ATAILA Newsroom · Budapest · 2026-08-20

IDC’s three numbers point at the same single gap. The market is not missing a portfolio — it is missing operations.

On 15 August the Hungarian IT trade site computertrends.hu presented Lenovo’s Hybrid AI Advantage offering. The article builds its case on Lenovo’s study CIO Playbook 2026: The Race for Enterprise AI, for which more than three thousand IT and business leaders answered IDC’s questions. The assessment stands, and the numbers speak for themselves. We read the same numbers — just not from the hardware side. In our view what the market is missing is not a portfolio, it is operations.

What the survey found

According to the research, enterprise AI has left the experimental phase — just not evenly.

On top of that, 96 percent of respondents plan more AI investment next year — 13 percent more on average. The reasons behind the hybrid preference, per the article, are business and financial on one side, data protection and cybersecurity on the other. On agentic AI the picture is inverted: next to the 21 percent with significant production use, 60 percent of respondents are still running pilots or examining use cases.

From these numbers the article derives Lenovo’s hybrid AI portfolio, which the vendor builds with partners — Red Hat, Intel, NVIDIA and ServiceNow among them — and in which private, on-premise AI features prominently. It also cites another survey: 90 percent of IT leaders admit their company’s defences against AI-powered attacks could be better.

We agree with the numbers, not the conclusion

The 62 percent is the article’s most important number, and its least surprising one. An organisation that would run its own contracts, patient records or customer financials through a model does not want local or hybrid deployment out of ideology — it wants it because the risk cannot be outsourced. We have been writing that same sentence since day one: private AI for the work you cannot put into the cloud.

The 46 percent is an honest number too — which is exactly why the other half deserves attention. Most pilots still never reach production; we wrote about the reasons in an earlier response to another market study. Our experience has not changed: what stands between a prototype and production is not model selection but platform questions — environments, releases, permissions, backups, accountability.

Where we part ways is the conclusion. The article treats the 27-percent governance shortfall as a missing framework to be placed next to the portfolio. We see it differently: governance is not a document, it is the point where a release either goes through or does not.

Everyone builds one column — the value is in the combination.

What ATAILA does with these gaps

The hybrid demand: a sovereign EU cloud, not rented iron

ATAILA Cloud is an operated sovereign environment running in the EU, where the private models run on our own hardware too. Behind it is not a souped-up VPS but the environment we operate our own systems on, day in and day out. For the leaders behind that 62 percent, the difference is that they are not buying GPU capacity but a working service — hardware, model, network and backups in one contract, at a fixed monthly fee, with no token-based surprises. The data, the model and the logs stay in one place, with a known operator.

The pilot gap: production is the starting point, not a phase

Pilots stall because production is scheduled for the end of the project, and that is where money and attention rarely remain. ATAILA Factory turns this around: we describe the application once, in a manifest, and the engine builds, releases and audits it in every environment. Four environments — sandbox, dev, uat, prod — are not an option but the default state. “Going live” stops being a separate project phase and becomes a repeatable step.

This is not theory for us: our open*.hu family of reference products runs in production on this platform, and its newest member went live in hours — by reusing what was already built.

The governance gap: governance lives where releases happen

The 27 percent does not grow because someone writes a policy. It grows when the rule is in the system: who may release, what went out, when, with what data, where to. On our platform the release path is audited, data stays inside the EU, and there is no American data processor in the path of a model call. On top of that comes the everyday side of operations — monitoring, backups, security patching, model lifecycle — committed to in writing. Together these do not replace a company’s AI policy, but without them the policy is unenforceable.

The client side: private AI for developers too

The article counts client devices into the hybrid model, and with good reason. In practice most confidential data does not leak from the server but from the developer’s machine — source code, production data samples, customer documents, pasted into a public assistant. ATAILA Studio is our answer: a managed developer workstation with private AI coding. The machine is in our hands — updates, backups and access management included.

Where we say no

We do not claim every workload belongs on-premise. The hybrid model won the survey because it is selective: some things are cheaper and better run in a public cloud. We answer for the part you cannot let out of your hands.

Nor do we claim a platform solves governance. Classification, retention rules and lines of responsibility come from the customer; what we can commit to is that they are technically enforceable and provable after the fact.

And we do not claim a long list of customer references. We have one customer case in public, anonymised at the customer’s request — our named references, for now, are our own products. We work for data-sensitive organisations of twenty to two hundred people, on one concrete, expensive, recurring workflow at a time. If you are looking for fifty regions and exabytes of storage, we are not the right answer.

If this is where you are

If you have a workflow you have long wanted to automate but would never paste into a public assistant, that is worth a conversation — confidential, no strings attached. We are not selling a demo: our first question is which step costs you the most today, and what data it works on. If we are not the right answer, we will say so.

Get started →