← All briefs

Who has to be ready for AI in a project?

5 min read

AI is now used across project delivery, in scheduling, estimating, risk analysis, design coordination and project documentation. Research has followed quickly, and the question it keeps returning to is readiness: what has to be in place before AI can be used well? In a paper for the AUBEA 2026 conference, my PhD student Ibrahim Dani, Jingxiao Zhang of Chang'an University, Liang Ma of Shanghai University and I ask a question that comes before that one. Ready by whom?

Why the question matters

Readiness has to belong to something. In the recent project management literature, it is attached in turn to teams, individuals, workflows, stakeholders, organisations and projects, sometimes within a single framework. A 2025 framework by Armanios and Tucci in the International Journal of Project Management names six kinds of readiness, from team readiness to ecological readiness. The project appears only as the setting where readiness is needed. Four responses followed in the same journal within nine months. Each moved readiness somewhere else, and three argued that readiness is not enough. None asked whose readiness it is.

This matters for two reasons. If readiness has no fixed owner, it cannot be assessed or compared, because there is nothing stable to assess. And most large projects are delivered by several firms working under contract. Each firm may be well prepared while the arrangement between them is not. Checking readiness firm by firm will then overstate what the project can actually do.

What we reviewed

We took every article with artificial intelligence, AI or GenAI in its title published since 2020 in four project management journals: International Journal of Project Management, Project Management Journal, International Journal of Managing Projects in Business and Project Leadership and Society. That gave 46 articles up to August 2026, 39 of them published in 2025 or 2026. We read each in full and asked three things: who or what is treated as ready, what the word "project" refers to, and whether the conditions of multi-firm delivery appear at all.

What we found

Readiness is placed almost everywhere except the project. Seventeen articles attach readiness, capability or maturity to the firm, eight to individuals and four to teams. Only one article takes the project as its unit of analysis, and it is about stakeholder roles, not readiness. No article treats the project as something that can be ready or unready.

The word "project" appears throughout but rarely means an organisation. In 25 articles it is the setting where work happens. In 12 it is a bundle of processes to coordinate or automate, in five an outcome to measure, and in three a level in a hierarchy. Only once is it an organisation with a state of its own.

The conditions of multi-firm delivery are mostly missing. Thirty-five articles do not mention them, nine describe them without building them into their explanation, and two theorise them. One study deliberately sampled only projects that did not involve multiple organisations, because contracts and ethical concerns made AI capability vary across them. The very variation a project-level view would explain was treated as noise.

Project AI readiness

Every firm in a project can be ready for AI while the project itself is not.

We define project AI readiness as the preparedness of a project, constituted as a temporary multi-firm organisation, to deploy and operate AI in pursuit of project-specific objectives within its lifecycle. Three features set it apart.

  1. It is held in the arrangement, not in the firms. It lies in the terms on which the parties share data, in who carries liability for errors that come from AI, and in who has the authority to act on an AI output. It is not the sum or the average of each firm's readiness.
  2. Its scope is set by organisational form, not sector. It applies wherever a project is delivered by contractually separate parties with a finite life. It does not apply to a project run wholly within one firm, where existing firm-level ideas are adequate.
  3. It is distinct from its neighbours. AI capability, maturity models, AI literacy and organisational AI readiness all belong to firms, standing organisations or individuals. Project AI adoption, its companion idea, describes a process; readiness describes a state.

We agree with critics that readiness alone does not explain what AI achieves in projects. We do not think that is a reason to move beyond readiness. It is a reason to be clear about whose readiness it is.

What it means for practice

A project's information is held across organisational boundaries, and its authority to act is shared among independent parties. Readiness assessed one firm at a time will therefore overstate what the project can do with AI. Clients and delivery partners should assess the arrangement itself, at procurement and again at mobilisation. That means looking at how data will be shared, who carries liability for AI-derived errors, and who can accept an AI output on behalf of others.

What the paper does not settle

This is a conceptual paper built on a bounded review. It specifies the construct but does not yet measure it. The review covers article titles only, four journals and one database, and it captures a fast-moving debate at a fixed date. Three pieces of work follow:

Reference