The conversation around enterprise AI is shifting. For much of the past two years, organizations have focused on pilots – testing generative AI tools, exploring use cases, and proving value. Now, the focus is turning to production.
As businesses begin to operationalise platforms such as Microsoft Azure AI, Azure OpenAI Service, and Copilot, a new set of challenges is emerging that goes far beyond experimentation.
“Our customers are no longer asking what AI can do,” Microsoft noted in a recent partner blog. “They are asking how to deploy it securely, govern it consistently, and generate measurable business outcomes.”
That shift is exposing structural gaps across the Microsoft partner ecosystem.
For partners on the ground, the transition from proof of concept to production is proving far from straightforward.
“AI pilots work if you can demonstrate value to the users, the departments, the business, and ultimately the P&L,” said Rupert Davey, managing director at CTM IT. “Problems come in moving to production when security and governance controls are introduced, and these are seen as classic cybersecurity blockers.”.
This experience is widely reflected in the market. Research from Omdia shows enterprises are running dozens of AI pilots simultaneously, but conversion to production remains inconsistent, with many organizations seeing only a small proportion progress beyond experimentation.
IDC research suggests up to 90% of AI pilots fail to reach production, highlighting a significant gap between experimentation and operational deployment..
While early pilots often operate in relatively unconstrained environments, production deployments must integrate with enterprise identity, compliance, and security frameworks, which introduces a new layer of operational complexity.
At software marketplace Pax8, Kristen Fehrenbach, vice president of global Microsoft alliance, said readiness varies across the MSP landscape.
“MSPs in the Pax8 ecosystem are moving quickly from experimentation to execution, but readiness varies – those making the most progress are the ones building repeatable, managed approaches rather than treating AI as a one-off project,” she said
As AI systems move into production, requirements around governance, lifecycle management, and cost control are becoming unavoidable.
“The biggest operational friction shows up in turning working demos into governed, secure, supportable services. Cost predictability is also a pressure point as AI usage grows,” said Fehrenbach.
Unlike traditional software deployments, AI workloads are typically consumption-based, introducing new financial dynamics and new challenges around cost management.
“The ramp up of adoption means closely monitored PAYG options are key, and once adoption plateaus, then review and look for prepaid options,” said Davey.
These requirements are forcing a rethink of the traditional Microsoft partner model. Historically, many partners have focused on licensing, migrations and implementation projects. But AI is pushing the channel toward ongoing operations.
“Users at businesses are quick to find their own gen AI tooling in the wild,” said Davey.
“So if Microsoft partners don’t get in front of this with managed services… businesses will organically move to an uncontrolled set of heterogeneous generative AI tooling, risking the farm on democratised technology selection.”
Fehrenbach believes AI is already shifting where value is created. “Partners who can run AI as an ongoing managed operation will be better positioned than those limited to resale or one-time projects,” she said.
New skills, new responsibilities
Supporting production AI environments requires new capabilities, many of which sit at the intersection of cloud and AI operations. These include governance by design, lifecycle management, and the ability to deploy and manage agentic workflows.
For smaller partners, the shift is particularly significant.
“For smaller, more agile Microsoft partners, like us, grassroots adoption and AI enablement services are going to be key to getting from pilot to production. This includes training and coaching with feedback loops from monitoring usage. This will be new to a lot of partners who’ve evolved from a break/fix mentality,” explained Davey.
Microsoft is also pushing partner skilling as a priority. “In an AI-first world, skilling is the new currency,” said the company.
As the ecosystem evolves, a divide may be emerging between partners that can support AI in production and those that cannot.
“Partners who can run AI as an ongoing managed operation will be better positioned,” said Fehrenbach.
But Davey suggests the implications could go further. “If AI, specifically agents, replace bums-on-seats… who, or what, does the Microsoft partner support? The Microsoft partner ecosystem needs to embrace the idea that we support identities, not people. In the future, agent support will be as important as user support,” he said.
Looking ahead, AI is expected to reshape the Microsoft channel around ongoing operations, governance, and outcomes rather than one-off deployments.
“AI will push the Microsoft partner ecosystem toward partners delivering intelligence-driven outcomes, where governance, adoption, and lifecycle management become as central as licensing and security are today,” said Fehrenbach.
For partners, this means the era of AI experimentation is ending, and what comes next is operational. Ultimately, for a channel built on projects and resale, this represents both a challenge and a significant opportunity.