — and challenging — part is happening right now. 🚀
