Qlik Is Right About the Hard Part of AI
Qlik has published research identifying what it calls “the hard part of AI” in enterprise deployments. The framing is self-serving — Qlik sells data integration and analytics software — but the underlying observation is accurate and underreported.
The hard part of AI is not the model. It is everything the model depends on.
Enterprise AI projects fail at a predictable set of choke points. Data that exists in the organization but is not accessible to the model. Data that is accessible but inconsistent, unlabeled, or structured in ways that the model cannot parse reliably. Outputs that are technically correct but operationally useless because they do not map to the decisions the organization actually makes. Governance requirements that make deployment legally or politically untenable. Integration costs that exceed the value of the capability being deployed.
These are not AI problems. They are data infrastructure problems, organizational problems, and workflow design problems that AI adoption makes visible. Most of them existed before anyone deployed a model. The model just makes them expensive.
The post-hype reckoning in enterprise AI is producing a useful clarification. The organizations that deployed AI fastest, with the least rigor, are now quantifying failure costs and recalibrating. The organizations that treated AI deployment as primarily a data and integration problem from the start are in a structurally better position — not because they moved slowly, but because they invested in the layer the model depends on.
Qlik’s commercial interest is transparent: it wants to sell the data infrastructure that sits under AI deployments. That does not make the diagnosis wrong. The market for “AI readiness” infrastructure — data pipelines, integration middleware, governance tooling, and observability — is real and underdeveloped relative to the market for model access. That gap closes next, with or without Qlik.