Artificial intelligence has moved from novelty to normal in the space of a few years. Predictive scheduling, automated status reporting, risk scoring, resource optimization — tools that touch portfolios, programs, and projects (PPPM) are already in daily use. What has been missing is an agreed, vendor-neutral answer to a simple question: how do we use AI responsibly in our profession? In 2026, PMI answered it with the first edition of The Standard for Artificial Intelligence in Portfolio, Program, and Project Management.
This article walks through what the standard actually says, how it is organized, and — from my own perspective as a PMP-certified practitioner — what it means for the way we work. It is an editorial overview, not a substitute for the standard itself.
What the standard is, and why now
The standard is a consensus document that sets out good practice for adopting, using, and integrating AI across the three PPPM levels. Its stated aim is ethical, efficient, and effective AI adoption that stays aligned with organizational values and applicable regulation.
Two design choices are worth flagging up front. First, it is deliberately technology-agnostic: it names no specific tools or vendors, and it explicitly encourages organizations to revisit the guidance as regulation and technology change. Second, it is human-centered. A recurring theme — the “human-in-the-loop” (HITL) approach — treats human oversight not merely as a safety brake but as a genuine source of value, since judgment, empathy, context, and institutional knowledge are things models do not hold.
The intended audience is broad: organizational leaders and sponsors, portfolio and program managers, project managers, and delivery teams including data scientists. The standard is framed as a supplement to PMI’s existing global standards for project, program, and portfolio management rather than a replacement for any of them.
The eight principles
The heart of the document is a set of eight principles, presented without ranking or weighting. Each is a value statement meant to shape decisions rather than prescribe steps.
- Strategic Value — the value-creation anchor. Every AI initiative, whatever its scale, should trace back to organizational goals. The practical test: start with the value case, not the technology. AI is a means to measurable benefit, not an end in itself.
- Risk — evolving risk adaptation. AI brings its own risk profile: algorithmic bias, data-privacy exposure, cybersecurity vulnerabilities, opaque model behavior. These need proactive, continuous management, embedded within existing governance rather than run as a side process.
- Governance and Compliance — responsible AI management. Clear structures, roles, authority, and guardrails define what “responsible use” looks like. The standard is explicit that AI governance should fold into the organization’s wider governance, risk, and compliance (GRC) machinery, not sit in a silo.
- People and Culture — innovation fosters empowerment. As AI reshapes roles, organizations need AI literacy, skill-building, and — a point I’d underline — psychological safety, so people can raise concerns and experiment without fear of blame.
- Ethics and Professional Responsibility — ethical AI stewardship. Transparency, fairness, explainability, auditability, and accountability run from the first design decision through the whole life cycle. Nondiscrimination and honest communication of a model’s limits are treated as professional duties.
- Stakeholder Engagement — connect, align, and deliver. Communicate the purpose and vision for AI early, keep the dialogue continuous, and treat feedback as a steering input. Notably, the standard expects you to engage stakeholders who oppose AI adoption, not just the enthusiasts.
- Optimization and Innovation — evolving through innovation. Continuous improvement backed by clear metrics and feedback loops. Automation frees teams from routine work so they can focus on higher-value activity, always with human judgment in the loop.
- Data Quality — input for impact. Models learn from their data, so inaccurate, biased, or incomplete inputs propagate straight into outputs and decisions. Robust data governance, defined stewardship roles, lineage and traceability, and regular audits are the foundation everything else rests on.
The five performance domains
Where the principles describe what to value, the performance domains describe where the work happens. Again, they are presented without a fixed order and are meant to operate together.
- Managing Stakeholder Expectations About AI. Identifying stakeholders, understanding their needs and concerns, and building consensus on how AI will be used and how AI-influenced decisions will be made — including agreement on data-privacy and governance policies.
- Defining the Scope for AI. Establishing and maintaining a clear vision and mission for AI initiatives, and the measurable activities that keep AI adding value across roles and functions rather than sprawling without purpose.
- Designing AI Architecture With Quality and Reliability. Making sure systems perform accurately and consistently against requirements (quality) and hold up dependably over time (reliability), underpinned by data governance and quality assurance.
- Executing Strategic AI Goals. A structured approach to delivery: strategic alignment, benefits realization, change management, governance, and proactive risk management working in concert.
- Managing AI Risks and Uncertainties. Identifying, assessing, and managing the dynamic risks AI introduces — and recognizing that these are as much opportunities as threats, and that legacy risk approaches may not fully cover them.
Chapters beyond these (life cycle management and tailoring, AI in the PPPM context, an implementation framework, and ethical and legal considerations) round out the standard, but the principles and domains are where most practitioners will start.
What this means in practice
If the architecture feels familiar, that is not an accident. PMI’s PMBOK Guide moved to a principle-and-domain structure with the 7th edition (12 principles, 8 performance domains), and the 8th edition refined that to 6 principles and 7 performance domains while explicitly emphasizing AI, sustainability, and tailoring. The AI standard slots neatly into that mental model: if you already think in principles and domains, you can absorb this without relearning your vocabulary.
From my own PMP perspective, the reassuring message is that the core competencies we already practice — strategic alignment, risk management, stakeholder engagement, governance — are exactly the foundation the standard builds on. AI does not replace project management judgment; it raises the stakes on it.
A few things I would watch for:
- Treating AI as an IT problem. The standard is emphatic that governance belongs in the broader GRC structure and that people and culture matter as much as models. Delegating it entirely to a technical team misses the point.
- Skipping the value case. “We should use AI” is not a business case. Start with the value the Strategic Value principle demands, or you will end up optimizing something nobody needed.
- Under-investing in data. Data Quality is a first-class principle for a reason. Poor inputs quietly undermine every downstream decision — and they are far harder to fix after the fact.
- Thinking HITL is just a rubber stamp. The standard frames human oversight as a value driver. Define real intervention triggers and escalation paths; do not just log that a human “reviewed” the output.
FAQ
Is this standard mandatory or a certification requirement?
No. It is a good-practice standard meant to be tailored to context. PMI notes it does not police or enforce compliance.
Does it replace the PMBOK Guide?
No. It supplements PMI’s existing standards for project, program, and portfolio management and is meant to be used alongside them.
Does it tell me which AI tools to buy?
No. It is deliberately technology-agnostic and recommends revisiting the guidance as tools and regulations evolve.
Who should read it first?
Anyone accountable for AI adoption — sponsors, PMO leaders, and portfolio/program managers — plus project managers introducing AI into delivery. The principles chapter is the most useful entry point.
By Tom, PMP-certified since 2004. Last updated: July 2026.