By Guest contributor |
Every operating leader I speak with is holding onto two truths. Clients now expect service that feels faster, more personal and more anticipatory than anything they experienced before. At the same time, the costs associated with that experience keep growing: adoption, data governance, security operations, and now AI token consumption that scales with every workflow you automate. Delight without cost discipline becomes a margin problem. Cost discipline without delight becomes a retention problem. The operator’s job is to design for both in the same motion, and that design work must happen before the first AI agent reaches a customer-facing process.
Coretek’s answer is to be our own first customer. Before an approach goes to market, we live with it inside our own operation. We invested more than 4,000 engineering hours building a unified, governed data estate: a production Microsoft Fabric lakehouse bringing together seven core enterprise data sources, deployed on Microsoft Foundry through infrastructure as code. Across fiscal year 2026 we moved 24 agents into production in eight departments, with 22 more in build and 47 in the pipeline. Agent identity is live, which means we have literal agents in our org chart, each with an owner, a defined scope and accountability, governed through a CEO-led AI governance and centre-of-excellence cadence.
The internal numbers gave us the confidence to scale: roughly 10,000 hours a year reclaimed, close to six full-time equivalents of capacity returned to higher-value work, and about $1.1 million in modelled annual return on investment. The more valuable output, though, was the hard-won operating knowledge. We learned where token costs spike and how to architect around them, which data sources need remediation before an agent is ever pointed at them, and how to sequence a security review so it accelerates a project rather than stalling it at the end. Clients feel that as a better project experience. Fewer surprises, fewer rebuilds, a realistic view of run-rate cost, and a governance model handed over rather than invented mid-flight.
For many of our clients the entry point is Microsoft 365 Copilot, with Agent 365 giving leaders a way to manage agents as first-class identities alongside their people. That path is fast, and it is the right answer when the work lives in mail, meetings, documents and Microsoft Teams. It is also where the trust conversation starts, because Copilot inherits whatever permission sprawl and unlabelled content already exists in the tenant. Getting data governance and identity right first is what turns a promising pilot into something an executive is willing to put in front of customers.
Just as often, the business need does not fit the productivity surface. A supply chain workflow, claims process, field service dispatch or revenue operations model needs a purpose-built application. Here we work in Azure AI Foundry, and our AI Factory approach takes the use cases one at a time: assess readiness, prioritise by value and feasibility, then determine whether the right answer is Copilot, a custom Foundry solution, or simply better automation with no model involved. Taking those use cases one by one keeps token and infrastructure spend attached to a measurable outcome, and it stops companies from buying a platform in search of a problem.
Security is the layer that makes speed possible. As an Azure Expert Managed Services Provider and a member of the Microsoft Intelligent Security Association, we build AI on the same foundation we defend around the clock through our 24/7, year-round security operation. Application development, data governance and managed security are not sequential phases in our model. They are the same conversation, held early.
Managing change in the workplace is, in the end, a trust exercise. People adopt an agent when they know who owns it, what data it can see and who is watching it. Put agents in the org chart, give them the same accountability as a new hire, and adoption stops being a campaign and starts being how the work gets done. That is when the customer finally feels it.
Michael Swiencki is chief operating officer at Coretek
Discover more insights from experts at Microsoft Partner businesses in the Autumn 2026 issue of Technology Record. To get content like this on a regular basis, subscribe to the print edition or free digital edition of the quarterly Technology Record magazine.